DOI: 10.3897/jucs.160204

Authors: Abdulhadi Altherwi, Md. Mottahir Alam, Mastoor M. Abushaega, Ahmed Hamzi, Abdulmajeed Azyabi, Shabbir Hassan, Asif Irshad Khan

Abstract: The integration of hybrid renewable energy systems (HRES) has introduced both opportunities and challenges in managing multisource power systems such as wind and solar. Accurate forecasting of HRES performance is critical to efficient planning and grid stability. This paper proposes a data efficient hybrid framework that combines Grey Wolf Optimization (GWO) for feature selection with Deep Belief Networks (DBN) for predictive modeling. GWO effectively selects relevant features from high dimensional environmental and system parameters, reducing computational burden and enhancing learning performance. The DBN is then trained on the optimized input set to forecast system performance. Two public datasets capturing wind and solar power production across distinct geographic conditions were used for validation. The proposed model significantly outperforms conventional methods, achieving a mean square error of 0.0207, RMSE of 0.144, and an energy efficiency of 98.32%. These results demonstrate the framework’s potential for deployment in smart grid forecasting environments.

HTML

PDF

DOI: 10.3897/jucs.152399

Authors: Abdulaziz Shehab, Abdelhady Naguib

Abstract: The importance of localization algorithms is due to their uses in various wireless sensor network applications. A single anchor movement can be used to aid in localization to reduce the cost of using multiple anchors or equipping sensor nodes with GPS units, but the main challenge here is choosing the best path of movement while avoiding potential obstacles. This paper proposes a path planning algorithm called Square Spiral with Obstacle Avoidance (SQSPOA) which allows a mobile anchor node to track an optimal path while broadcasting its current coordinates to the unknown sensor nodes. During its movement, the mobile anchor node faces many obstacles that may hinder its mobility; but as a result of the superiority of the proposed algorithm the mobile anchor can avoid these obstacles while still broadcasting its coordinates to sensor nodes. The performance of the proposed algorithm was evaluated at the presence of variable-sized obstacles and was compared with recent path planning algorithms. Simulation results proved the superiority of the proposed algorithm with respect to localization error, percent of localized sensor node and trajectory length.

HTML

PDF

DOI: 10.3897/jucs.130529

Authors: Moulay Youssef Ichahane, Noureddine Assad, Hassan Ouahmane

Abstract: The diagnosis of rheumatoid arthritis (RA) diagnosis demands precise detection methods due to its complex symptomatology. This study presents a novel hybrid diagnostic framework that is the first to integrates Case-Based Reasoning (CBR) with deep learning and introduce three key innovations: (i) a dual-pathway architecture that combine electronic health records with imaging analysis, (ii) an Enhanced Clustering-Based K-nearest neighbors (ECB KNN) model for optimal feature selection, and (iii) a dynamic K-means clustering approach for handling class imbalance. We evaluated our framework using two comprehensive datasets: MIMIC-IV-Hosp, containing clinical data and MIMIC-CXR containing 377,110 chest X-ray images. The model employs a VGG16-based CNN for radiological feature extraction, with a particular focus on pulmonary manifestations, which is combined with our ECB KNN algorithm for patient-specific clinical data analysis. Using five-fold cross-validation, our framework is shown to achieve superior performance metrics (precision: 0.90-0.95, recall: 0.89-0.93, F1-score: 0.91) compared to baseline methods (traditional CNN: precision 0.82, recall 0.79; standard CBR: precision 0.85, recall 0.83). This significant improvement in diagnostic accuracy demonstrates the potential of our framework in terms of enhancing early RA detection and clinical decision support. The architecture of the model architecture is designed to allow for extensibility to other rheumatic conditions, thereby offering a comprehensive solution for multi-disease diagnosis in rheumatology.

HTML

PDF

DOI: 10.3897/jucs.156960

Authors: Anuradha Mahasinghe, Ayodhya Liyanage

Abstract: A quantum Turing machine (QTM) is an abstract model of computing that serves as the quantum counterpart of a classical Turing machine. Closely related to the probabilistic Turing machine, a QTM utilises quantum effects such as superposition, entanglement and unitary evolution. Despite its historical role as a framework for devising quantum algorithms and the significance of its classical counterpart in theoretical computing, very little attention has been paid in literature to the state transition of QTM’s for basic quantum gates. In this paper, we construct the state transition diagrams for a set of elementary quantum gates that consists of the Hadamard, CNOT, and T gates, providing a universal basis for fault tolerant quantum computation. We verify the necessary conditions to ensure that the designed state transition diagrams comply with the postulates of quantum mechanics, verifying their well formedness.

HTML

PDF

DOI: 10.3897/jucs.145075

Authors: Jorge Arthur Schneider Aranda, Ricardo dos Santos Costa, Vitor Werner de Vargas, Paulo Ricardo da Silva Pereira, Jorge Luis Victória Barbosa, Marcelo Pinto Vianna, Eleandro Luis Marques da Silva

Abstract: Power utilities demand large volumes of data used in power distribution networks. Among them are parameters representing possible technical failures, such as network’s short circuit current and voltage sag. Specialists find these parameters and detect technical failures. However, this process can become time-consuming. Thus, this article proposes an ontology called OntoFreya, which classifies voltage, current, or any electric metric, following the definitions of the regulatory agencies and reducing the time spent on this task. A series of 4402 axioms, 132 classes, and 40 data properties comprises OntoFreya. The ontology automatically inferred classifications for four hundred readings from energy samples, validating OntoFreya across three scenarios. The first and second scenarios classified current in amperes, and the third classified voltage in per-unit system (pu). The scenarios showed that OntoFreya automates the classification of electric metrics, reducing specialists’ time in detecting technical failures in a distribution network.

HTML

PDF

DOI: 10.3897/jucs.139707

Authors: Chaimae Moumouh, José A. García-Berná, Begoña Moros, Juan M. Carrillo de Gea, Mohamed Yassin Chkouri, José L. Fernández-Alemán

Abstract: The impact of technology on improving health and well-being of individuals is remarkable. EHealth boosts the transition from paper-based health records to Electronic Health Records (EHRs). The use of EHRs can lead to improve quality of care, costs and time. In eHealth systems the health data is stored in digital form, and can be exchanged or accessed securely by authorised users. It is worth noting that medical data is considered very confidential information. However, the privacy and security of medical data remains a critical issue. Any leak or breach in security can lead to serious privacy damages for patients. Despite the safeguards, training courses and the consciousness on keeping data safe, the human error continues to be a problem. The main purpose of this paper is to present a bibliometric overview on the academic research related to privacy and security in EHRs. For this purpose, the papers of this study were searched in the Scopus. A period of 24 years was considered for selecting the papers. The information gathered in the database identified a total of 3,077 publications. Some key findings revealed that in the year 2015 the highest number of publications was produced. The Harvard Medical School was the most prolific institution with 2.44% papers from the total number of publications. A total of 97.21% of the documents were written in English. Finally, the results provided in this manuscript allowed us to make a picture on the current relevance in academic literature on privacy and security in EHRs.

HTML

PDF

DOI: 10.3897/jucs.192427

Authors: Christian Gütl

Abstract: Dear Readers,It gives me great pleasure to announce the third regular issue of 2026. I would like to thank all the authors for their sound research and the editorial board for the extremely valuable reviews and suggestions for improvement. These contributions together with the support of the community and the generous support of the KOALA initiative enable us to run our journal and maintain its quality.I would still like to expand our editorial board: If you are a tenured associate professor or above with a good publication record, please apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and emerging trends.In this regular issue, I am very pleased to present 6 accepted papers by 27 authors from 6 countries: Brazil, India, Kingdom of Saudi Arabia, Morocco, Spain and Sri Lanka.In a collaborative effort between researchers from Morocco and Spain, Chaimae Moumouh, José A. García-Berná, Begoña Moros, Juan M. Carrillo de Gea, Mohamed Y. Chkouri, and José L. Fernández-Alemán address in their paper the critical challenges of privacy and security in Electronic Health Records (EHRs) by presenting a comprehensive bibliometric analysis of academic research in this field based on 3,077 publications indexed in Scopus over a 24-year period. The findings identify major publication trends, leading institutions, dominant languages, and the year with the highest research output, highlighting the growing academic relevance of EHR privacy and security and providing a structured overview of the field’s development.Jorge Arthur Schneider Aranda, Ricardo dos Santos Costa, Vitor Werner de Vargas, Paulo Ricardo da Silva Pereira, Jorge Luis Victória Barbosa, Marcelo Pinto Vianna, and Eleandro Luis Marques da Silva from Brazil research in their work the challenge of efficiently classifying electrical metrics in power distribution networks by proposing OntoFreya, an ontology-based model that applies semantic reasoning to interpret voltage, current, and contextual data. The results demonstrate that OntoFreya enables precise and scalable automatic classification, reducing specialist analysis effort while supporting context-aware inference across large datasets.Ayodhya Liyanage and Anuradha Mahasinghe from Sri Lanka investigate in their paper the lack of well‑posed state transition diagrams for basic quantum gates in the standard Quantum Turing Machine model by constructing rigorous diagrams for a universal quantum gate set. The results demonstrate that these diagrams satisfy the postulates of quantum mechanics, thereby providing a universal, fault‑tolerant basis for simulating quantum computations within the QTM framework.Moulay Youssef Ichahane, Noureddine Assad, and Hassan Ouahmane from Morocco present in their work a multimodal diagnostic framework that combines case-based reasoning with deep learning to address the complexity and heterogeneity of rheumatoid arthritis diagnosis, integrating electronic health record data with deep learning–based analysis of chest X-ray images. Experimental evaluations show that the proposed approach significantly improves diagnostic accuracy and robustness compared to conventional CNN- and KNN-based methods, highlighting the framework’s relevance for advanced computer-aided medical diagnosis.Abdelhady Naguib and Abdulaziz Shehab from Saudi Arabia tackle in their article the problem of robust and energy-efficient localization in obstacle-rich wireless sensor networks by introducing a deterministic mobile anchor trajectory model, that integrates square spiral coverage with lightweight obstacle avoidance. A range of simulations demonstrate that the proposed approach achieves superior localization accuracy, higher node coverage, and reduced trajectory length compared to state-of-the-art path planning schemes.In a collaborative research between Saudi Arabia and India, Abdulhadi Altherwi, Md. Mottahir Alam, Mastoor M. Abushaega, Abdulmajeed Azyabi, Ahmed Hamzi, Shabbir Hassan and Asif Irshad Khan research the challenge of accurate and computationally efficient forecasting in Hybrid Renewable Energy Systems by proposing a hybrid Grey Wolf Optimization–Deep Belief Network (GWO-DBN) framework that integrates metaheuristic feature selection with deep learning. Validation on two real-world datasets demonstrates reduced prediction error and computational time, achieving high forecasting accuracy and improved energy efficiency for smart grid applications.Enjoy Reading!Warm regards,Christian Gütl, Managing Editor-in-ChiefGraz University of Technology, Graz, Austria

HTML

PDF

DOI: 10.3897/jucs.157024

Authors: Mashael M. Alsulami, Kholoud Althobaiti, Haneen Algethami

Abstract: Job recommendation systems play a critical role in matching individuals with relevant career opportunities based on their skills and experiences. However, many existing systems struggle to balance precision and contextual relevance, leading to mismatches in job recommendations. In this paper we introduce JobMatcher, a multilayered recommendation system that integrates a well established technique, cosine similarity and KNN clustering with ChatGPT based evaluation. Initial recommendations are generated through content-based filtering and refined via clustering similar job descriptions aligned with user profiles by seniority and trajectory. To enhance contextual accuracy, GPT 3.5 turbo was prompted to act as an expert evaluator, scoring top recommendations based on skill relevance and career fit using structured and unbiased prompts. In a user study with seven domain experts and ten user profiles, system-selected jobs scored significantly higher (mean = 3.43 compared to 2.99 for KNN clustering, p = 0.0035), with moderate inter-rater agreement (Kendall’s W = 0.417). JobMatcher bridges algorithmic filtering with human like evaluation, offering a scalable, intelligent solution for improved job matching.

HTML

PDF

DOI: 10.3897/jucs.156911

Authors: Gamze Cabadag, Ali Degirmenci, Omer Karal

Abstract: Fast Fourier transform (FFT) is a widely used method for frequency estimation in electronic support systems. However, when the intermediate frequency (IF) of the radar signal is not an exact multiple of the FFT resolution, the correct frequency value cannot be obtained in the FFT computation. Therefore, interpolation methods are used to improve the frequency obtained from the FFT result. In this study, 12 different interpolation techniques (Jain, corrected Jain, Quinn, improved Quinn, Jacobsen, Macleod, Ding, Voglewede, mobile industrial (MI), Candan, rectangular-window-based interpolation, and Hanning window based interpolation) used in the literature have been extensively analyzed on radar signals contaminated with Laplace and Gaussian noise at different SNR values. In addition, in order to observe the performance of the techniques in different frequency bands, the bandwidth was changed to between 100 and 1000 MHz, and 100 Monte Carlo simulations were applied for each frequency. From the experimental analysis results, the improved Quinn technique showed the best performance for both noises. In addition to accuracy evaluations, the computational complexity of each interpolation technique was analyzed in terms of floating-point operations (FLOPs). The FLOPs cost of the FFT was uniformly included in all methods to ensure fair comparison. Results showed that while all techniques operate within a similar computational range, methods like Jain and Candan exhibit lower FLOPs costs, whereas the improved Quinn method, despite its higher complexity, achieves the best estimation accuracy.

HTML

PDF

DOI: 10.3897/jucs.141277

Authors: Kausthav Pratim Kalita, Debojit Boro, Dhruba Kumar Bhattacharyya

Abstract: Purpose: Big data applications enable organizations to derive actionable insights that inform strategic decision making and enhance operational efficiency in real time. Hadoop’s architecture features a distributed file system that stores voluminous data across multiple machines within a cluster. However, the management and control of access to this data can be viewed as centralized, as Hadoop relies on a central coordination system to manage tasks and resources across the cluster.Design / methodology / approach: To address the limitation in Hadoop, this paper proposes integrating blockchain technology to establish strict authentication procedures through smart contracts, enabling controlled access to the Hadoop platform. The proposed platform allows organizations to access Hadoop clusters through participation in a blockchain network, enabling efficient data storage mechanisms and model training capabilities.Findings: The performance of this integrated system is evaluated through simulations leveraging Ethereum based smart contracts. The findings suggest that implementing appropriate indexing mechanisms and hashing techniques can enable sufficient access control, thereby facilitating controlled access to Hadoop clusters. The paper presents the simulation results in terms of execution cost and execution time.Originality/value: This paper addresses the identified need for a transparent and reliable access control system that leverages blockchain’s smart contracts to enable controlled and restricted access to Hadoop clusters.

HTML

PDF

DOI: 10.3897/jucs.140733

Authors: Hanan Hafiz, Maher Alharby

Abstract: The rapid adoption of cloud computing has revolutionized how businesses and consumers access and utilize resources, offering scalability, flexibility, and cost effectiveness. However, this increased reliance on cloud services has also led to a rise in Distributed Denial of Service (DDoS) attacks, which can severely impact the availability and performance of these services. This study aims to address the critical need for effective detection and classification of DDoS attacks in cloud environments using machine learning techniques. We conducted binary and multiclass classification experiments using the CICDDoS2019 dataset, focusing on three specific types of attacks. Four machine learning models, namely Random Forest, K Nearest Neighbor, Naïve Bayes, and Logistic Regression, were implemented in a Kaggle notebook using Python. Feature selection techniques, including Chi square and Principal Component Analysis, were employed to identify the most relevant features, while the oversampling technique was used to handle imbalanced data. The experiments yielded impressive results, with Random Forest and K-Nearest Neighbor achieving the highest accuracy rates of 100% and 99.72% in binary classification, and 100% and 99.66% in multiclass classification, respectively. The study also measured training and testing times, along with other performance metrics. These findings highlight the effectiveness of machine learning approaches in tackling cloud based detection challenges while ensuring computational efficiency tailored for dynamic cloud environments.

HTML

PDF

DOI: 10.3897/jucs.154672

Authors: Aleksandar Velinov, Aleksandra Mileva, Simon Volpert, Sebastian Zillien, Steffen Wendzel

Abstract: Network steganography has existed for several decades and it uses network traffic and network protocols as carriers for embedding secret messages in a stealthy manner. Quick UDP Internet Connections (QUIC) is a novel secure and reliable transport layer network protocol that is encapsulated in the User Datagram Protocol (UDP) and utilizes the Transport Layer Security Version 1.3 (TLSv1.3) standard. In addition, Hypertext Transfer Protocol Version 3 (HTTP/3) employs QUIC. In this paper, we present a systematic analysis of the covert channels that can be found in QUIC. Twenty novel covert channels are identified by applying the latest covert channel pattern based taxonomy, and an analysis of their transmission rate, undetectability, and robustness is presented, together with suggested countermeasures. A single covert channel is implemented as a proof of concept tool and is appropriately evaluated.

HTML

PDF

DOI: 10.3897/jucs.147898

Authors: Gustavo Lazarotto Schroeder, Wesllei Felipe Heckler, Rosemary Francisco, Jorge Luis Victória Barbosa

Abstract: Smartphone use has increased globally and has become essential in daily life. Although benefits exist, concerns arise about the negative effects of prolonged hyperconnectivity. The excessive use of smartphones combined with demographic and mental health related risk factors can lead to problematic smartphone use (PSU). PSU is characterized as the compulsive use of smartphones that disrupts an individual’s daily life, work, and relationships. Considering this scenario, the present paper proposes OntoKratos as an ontology designed to detect and prevent PSU. The ontology enables inferences, such as determining the individual’s mental health and PSU state, inferring context information, identifying PSU demographic and emotional risk factors, and suggesting interventions. OntoKratos includes 89 classes, 43 object properties, 35 data properties, and 1,113 axioms. Evaluations performed through a simulated dataset demonstrated the ontology’s effectiveness regarding PSU identification and interventions for PSU behaviors. Ontology’s rules allowed the definition of accurate axioms, improving the correct classification and inference of eight instantiated individuals. This study presents the first ontology for PSU identification and intervention suggestions on PSU behaviors. OntoKratos allows to identify and assist individuals by considering mental health and PSU status, inferring potential PSU risk factors, and providing tailored intervention suggestions to cope with PSU.

HTML

PDF

DOI: 10.3897/jucs.189356

Authors: Christian Gütl

Abstract: Dear Readers,It gives me great pleasure to announce the second regular issue of 2026. I would like to thank all the authors for their sound research and the editorial board and guest reviewers for the extremely valuable reviews and suggestions for improvement. These contributions together with the support of the community enable us to run our journal and maintain its quality. I would still like to expand our editorial board: If you are a tenured associate professor or above with a good publication record, please apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and emerging trends. In this regular issue, I am very pleased to present 6 accepted papers by 20 authors from 6 countries: Brazil, Germany, India, North Macedonia, Saudi Arabia, Türkiye. Gustavo Lazarotto Schroeder, Wesllei Felipe Heckler, Rosemary Francisco, and Jorge Luis Victória Barbosa from Brazil address in their manuscript the growing problem of problematic smartphone use (PSU) by proposing OntoKratos, an ontology-based approach that models contextual, demographic, and mental health information to identify PSU and recommend personalized interventions through semantic reasoning. The research contributes a formal and reusable ontology with SWRL-based inference mechanisms, demonstrating through simulated data that OntoKratos effectively classifies PSU states, infers risk factors, and generates evidence-based intervention suggestions. In a collaborative research between colleagues from North Macedonia and Germany, Aleksandar Velinov, Aleksandra Mileva, Simon Volpert, Sebastian Zillien, and Steffen Wendzel look into the steganographic analysis of different network protocols which becomes a necessary part of their security evaluation, to prevent their abuse as carriers of hidden messages. In this manuscript, twenty novel covert channels are identified in QUIC, with an accent on their transmission rate, undetectability, and robustness, suggested countermeasures, and one implemented covert channel as a proof-of-concept.Hanan Hafiz and Maher Alharby from Saudi Arabia introduce in their work a study that aims to develop efficient machine learning models for detecting DDoS attacks in cloud environments by addressing challenges related to multi-tenant traffic patterns and virtualized infrastructure constraints. The main contributions of this study include binary and multiclass DDoS classification with feature selection, evaluation of model performance and computational efficiency, and mitigation of data imbalance using oversampling techniques.Kausthav Pratim Kalita, Debojit Boro, and Dhruba Kumar Bhattacharyya from India investigate in their research the issue that big data platforms face limitations in centralized access control despite their distributed architecture and propose integrating blockchain technology using smart contracts to enable secure and controlled access to cluster resources. Through Ethereum-based simulations, the study demonstrates that appropriate indexing and hashing mechanisms can effectively enforce access control while maintaining acceptable execution cost and execution time.Gamze Cabadag, Ali Degirmenci, and Omer Karal from Türkiye research in their work FFT-based radar frequency estimation errors arising from non-integer FFT bin alignment and evaluate twelve interpolation techniques under Gaussian and Laplace noise over varying SNRs and bandwidths. Monte Carlo analyses combined with FLOPs-based complexity evaluation show that the improved Quinn method achieves the highest estimation accuracy for both noise types, while simpler methods offer lower computational cost with reduced performance.Last but not least, Mashael M. Alsulami, Kholoud Althobaiti and Haneen Algethami from Saudi Arabia address in their paper the limitation of traditional job recommendation systems by introducing JobMatcher, a multi-layered framework that combines content-based filtering-KNN, and large language model–based evaluation to better capture career context and progression. The findings show that utilizing ChatGPT as a refinement layer improves alignment with expert judgments, resulting in more relevant and realistic job recommendations.Enjoy Reading!Best regards,Christian Gütl, Managing Editor-in-ChiefGraz University of Technology, Graz, Austria

HTML

PDF

DOI: 10.3897/jucs.148927

Authors: İbrahim Çakırlar, Sevcan Emek, Şebnem Bora, Oğuz Dikenelli

Abstract: This study introduces a novel methodology and framework for the verification, validation, and testing of agent-based simulation models: RatKit. Building on repeatable automated testing in ABMS, the present contribution significantly extends the foundation by proposing an integrated metamodel and systematic development methodology that embeds these activities throughout the simulation lifecycle. The RatKit methodology is both general, in that it applies to a wide range of agent-based simulation models using a well-defined metamodel, and comprehensive, in that it addresses the macro-level (societal), the meso-level (interaction) and the micro-level (agent) aspects of simulations. It also provides a generic infrastructure to be able to support various VV&T techniques. RatKit is designed as a general VV&T framework for all ABM frameworks. The methodology comes with a dedicated implemented framework. It is implemented by selecting the Repast ABM development framework. RatKit is demonstrated through a detailed case study of the Boids model, where the dynamics of alignment, cohesion, and separation are examined. Results from the case study show that a test-driven approach can enhance model reliability and ensure that individual agent behaviors coalesce into realistic emergent phenomena. Experiences and feedback obtained during the case studies show that developing ABM with a test-driven method based on VV&T facilitates the creation of desired models.

HTML

PDF

DOI: 10.3897/jucs.146099

Authors: Hamdullah Karamollaoğlu, İbrahim Yücedağ, İbrahim Alper Doğru, Sinan Toklu, İsmail Atacak

Abstract: The rapid expansion of IoT devices has transformed industries while simultaneously introducing critical security vulnerabilities, particularly Distributed Denial-of-Service (DDoS) attacks that exploit the constrained resources of IoT systems. To address this challenge, a novel intrusion detection system (CBM-IDS) is proposed for the effective identification and mitigation of DDoS attacks in IoT environments. A hybrid deep learning framework is employed, integrating Convolutional Neural Networks (CNN) for spatial feature extraction, Bidirectional Long Short-Term Memory (BiLSTM) for temporal dependency analysis, and a Multi-Head Attention Mechanism (MHAM) to prioritize critical network traffic patterns. Model robustness is enhanced through Adaptive Synthetic Sampling (ADASYN) and One-Sided Selection (OSS) for class imbalance mitigation, along with dimensionality reduction using an Autoencoder combined with ANOVA F-test-based feature selection. The proposed system is evaluated on the CICDDoS2019 benchmark dataset, achieving a detection accuracy of 99.93%, which demonstrates its efficacy in real-world IoT security applications.

HTML

PDF

DOI: 10.3897/jucs.142860

Authors: Banani Ghose, Zeenat Rehena

Abstract: Due to the advancement of industrialization and urbanization, air pollution become a serious issue in recent decades. To get rid of this problem Air Quality Monitoring Stations (AQMSs) are established that can asses the air quality and provide some measures to control it. These AQMSs capture the time series data through sensors and form an IoT based network to send the data to the cloud for further analysis. These IoT devices are paired with low capacity batteries and limited memory, transmission, and computational components. Transmitting the high volume data to the cloud for successive analytical purposes demands high energy dissipation and higher bandwidth. Moreover, for storing the big volumed data the system needs larger storage space. The single solution for all these constraints is data compression. Data compression reduces the volume of the datasets. The reduced volume of data saves energy and it is easier to transmit the compressed data through the limited bandwidth. In this paper, a lossless data compression algorithm using Genetic Algorithms is proposed. On successful implementation of the algorithm, the air quality time series data is compressed with absolutely no data loss. To evaluate the efficiency of the algorithm, its performance is compared with some classical and several state of the art compression schemes. The experimental results show that the proposed compression algorithm outperforms the classical as well as state of the art models concerning the performance evaluating parameters like Compression Ratio, Compression Factor, and most importantly Power Saving.

HTML

PDF

DOI: 10.3897/jucs.140979

Authors: Kadir Ileri, M. Şamil Balcı, Adem Dalcalı

Abstract: Energy harvesting is an effective solution, especially in scenarios with low power requirements, using sources such as magnetic fields, vibrations, and wind. This study focuses on predicting harvested power of toroidal electromagnetic energy harvesters using various machine learning methods, including Least Absolute Shrinkage and Selection Operator (Lasso), Adaptive Boosting (AdaBoost), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest, Gradient Boosting Machine (GBM), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM). To enhance the performance of these models, Artificial Bee Colony (ABC) optimization has been applied. The experiments were conducted using 1,300 trials across seven toroidal cores with varying sizes and magnetic permeabilities. During each experiment, the line current was varied between 0–100 A, and the resulting induced voltage and current were recorded. These measurements were used to create a comprehensive dataset named the Toroidal-Energy-Harvesting Dataset, enabling accurate power prediction. The performance of the machine learning models was assessed using statistical metrics, including R², MSE, MAE, and RMSE. Among the evaluated models, the ABC-optimized XGBoost (ABC-XGBoost) demonstrated the highest performance, achieving an R² value of 0.9993, an MSE of 247.1, an MAE of 9.8, and an RMSE of 15.7, indicating superior accuracy and minimal error. The comparative analysis clearly shows that proposed ABC-XGBoost outperformed the other models, making it the most effective solution for accurate power prediction in the Toroidal-Energy-Harvesting Dataset.

HTML

PDF

DOI: 10.3897/jucs.137000

Authors: Jinghan Liu, Hui Zhao, Chenyang Lin, Dan Wang, Shufan Li

Abstract: The current security problem of smart contracts is becoming a common concern for researchers and developers. Existing smart contract vulnerability detection methods rely heavily on fixed expert rules, resulting in low detection accuracy. In order to cope with complex and changing smart contract application scenarios, we chose to use graph neural networks to detect vulnerabilities. In this paper, we proposed a vulnerability detection model called ESA based on the enhanced sequential algorithm. During the coding process, the contract function source code is described as a contract graph, which increases the model’s global insight into node features during the learning process and reduces the number of noise nodes unrelated to vulnerabilities while retaining sufficient contextual semantic features. Compared to the cutting-edge methods, our model has significantly improved the accuracy of reentrant and timestamp dependency vulnerabilities, with detection accuracies of 89.09% and 88.49%, respectively.

HTML

PDF

DOI: 10.3897/jucs.150679

Authors: Stefan Lengauer, Lin Shao, Hossein Miri, Michael Bedek, Cordula Kupfer, Maria Zangl, Bettina Kubicek, Barbara Dienstbier, Klaus Jeitler, Cornelia Krenn, Thomas Semlitsch, Carolin Zipp, Dietrich Albert, Andrea Siebenhofer, Tobias Schreck

Abstract: The effective and targeted provision of health information to consumers, specifically tailored to their needs and preferences, is indispensable in healthcare. With access to appropriate health information and adequate understanding, consumers are more likely to make informed and healthy decisions, become more proficient in recognizing symptoms, and potentially experience improvements in the prevention or treatment of their medical conditions. Most of today’s health information, however, is provided in the form of static documents. In this paper, we present a novel and innovative visual health information system based on adaptive document visualizations. Depending on the users’ information needs and preferences, the system can display its content with document visualization techniques at different levels of detail, aggregation, and visual granularity. Users can navigate using content organization along sections or automatically computed topics, and choose abstractions from full texts to word clouds. Our first contribution is a formative user study which demonstrated that the implemented document visualizations offer several advantages over traditional forms of document exploration. Informed from that, we identified a number of crucial aspects for further system development. Our second contribution is the introduction of an interaction provenance visualization which allows users to inspect which content, in which representation, and in which order has been received. We show how this allows to analyze different document exploration and navigation patterns, useful for automatic adaptation and recommendation functions. We also define a baseline taxonomy for adapting the document presentations which can, in principle, be leveraged by the observed user patterns. The interaction provenance view, furthermore, allows users to reflect on their exploration and inform future usage of the system.

HTML

PDF

DOI: 10.3897/jucs.185149

Authors: Christian Gütl

Abstract: Dear Readers,First of all, I would like to wish you all the best for the New Year! It is with great pleasure that I welcome you to our first regular issue in 2026.Looking back on the past year, we have further increased our visibility and taken steps to fully comply with the Diamond Open Content Standard and align our journal with the KOALA requirements. Thanks to the combined efforts of the Pensoft team and the J.UCS publishing team, we are listed and indexed in more than 40 indexing services worldwide, including DOAJ, Web of Science, and Scopus. The increased visibility and presence on social media have also led to a further increase in page views and article downloads.With around 340,000 unique views, reader interest in J.UCS publications increased by almost 200% compared to the previous year. We can also look back on an increasing number of submitted articles and special issue proposals. This interest is reflected in a notable impact factor with a Scopus Cite Score of 2.5 and a Web of Science Journal Impact Factor of 0.9, both metrics have improved slightly since last year. We proudly look back on a total of 14 issues – 12 regular and 2 special issues – with 70 articles by 247 authors from 34 countries on novel aspects of various topics in computer science. The acceptance rate has fallen to below 6 per cent.These great achievements were only possible thanks to the commitment and interest of the community, the valuable support of the Editorial Board and the financial supporters of J.UCS through the KOALA Computer Science Cluster from TIB - Leibniz Information Centre for Science and Technology University Library, Germany.In 2025, we welcomed 10 new members to the Editorial Board, bringing our total number of Editorial Board members to 222. We would also like to gratefully acknowledge the support of 78 guest reviewers over the past year.In particular, I would like to thank Dr. Ulrike Krießmann from the Library of the Graz University of Technology, the TIB - Leibniz Information Centre for Science and Technology University Library in Germany, and the Institute of Human-Centred Computing (HCC) from TU Graz for their financial support.I would also like to thank the J.UCS team, Johanna Zeisberg for taking care of the publication process, David Kerschbaumer for his social media support, and Sebastian Gürtl and Alexander Nussbaumer for their technical support, as well as Pensoft Publishers Ltd. for hosting our journal.I look forward to continuing to work with our editors, editorial team and technical support to maintain the success of J.UCS. I would be very grateful for suggestions and feedback on how we can improve and develop J.UCS in the future. We also greatly appreciate the generous support of the J.UCS community, especially in promoting the journal and citing relevant articles in their research papers.In this regular issue, I am very pleased to present 6 accepted articles by 34 authors from 8 different countries, namely Austria, China, France, Germany, India, Thailand, Türkiye, United Kingdom.In a collaborative research effort from several research institutions from Austria, Thailand and Germany, Stefan Lengauer, Lin Shao, Hossein Miri, Michael Bedek, Cordula Kupfer, Maria Zangl, Bettina Kubicek, Barbara Dienstbier, Klaus Jeitler, Cornelia Krenn, Thomas Semlitsch, Carolin Zipp, Dietrich Albert, Andrea Siebenhofer, and Tobias Schreck present an advanced and innovative visual health information system - A+CHIS - based on adaptive document visualizations. Depending on the users’ information needs and preferences, the system displays its content at different levels of detail, aggregation, and visual granularity, addressing individual needs and preferences.Jinghan Liu, Hui Zhao, Chenyang Lin, Dan Wang, and Shufan Li from China address in their research the issue that existing smart contract vulnerability detection methods overly rely on expert rules and struggle to adapt to complex application scenarios. The article proposes a vulnerability detection model called ESA based on the enhanced sequential algorithm. Experimental results demonstrate that ESA significantly outperforms cutting-edge methods, achieving detection accuracies of 89.09% for reentrancy vulnerabilities and 88.49% for timestamp dependency vulnerabilities.Kadir Ileri, M. Şamil Balcı, and Adem Dalcalı from Türkiye focus in their research on predicting the harvested power of toroidal electromagnetic energy harvesters using machine learning models optimized with the Artificial Bee Colony algorithm. The findings show that the ABC-optimized XGBoost model provides highly accurate and robust predictions, outperforming the other evaluated approaches.Banani Ghose and Zeenat Rehena from India propose in their research work a Genetic Algorithm-based lossless data compression technique to reduce the size of the time-series data while conserving energy of the sensor-based system. The proposed lossless data compression technique can not only compress the air quality time series data with a high compression ratio, but it is also energy efficient, both contributing finally towards greater efficiency and longer lifetime of the sensor network.Hamdullah Karamollaoğlu, İbrahim Yücedağ, İbrahim Alper Doğru, Sinan Toklu and İsmail Atacak from Türkiye cover in their study the critical security challenge of DDoS attacks in resource-constrained IoT environments by proposing a novel hybrid deep learning model (CBM-IDS) that integrates Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and a Multi-Head Attention Mechanism for robust intrusion detection. The proposed model, evaluated on the CICDDoS2019 benchmark dataset and enhanced through advanced feature reduction and data balancing techniques, achieved a detection accuracy of 99.93%, demonstrating its significant potential for real-world IoT security applications.In a collaborative research effort between France and Türkiye, İbrahim Çakırlar, Sevcan Emek, Şebnem Bora, and Oğuz Dikenelli introduce RatKit, an advanced framework and methodology for verification, validation and testing of agent-based simulation models, which is based on the Boids model. The findings demonstrate that a test-driven approach can enhance model reliability and ensure that individual agent behaviors coalesce into realistic emergent phenomena.Enjoy Reading!Best wishes, Christian Gütl, Managing Editor-in-Chief Graz University of Technology, Graz, Austria

HTML

PDF

DOI: 10.3897/jucs.154610

Authors: Claudio Gutiérrez-Soto, Marco A. Palomino, Patricio Galdames, Cristian Duran-Faundez

Abstract: The growth of online learning in higher education, particularly after the COVID-19 pandemic, has fostered the advancement of learning analytics, which nowadays relies greatly on capturing and mining data derived from systems such as Blackboard and Moodle. However, it remains difficult to identify all the variables having a direct bearing on academic success, and drawing advice from machine learning models trained to support data-driven decision making is challenging. Therefore, we have endeavoured to pair a descriptive model, which characterises the profiles of computer science students, with a predictive model, which relies on Bayesian networks to forecast academic success. To achieve this, we have looked for the factors directly influencing the academic performance of computing science students, and the common patterns of behaviour which characterise higher education students individually and as part of a cohort. Our approach has been tested with data provided by a Chilean institution—University of Bío-Bío. We have enhanced and supplemented the data employed in our investigation by means of two surveys distributed among all the different cohorts of the student population. Our predictive model can determine student outcomes with an accuracy rate above 97%.

HTML

PDF

DOI: 10.3897/jucs.153217

Authors: Przemyslaw Jatkiewicz

Abstract: The study aims to analyse the vulnerabilities of traditional access control methods and define optimization objectives, constraints, and decision-making processes based on data for the effective implementation of artificial intelligence to enhance corporate data protection. The research methodology addressed various approaches, including machine learning, user behaviour analysis and neural networks, and data protection methods such as anonymisation, encryption and federated learning. Traditional access control methods, such as passwords, biometrics and multi-factor authentication, were discussed, as well as their shortcomings, including vulnerability to data breaches, phishing attacks and infrastructure threats. The use of artificial intelligence to strengthen access control mechanisms, such as machine learning, user behaviour analysis and neural networks, was emphasised. Artificial intelligence significantly improves security by enabling the analysis and processing of large amounts of data, detecting anomalies and predicting threats based on the analysis of user behaviour and biometric data. The study also examined methods of protecting data used to train artificial intelligence, including anonymisation, differential privacy, encryption and federated learning. Privacy issues the risks of data leakage when using artificial intelligence and the need to comply with ethical norms and standards were addressed. The successful integration of AI-oriented solutions into corporate security systems in various industries, including the financial sector, healthcare, and retail, is presented. Evaluating the effectiveness of artificial intelligence in access control systems is based on indicators such as the speed of the system’s response to changes in user behaviour, the number of false positives and successfully prevented incidents. The study also developed recommendations for improving access control mechanisms using artificial intelligence, including the introduction of machine learning-based systems to detect anomalies in user behaviour, and the integration of AI with multi-factor authentication to create flexible and reliable data protection mechanisms.

HTML

PDF

DOI: 10.3897/jucs.150393

Authors: Alexandru Pintea

Abstract: Smart homes, live streaming IoT devices, and smart sensors can all be optimised to enhance energy efficiency. In order to offer a cheap alternative to the traditional real-time monitoring systems, this study proposes a sensor-based occupancy system. The evaluation in real time of the number of occupants in buildings/ rooms /houses is reflected in the energy usage. Sensor data can provide insight into many characteristics of a considered environment. The sensor dataset considered was collected with the aim of determining how many people are present in a given space/room. The sensor data does not portray the people present in the room, but rather their impact on it (e.g. CO2/ noise/ light level changes). The dataset was cleaned and preprocessed to optimise model performance. The results obtained by training several classifiers yielded accuracies that reach 98%-99%. The research provides an end-to-end solution for the considered problem, through data preprocessing/feature selection/outlier removal and model training/evaluation. Hyperparameters were tuned for more than twenty models. All chosen models and features were ranked based on performance and robustness. A novel solution for optimising sensor placement has also been proposed by this study, to further improve sensor-based monitoring systems.

HTML

PDF

DOI: 10.3897/jucs.137699

Authors: Wolfram Luther, Ashot Harutyunyan

Abstract: This article presents an extensive literature review on the importance of fairness in society, science, the world of work and leisure, with a focus on healthcare. Depending on the application area, fairness criteria and metrics play a major role in evaluation, classification, and allocation. Different approaches to a general definition of algorithmic fairness for individuals or groups are considered, and their measures from the perspective of the concerned sciences and requirements for the decision-making processes are also formulated. There are many reasons for the lack of fairness: inadequate data quality or low model performance, differences in understanding, competing standards, inappropriate measures in selection, classification and decision-making, lack of accuracy or performance of algorithms paired with insufficient communication, interaction or collaboration of stakeholders. The requirements are illustrated using the example of medical risk prediction tools, e.g., the individual and familial risk for the occurrence of pathogenic variants in BRCA1 (BReast CAncer 1) or BRCA2 genes with impact on early breast cancer (BC) and ovarian cancer (OC) disease, and the 5-year risk that an individual with ocular hypertension will develop Primary Open Angle Glaucoma (POAG), the leading global cause of irreversible blindness.

HTML

PDF

DOI: 10.3897/jucs.133397

Authors: Nadia van Niekerk, Brink van der Merwe, Louwrens Labuschagne

Abstract: In the rapidly evolving realm of blockchain technology, the pursuit of enhanced privacy, security, and scalability has propelled the exploration of cryptographic innovations. Zero-Knowledge Proofs (ZKPs) have emerged as a pivotal solution, addressing diverse challenges across decentralized applications and cryptographic systems. However, the intricate mathematical foundations of ZKPs can pose a barrier to widespread adoption. To bridge this gap, a spectrum of ZKP tools has been developed, abstracting mathematical complexities and enabling developers with varying levels of expertise to incorporate ZKPs into their projects.The exploration of the 5 W’s – Who, What, When, Where, and Why – guides developers in selecting ZKP tools aligned with their specific needs and understanding. This paper serves as a vital resource for developers entering the dynamic landscape of ZKP development. By answering crucial questions and providing nuanced insights into ZKP tools, it empowers developers to navigate this intricate domain effectively. As ZKP technology continues to evolve, our findings contribute to the ongoing dialogue surrounding its implementation, utilization and the ever-adapting toolkit shaping the future of cryptographic innovation.This paper employs a Mining Software Repositories (MSR) approach to unravel insights from the expansive landscape of ZKP development. By delving into GitHub repositories, we categorize author archetypes, discuss ZKP proof constructions, identify phases of tool development, explore the level of understanding required and examine the correlation between tool types and application purposes. Through a metrics-driven analysis, we unveil patterns in tool popularity, development trends, and historical perspectives, offering a comprehensive understanding of the ZKP tooling ecosystem.

HTML

PDF

DOI: 10.3897/jucs.143527

Authors: Nguyen Thi Hoang Phuong, Phan Minh Nhat, Nguyen Van Hieu

Abstract: Near-infrared (NIR) spectroscopy has emerged as a valuable analytical technique for assessing the composition and quality of various materials. This study proposes NirMACNet, a novel convolutional neural network (CNN) architecture that incorporates a residual-based multi-scale kernel mechanism for enhanced prediction of compositional attributes. The model is evaluated on two distinct NIR spectral datasets, milk and soil, to demonstrate its generalization capability across domains. By leveraging multiscale kernel operations, NirMACNet effectively captures diverse spectral patterns, while its deep architecture facilitates comprehensive feature extraction. To mitigate performance degradation commonly associated with deeper networks, residual learning is employed. Experimental results indicate that NirMACNet consistently outperforms state-of-the-art methods in terms of prediction accuracy. Future work will involve expanding the diversity of training datasets and investigating alternative architectural enhancements to further improve model robustness and applicability.

HTML

PDF

DOI: 10.3897/jucs.183050

Authors: Christian Gütl

Abstract: Dear Readers, At the end of this year, it gives me great pleasure to announce another regular issue of J.UCS. In this issue, various topical aspects of computer science are covered in 6 articles by 14 authors from 7 countries – Armenia, Chile, Germany, Poland, South Africa, United Kingdom, and Vietnam.I would like to thank all the authors for their sound research and the editorial board and guest reviewers for the highly valuable review effort and suggestions for improvement. These contributions, together with the generous support of the KOALA initiative, maintain the quality of our journal. I am looking forward to continuing my work as Editor-in-Chief with the J.UCS community also in 2026. In an ongoing effort to further strengthen our journal, I would like to expand the editorial board: If you are a tenured associate professor or above with a strong publication record, you are welcome to apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and trends. Please consider yourself and encourage your colleagues to submit high-quality articles or special issue proposals to our journal. Finally, we are preparing to offer a J.UCS portal on ResearchGate as well as the Altmetric and Web of Science reviewer recognition service with beginning of the next year. In the last regular issue in 2025, I am very pleased to introduce 6 accepted articles covering various aspects of computer science. Nguyen Thi Hoang Phuong, Phan Minh Nhat, and Nguyen Van Hieu from Vietnam introduce in their study NirMACNet, a multi-scale convolutional neural network incorporating a residual mechanism to address the limitations of existing methods in predicting compositional attributes from raw near-infrared spectra across diverse materials. Experimental evaluations on milk and soil datasets demonstrate that NirMACNet significantly enhances feature extraction, eliminates the need for complex preprocessing, and outperforms state-of-the-art techniques in prediction accuracy. Nadia van Niekerk, Brink van der Merwe, and Louwrens Labuschagne from South Africa introduce in their work a mining software repositories approach to unravel insights from the expansive landscape of zero-knowledge proofs development. Through a metrics-driven analysis, the authors unveil patterns in tool popularity, development trends, and historical perspectives, offering a comprehensive understanding of the ZKP tooling ecosystem. In a collaboration between researchers from Germany and Armenia, Wolfram Luther and Ashot Harutyunyan present in their article an extensive literature review on the importance of fairness and absence of bias in society, science, the world of work and leisure, with a focus on healthcare and risk prediction tools. Based on this findings, relevant approaches to a general definition of algorithmic fairness for individuals and groups are presented, assessed from the perspective of the concerned sciences, and requirements for the decision-making processes are formulated. Alexandru Pintea from UK researches in his article how sensor data can be used by robust AI models to accurately estimate the number of people present in a room while considering real-time monitoring as a primary use case. The study explores sensor positioning and model hyperparameters for numerous models in order to optimize the robustness and performance of inhabitance monitoring systems. Przemyslaw Jatkiewicz from Poland proposes in his work to integrate artificial intelligence techniques - specifically machine learning, neural networks, and user behaviour analysis - to enhance corporate data protection through real-time anomaly detection and adaptive authentication. The research demonstrates that AI-based access control systems significantly improve security by enabling dynamic threat detection and prediction, and provides practical recommendations for implementing machine learning-based anomaly detection systems in combination with traditional authentication methods, while ensuring data protection through techniques such as anonymisation, encryption, and federated learning. Claudio Gutiérrez-Soto, Marco A. Palomino, Patricio Galdames, and Cristian Duran-Faundez from Chile address in their study the challenge of identifying determinants of academic success in higher education by integrating descriptive student profiling with Bayesian predictive modelling. Using enriched datasets from the University of Bio-Bio, the approach achieves over 97% accuracy, advancing learning analytics through reliable forecasting.Season greetings to all of you, relaxing holidays and ‘Enjoy Reading’! Best regards, Christian Gütl, Managing Editor-in-ChiefGraz University of Technology, Graz, Austria

HTML

PDF

DOI: 10.3897/jucs.129768

Authors: Atefeh Parvin, Farahnaz Mohanna, Masoumeh Rezaei

Abstract: When reconstructing jigsaw puzzles, the state-of-the-art algorithms struggle to distinguish between identically colored pieces that belong to different objects. This limitation significantly impacts the accuracy of puzzle solvers, especially in complex images with repetitive colors or textures. To address this issue, we propose a new GA-based square jigsaw puzzle solver. A combined color and texture discriminator is incorporated into the proposed solver to prevent pieces that have the same color but come from distinct objects from being joined together incorrectly. Color and texture features are extracted separately using the sum of square distances and Gabor filter. To evaluate the performance of the proposed solver, we used a dataset consisting 66 images: 20 puzzles with 432 pieces from the MIT collection, 20 puzzles with 540 pieces, and 20 puzzles with 805 pieces from the McGill collection, and 3 puzzles with 2360 pieces, and 3 puzzles with 3300 pieces from the Pomeranz collection. For the direct, neighbor, and largest component comparisons, the proposed method’s accuracy is 92.91%, 96.66%, and 90.83%, respectively. The proposed method demonstrates an improvement of 11.9%, and 3.65% in accuracy based on direct and neighbor comparison criteria, on the database images when compared to current state-of-the-art GA-based square jigsaw puzzle solver.

HTML

PDF

DOI: 10.3897/jucs.152114

Authors: Vahide Nida Kılıç, Esra Saraç Eşsiz

Abstract: In previous studies, the focus has predominantly been on anomaly detection, with minimal attention given to anomaly prevention. However, anomaly prevention holds greater significance than anomaly detection. Preventing anomalous behavior before it occurs and identifying potential anomalies in advance to enable timely intervention is both challenging and crucial. In this study, a Suspicious Nodes Anomaly Prevention framework for anomaly prevention has been developed. First, a novel K-medoid based Salp Swarm Anomaly Detection method is proposed within the framework. This method reveals unclustered data by applying clustering and determines the boundaries of clusters using a nature-inspired algorithm that optimizes the threshold. Since threshold determination is an optimization problem, it aligns well with nature-inspired algorithms. Additionally, the Enron email dataset was selected as it is a real-world dataset with accessible content information. Initially, content and node features were extracted from the Enron email dataset. The proposed anomaly detection method was then applied separately to each of these features. Nodes identified as anomalous by one feature but normal by others were of particular interest. These nodes were labeled as “suspicious nodes,” and their connections were analyzed to detect potentially harmful email content. This framework fills a significant gap in the anomaly detection literature by contributing an unprecedented approach to anomaly prevention, offering early intervention capabilities in various sectors by identifying risks in advance. In this study, the proposed framework demonstrates high efficacy in detecting anomalies, achieving a True Positive Rate of 94% in node-based anomaly detection and 78% in content-based anomaly detection, indicating a robust capability for early intervention and risk identification.

HTML

PDF

DOI: 10.3897/jucs.141599

Authors: José de Oliveira Guimarães

Abstract: Certain languages allow a metaprogram to act as a compiler plugin and thus alter the compilation process. The metaprogram interacts with low-level details of the compiler, making its construction difficult and potentially leading to errors. Different parts of the metaprogram may have conflicting interactions, thus producing unintended outcomes. This article introduces metaprogramming in the prototype-based object-oriented language Cyan. This language provides the same core functionality as other metaprogramming systems while introducing features that improve interactions between the compiler and different components of the metaprogram. Further-more, Cyan incorporates security measures designed to circumvent typical issues encountered in metaprogramming.

HTML

PDF

DOI: 10.3897/jucs.145223

Authors: Yongbin Li, Xinyue Yang, Linhu Hui, Enlin Fu, Stephanie Chua

Abstract: False-positive reduction is a critical step in the automatic lung nodule detection system, playing a significant role in the early detection and diagnosis of lung cancer. However, accurately distinguishing between nodules and non-nodules remains a challenge due to their high morphological similarity. To effectively reduce the false positives (FPs) in automated nodule detection using computed tomography (CT), we propose an attention-guided multi-scale cross-fusion three-dimensional (3D) convolutional neural network (AMCF-CNN). The proposed network adopts an innovative multi-scale cross-fusion strategy to integrate nodule features across different scales and incorporates a SimAM-Res module to adaptively enhance key features, thereby improving overall feature representation. To further enhance global contextual awareness, a Global Modeling Module (GMM) based on self-attention mechanism is introduced, enabling complementary fusion of local structural details and global semantic context, thereby enhancing structural discrimination in complex imaging backgrounds. We conducted experiments on the LUNA16 dataset consisting of 888 chest CT cases, and evaluated model performance using 10-fold cross-validation. Experimental results demonstrate that the proposed model achieves sensitivities of 0.983 and 0.984 at 4 and 8 FPs/scan, respectively, with a balanced accuracy of 0.983 and a competitive performance metric (CPM) score of 0.936. Compared to existing methods, AMCF-CNN achieves competitive performance in false-positive reduction, offering a more accurate and robust solution for lung nodule detection and demonstrating strong potential for real-world clinical applications.

HTML

PDF

DOI: 10.3897/jucs.152864

Authors: Claudio Alvarez, Andres Carvallo, Gustavo Zurita

Abstract: As student cohorts grow, real-time case-based learning discussions generate increasing volumes of textual data, intensifying the orchestration load teachers must manage. Reviewing and providing feedback on student responses promptly becomes increasingly challenging, demanding efficient methods to assist educators in selecting relevant contributions to steer classroom discussions. This study proposes a low-footprint natural language processing (NLP) approach that leverages small-scale models running on commodity hardware, avoiding the computational overhead and cost associated with large language models. Our system, integrated into EthicApp, a collaborative learning platform, employs pre-trained language models such as the Universal Sentence Encoder (USE) and Bidirectional Encoder Representations from Transformers for Spanish (BETO), along with traditional text-mining techniques like Term Frequency-Inverse Document Frequency (TF-IDF). Through expert evaluations, we found that BETO exhibited superior performance in identifying relevant student responses but required GPU acceleration. At the same time, USE provided an efficient alternative that outperformed TF-IDF and remained feasible for CPU-based execution. Additionally, the methods showed a tendency—most notably BETO—to select longer responses, which, rather than introducing selection bias, was interpreted as an indicator of deeper student engagement. No significant semantic bias was found, ensuring a fair representation of students’ perspectives. Our findings suggest that low-footprint NLP can effectively reduce teacher orchestration load, enabling more targeted feedback without requiring extensive computational resources.

HTML

PDF

DOI: 10.3897/jucs.159191

Authors: Muhammad Saqlain, José M. Merigó, Keivan Amirbagheri, Hermann Maurer

Abstract: This study presents a comprehensive bibliometric analysis of the Journal of Universal Computer Science (JUCS) covering the period 1994-2024, based on data retrieved from the Scopus database in April 2025. The analysis has been conducted following the scientific procedures and rationales for systematic literature reviews (SPAR-4-SLR) protocol and using mathematical and statistical methods, VOS viewer, and bibliometrix. The article investigates the journal’s publication trends, citation structures, collaboration networks, and thematic evolution over three decades. A comparative review with the study published in 2021 by Baloain and collaborators, shows improved journal metrics, enhanced global recognition, and alignment with emerging research trends. The co-citation and bibliographic coupling analyses highlight JUCS’s strong intellectual connectivity across key domains of computer science, particularly in interdisciplinary areas. Thematic mapping and keyword analysis show a clear transition from classical computing themes to modern topics like artificial intelligence, deep learning, big data, and blockchain, demonstrating the journal’s responsiveness to evolving scientific priorities. The study concludes with general findings, practical implications, and future research directions, emphasizing JUCS’s role as a durable, adaptive, and impactful platform for scholarly output in the global computer science research community.

HTML

PDF

DOI: 10.3897/jucs.178548

Authors: Christian Gütl

Abstract: Dear Readers,It gives me great pleasure to announce the 13th J.UCS issue of 2025. I would like to thank all the authors for their sound research and the editorial board and guest reviewers for the extremely valuable reviews and suggestions for improvement. These contributions together with the support of the community and the generous support of the KOALA initiative enable us to run our journal and maintain its quality.I would still like to expand our editorial board: If you are a tenured associate professor or above with a good publication record, please apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and emerging trends.In this regular issue, I am very pleased to present 6 accepted articles by 18 authors from 9 countries: Australia, Austria, Brazil, Chile, China, Iran, Malaysia, Spain, and Turkiye.In a collaborative effort between researchers from Australia and Austria, Muhammad Saqlain, José M. Merigó, Keivan Amirbagheri, and Hermann Maurer provide a comprehensive bibliometric analysis of the Journal of Universal Computer Science (JUCS) from 1994 to 2024, employing the SPAR-4-SLR protocol with VOSviewer and Bibliometrix to examine publication patterns, collaboration networks, and thematic evolution. The findings reveal JUCS’s strengthened global impact, intellectual connectivity, and transition toward modern computer science domains such as artificial intelligence, deep learning, and big data, underscoring its sustained relevance and adaptability over three decades.Claudio Alvarez, Andres Carvallo, and Gustavo Zurita from Chile address in their research the growing orchestration load teachers face in real-time case-based learning discussions by introducing a low-footprint natural language processing approach that can run on standard hardware without requiring large-scale models. Expert evaluation shows that small pre-trained models, particularly BETO and the Universal Sentence Encoder, effectively identify relevant student responses while maintaining low computational cost and minimizing bias, enabling scalable and equitable AI support for educators in the Global South.In a collaborative research effort between Malaysia and China, Yongbin Li, Xinyue Yang, Linhu Hui, Enlin Fu, and Stephanie Chua focus on Lung Nodule Detection. To reduce false positives on CT scans, the authors propose AMCF-CNN, a 3D attention-guided multi-scale cross-fusion network that effectively integrates local features and global contextual information through SimAM-Res and the Global Modeling Module. Evaluated on the LUNA16 dataset, AMCF-CNN achieves a CPM of 0.936 and a balanced accuracy of 0.983, outperforming most existing methods. José de Oliveira Guimarães from Brazil focuses his research on aspects of metaprogramming in Cyan. Most compile-time metaprogramming languages allow unrestricted modifications to the in-memory representation of the base program by providing largely unconstrained access to the compiler’s internal data structures. The Cyan metaprogramming system, in contrast, uses a sandboxed model that prevents errors caused by such unrestricted access while retaining most of the expressive power of other systems.Vahide Nida Kılıç and Esra Saraç Eşsiz from Turkiye propose in their research a novel anomaly prevention framework that combines clustering-based nature-inspired algorithms with both node and content features to identify suspicious instances in communication networks. The approach advances the field by shifting from traditional anomaly detection to proactive prevention through the early classification of suspicious nodes, threshold-based risk assessment, and link analysis to flag potential anomalies before escalation.Atefeh Parvin, Farahnaz Mohanna, and Masoumeh Rezaei from Iran discuss in their research a genetic-based square jigsaw puzzle solver. To address the challenge of distinguishing identically colored pieces from different objects in jigsaw puzzle solving, they propose a genetic algorithm-based solver that integrates a novel color and texture compatibility criterion using Sum of Squared Distances and Gabor filter features. This approach improves accuracy by 11.9% and 3.65% in direct and neighbor comparison criteria across 66 puzzles, offering a data-efficient solution for type 1, 2, and 3 square jigsaw puzzles.Enjoy Reading!Cordially, Christian Gütl, Managing Editor-in-ChiefGraz University of Technology, Graz, Austria

HTML

PDF

DOI: 10.3897/jucs.141304

Authors: Fouzi Lezzar, Seif Eddine Mili

Abstract: Abstract: Physical rehabilitation is essential for restoring motor function; however, traditional methods often require therapist supervision, which can be costly and inaccessible. Home-based rehabilitation offers a practical alternative, but without real-time guidance, patients may develop incorrect movement patterns that impede progress. Existing approaches typically provide feedback only after exercises are completed, reducing their effectiveness. To overcome this limitation, we propose a Temporal Conditional Generative Adversarial Network (TCGAN)-based motion generation system that delivers real-time skeletal guidance tailored to each patient’s body structure and positioning. By detecting key anatomical landmarks and generating adaptive motion sequences, the system ensures precise movement execution, minimizing errors and improving rehabilitation outcomes. Patients can mimic these movements, enabling them to perform exercises with greater accuracy and confidence. Quantitative and qualitative evaluations confirm the effectiveness of the generated exercises, thanks to an optimized architecture, an improved loss function, a refined training process, and fine-tuned TCGAN hyperparameters. Experimental results demonstrate a high degree of similarity between generated and real movements, with a Fréchet Inception Distance (FID) score of 0.89 and strong temporal alignment, as shown by Dynamic Time Warping (DTW) scores ranging from 2.9 to 5.6 across nine rehabilitation exercises. These metrics underscore the system’s realism and reliability.

HTML

PDF

DOI: 10.3897/jucs.146407

Authors: Zhichao Chen, Zixi Han, Bingqing Shen, Yuxin Zeng, Min Wang, Hongming Cai, Minqi Wang

Abstract: Abstract: The Integrated Process Planning and Scheduling (IPPS) problem has long been a practical concern in industry, with numerous problem modelling and solving methods to optimize multiple objectives, such as makespan and cost. As yet, existing modelling methods have difficulties in satisfying new modelling requirements for adequately representing real complex manufacturing jobs. This paper proposes a novel computing framework for modelling and solving complex IPPS problems based on a Supplier-Input-Process-yield-Output-Customer-Resource-Control (called SIPyOC-RC) model. This framework includes a SIPyOC-RC modelling method to address the modelling adequacy problem. With the proposed method, complex IPPS problems in real manufacturing scenarios, such as commercial airplane component production and assembly, can be modeled with the key demanding features. Next, the SIPyOC-RC model can be transformed into a Constraint Programming (CP) model, which can then be solved with any CP-solver. Based on an implemented testbed, the modelling adequacy and computation performance of the proposed approach has been evaluated. The experiment results show that it can successfully solve complex IPPS problems without significant computation cost. Moreover, its compatibility with conventional IPPS problems is also discussed. Overall, the outcome of this study contributes both theoretical methodology and practical experience to industrial computing.

HTML

PDF

DOI: 10.3897/jucs.141629

Authors: Samet Aymaz

Abstract: oronary artery disease (CAD) is a major global health issue. Early detection plays a crucial role in reducing risk and improving patient outcomes. This study proposes a novel, efficient approach to CAD diagnosis by integrating a histogram-based feature selection method with a specially designed long short-term memory (LSTM) classifier. The method is evaluated on two benchmark datasets: Z-Alizadeh Sani and Cleveland. Imbalanced class distribution, a common challenge in medical datasets, is addressed using the synthetic minority over-sampling technique (SMOTE). The proposed feature selection technique offers a fast and simple alternative to traditional optimization methods like particle swarm optimization (PSO), teaching-learning-based optimization (TLBO), and the whale optimization algorithm (WOA), which typically require extensive parameter tuning and longer processing times. The histogram-based method selects features based on their distribution similarity to a Gaussian profile, aiming to enhance classification performance and computational efficiency. The selected features are then classified using a custom-designed LSTM architecture optimized through Grid Search and validated via k-fold cross-validation (k-fold). The effectiveness of the proposed method is demonstrated by comparing it with other feature selection approaches using metrics such as accuracy, precision, sensitivity, and the F1-score (f-score). Experimental results show that the histogram-based method significantly improves classification accuracy and reduces computational time. This approach offers a promising, low-cost, and scalable solution for CAD diagnosis, especially in resource-constrained settings, and provides valuable contributions to the field of medical data analysis.

HTML

PDF

DOI: 10.3897/jucs.143553

Authors: Aluizio Haendchen Filho, Jonathan Nau, Hércules Antonio do Prado, Edilson Ferneda

Abstract: This paper explores Natural Language Processing (NLP) in automatic comprehension and discourse analysis, focusing on argument mining. While previous works have focused on English, this study addresses the lack of adequate corpora and methodologies for Brazilian Portuguese. The researchers employed a corpus of essays from Brazil's National High School Exam (ENEM) to investigate the impact of discourse markers on identifying argumentative structures using feature engineering with machine learning. The proposed methodology offers key advantages over transformer-based approaches: enhanced interpretability of feature selection, computational efficiency, and improved adaptability across different linguistic domains. By systematically 'opening the black box' of machine learning models, this approach provides insights into the discourse marker identification decision-making process, in contrast to the opaque neural network models. Unlike the transformers-based solutions, this approach offers a transparent solution based on feature engineering allowing insights into the linguistic patterns underlying argumentative structures in Portuguese. While acknowledging the relatively small corpus size as a limitation, the researchers suggest that future work should focus on expanding the dataset for further evaluation. This work lays the groundwork for advancing NLP in Portuguese by providing valuable features and methodologies for feature engineering in automated linguistic analysis tasks such as essay scoring, opinion mining, and text summarization. The findings demonstrated a significant breakthrough, revealing that a concise set of only five argument mining-derived features dramatically improved the model accuracy, surpassing the performance of an initial, extensive set of over 600 features. These features specifically enhanced the evaluation of Competence 5, which assesses students' ability to develop intervention proposals grounded in scientific concepts.

HTML

PDF

DOI: 10.3897/jucs.151691

Authors: Irene Ruiz-Pozo, Juan Morales-García, Claudia Ximena Aguirre-Mejía, Antonio Serrano

Abstract: The use of Generative Artificial Intelligence (GenAI) is revolutionising how companies create digital content on social networks. However, how users respond to this content remains less explored. The aim of this article is to compare GenAI-generated content with human-generated content on Instagram. First, through a survey of 273 students, we explored user engagement with the eSports team’s content on Instagram. Second, we measured followers’ engagement behaviors (i.e., “likes”, comments, shares and reach) with GenAI-generated content versus human-created content. Results indicate that both GenAI tools (ChatGPT and Gemini) achieved similar levels of engagement in terms of “likes”. ChatGPT stood out for its use of interactive features like opinion polls, particularly in “stories”, while Gemini excelled in total reach and visibility. In contrast, human-generated content attracted a higher proportion of non-followers, suggesting a stronger potential to expand the existing audience. These findings underscore the advantages of adopting a hybrid strategy that combines the scalability and speed of GenAI with the contextual relevance and authenticity of human-created content.

HTML

PDF

DOI: 10.3897/jucs.175734

Authors: Christian Gütl

Abstract: Dear Readers,It is a great pleasure to announce the twelfth J.UCS issue of 2025. As always, I would like to thank all the authors for their sound research and the editorial board and guest reviewers for the extremely valuable reviews and suggestions for improvement. These contributions together with the support of the community and the generous support of the KOALA initiative enable us to run our journal and maintain its quality. I would still like to expand our editorial board: If you are a tenured associate professor or above with a good publication record, please apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and emerging trends. In this regular issue, I am very pleased to present 5 accepted papers by 18 authors from 5 countries: Algeria, Brazil, China, Spain, and Türkiye. Irene Ruiz-Pozo, Juan Morales-García, Claudia Ximena Aguirre-Mejía, and Antonio Serrano from Spain explore in their study the application of generative AI tools such as ChatGPT and Gemini - in creating Instagram marketing content for eSports, comparing them with human-generated campaigns. Findings reveal that AI-generated content achieves comparable or higher engagement metrics, while hybrid strategies combining AI and human creativity maximize reach, authenticity, and audience growth. Aluizio Haendchen Filho, Jonathan Nau, Hércules Antonio do Prado, and Edilson Ferneda from Brazil tackle in their research the scarcity of methodologies for argument mining in Brazilian Portuguese by analyzing student essays from the National High School Exam, and introducing a feature-engineering approach based on discourse markers. This method enhances automated essay scoring while providing a transparent and computationally efficient alternative to black-box transformer models. Findings show that a streamlined set of five argument-mining features significantly improves scoring accuracy for the competency of developing intervention proposals grounded in scientific concepts. Samet Aymaz from Türkiye discusses in the article a histogram-based feature selection method with a custom LSTM model to improve early detection of coronary artery disease. The proposed approach achieves convincing classification results on benchmark datasets, offering a fast, accurate, and scalable diagnostic solution. Zhichao Chen, Zixi Han, Bingqing Shen, Yuxin Zeng, Min Wang, Hongming Cai, and Minqi Wang from China address in their study the issue of complex Integrated Process Planning and Scheduling (IPPS) modeling that existing methods have difficulties in satisfying new modeling requirements for adequately representing the complex scenarios and settings in real-life production. This study proposed an enhanced modeling and computing framework for solving complex IPPS problems, and empirically demonstrated its effectiveness and efficiency in solving complex IPPS problems with the key demanding features. Fouzi Lezzar and Seif Eddine Mili from Algeria address in their research the lack of real-time guidance in home-based physical rehabilitation by proposing a Temporal Conditional Generative Adversarial Network (TCGAN) that generates patient-specific skeletal motion sequences for exercise execution. The system demonstrates high accuracy and realism, achieving strong alignment between generated and real movements, and significantly improves the precision of rehabilitation exercises compared to existing methods.Enjoy Reading!Cordially,Christian Gütl, Editor-in-Chief

HTML

PDF

DOI: 10.3897/jucs.130450

Authors: Jafseer KT, Shailesh S, Sreekumar A

Abstract: Recent advancements in data mining and knowledge discovery have created numerous research opportunities in streaming data analysis. One critical challenge is developing machine learning models that can efficiently handle changes in features and dynamic concepts, including concept drift, feature drift, and feature evolution. State-of-the-art techniques proposed to address these anomalies in data streams often assume that a constant set of features is available for processing. However, in real-time scenarios, the situation is quite different, as the set of features in a stream may vary over time due to factors such as the disappearance of existing features or the emergence of new ones. The proposed work focuses on handling dynamically evolving features by introducing a novel solution that leverages a Dynamic Autoencoder DAE and ensemble learning. Additionally, adaptive windowing and concept-preserving mechanisms improve the proposed architecture by retaining the concept information from previous data windows. The ensemble technique used in the proposed classification framework demonstrates promising performance in diverse datasets, achieving accuracies of 86%, 94%, and 95% in the Weather, Electricity and Forest Cover Type datasets, respectively. This innovative integration of deep learning and traditional methods effectively addresses various challenges in streaming data analysis.

HTML

PDF

DOI: 10.3897/jucs.135070

Authors: Ming-Lung Hsu, Yu-Wei Liu, Sheng Tun Li

Abstract: Data mining can be considerably improved with the inclusion of prior domain knowledge; such knowledge reveals complex patterns that might otherwise remain hidden. Among such patterns, monotonic relationships between variables are crucial because of their applicability in real-world contexts. Although considerable growth has occurred in the development of monotonic classification models, many of these models excel in binary or multiclass classification but falter in one-class classification. To address this problem, we developed a monotonicity-constrained support vector domain description (MC-SVDD) model in this study. This model is an innovative evolution of the monotonicity-constrained support vector machine model and is specifically designed for one-class classification with strict adherence to monotonicity constraints. In the developed MC-SVDD model, monotonicity constraints are integrated into the well-established support vector domain description (SVDD) framework. Moreover, methods such as quadratic programming and data visualization are incorporated into the MC-SVDD model. In extensive evaluations, the MC-SVDD model outperformed a conventional SVDD model in prediction performance. This study makes a key contribution to domain-driven data mining.

HTML

PDF

DOI: 10.3897/jucs.150681

Authors: Ashish Ranjan Mishra, Rakesh Kumar, Rajkumar Saini

Abstract: Ensuring secure and reliable person authentication is a critical challenge in modern security systems. Traditional biometric systems relying on physiological traits like fingerprints, iris, and facial recognition often suffer from spoofing vulnerabilities. In contrast, electroencephalogram (EEG) signals, characterized by unique temporal and cognitive patterns, provide a robust authentication mechanism. This paper introduces DeepV-Net, a multimodal fully convolutional neural network that leverages both EEG signals and dynamic handwritten signature data acquired from Wacom devices. The proposed model integrates spatial and temporal features of EEG signals with distinctive movement-based signature patterns through an end-to-end multimodal fusion strategy. Experimental evaluations on benchmark datasets demonstrate that DeepV-Net outperforms unimodal approaches and state-of-the-art authentication methods, achieving a training accuracy of 99.1% and a validation accuracy of 93.3%. These findings highlight the complementary nature of EEG and signature modalities, paving the way for more secure and efficient biometric authentication systems.

HTML

PDF

DOI: 10.3897/jucs.135907

Authors: Emine Cengiz, Murat Gök

Abstract: Money laundering seriously threatens economic stability by legitimizing illegal gains. Despite the transparency and security advantages offered by the blockchain technology, anonymity can create a platform for concealing illegal activities. Therefore, detecting and preventing money laundering activities in blockchain networks are of great importance. This study classifies money transfers during Bitcoin transactions as licit or illicit. By working on the Elliptic dataset to detect money laundering activities in the Bitcoin network, we examined money laundering traffic data using a graph data structure. This study presents a novel method for analyzing complex networks in money laundering as a chaotic time series. First, we increase the number of features of the graph nodes and convert them into a time series. By transferring the obtained time series to the phase space, we calculated the Lyapunov Exponents and aimed to capture the changes and uncertainties in the dynamic structure of the system more accurately using different embedding dimensions. We reconstructed the graph structure representing the transactions based on the feature vectors of these exponents, and classified the transactions using the Graph Convolutional Network method. In our study, we achieved a precision of 92.5%, recall of 92.1%, F1-score of 92.3%, and accuracy of 86.2%. These results demonstrate the effectiveness and reliability of our model in detecting money laundering. This study offers a novel approach for classifying chaotic structures in anti money laundering.

HTML

PDF

DOI: 10.3897/jucs.146032

Authors: Timothy Louis Scott, Wei Wei Goh, Navid Ali Khan

Abstract: On Amazon, buyers can submit reviews on products they have purchased. These reviews contribute to a potential buyer’s decision-making process, as buyers read reviews to decide whether to buy a product. Additionally, sellers depend on reviews to improve their product offerings. Amazon’s summary of reviews does not clearly indicate if an aspect of a product is mentioned positively or negatively. Buyers can manually read a small number of reviews to understand the overall sentiment towards a product, but reading reviews becomes progressively more difficult as the number of reviews increases, as it can lead to information overload. To address this problem, a hybrid machine learning classification algorithm that employs a branch of natural language processing, specifically aspect-based sentiment analysis, was developed to detect the polarity and key aspects mentioned in Amazon product reviews. Naïve Bayes, SVM, Decision Tree and Random Forest were compared to determine the two best algorithms for this purpose. The hybrid algorithm, named Soft Voting Hybrid Algorithm (SVHA), was implemented by training and testing a voting classifier using soft voting, which produced the final prediction by selecting the class with the highest average sum of probabilities from two base classifiers with the highest accuracies and macro F1-scores. Based on the experiments conducted, SVHA attained higher accuracies and macro F1-scores compared to the other four algorithms, showing its suitability in conducting aspect-based sentiment analysis.

HTML

PDF

DOI: 10.3897/jucs.171956

Authors: Christian Gütl

Abstract: Dear Readers,Welcome to another J.UCS regular issue covering 5 articles on topical research areas in computer science. As part of our continuous improvement process, we have decided to provide more information about the accepted papers in the editorials starting with this issue.I would like to thank all the authors for their sound research and the editorial board and guest reviewers for their extremely valuable review effort and suggestions for improvement. These contributions, together with the generous support of the KOALA initiative, contribute to the quality of our journal.In an ongoing effort to further strengthen our journal, I am continuously looking for new editorial board members: If you are a tenured associate professor or higher with a good publication record, you are welcome to apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and trends.It gives me great pleasure to announce the eleventh J.UCS issue of 2025. In this issue, 5 papers by 14 authors from 5 countries - India, Malaysia, Sweden, Taiwan, Türkiye - cover various topical aspects of computer science.Timothy Louis Scott, Wei Wei Goh, and Navid Ali Khan from Malaysia introduce their research on aspect-based sentiment analysis for product reviews. To address the problem of information overload on e-commerce platforms, a hybrid machine learning classification algorithm that employs aspect-based sentiment analysis and soft voting, was developed to detect the polarity and key aspects mentioned in Amazon product reviews. Based on the experiments conducted, SVHA attained higher accuracies and macro F1-scores compared to four other algorithms, showing its suitability in conducting aspect-based sentiment analysis.Emine Cengiz and Murat Gök from Türkiye propose in their study an enhanced approach for detecting money laundering in blockchain networks by representing transaction graphs as chaotic time series, extracting Lyapunov Exponents through phase space reconstruction, and classifying them with Graph Convolutional Networks. The main contribution is a feature expansion and chaotic analysis framework that improves blockchain transaction representation and enables more effective detection of illicit activities.In a collaborative effort, researchers from India and Sweden, Ashish Ranjan Mishra, Rakesh Kumar, and Rajkumar Saini introduce a deep learning technique for multimodal biometric authentication. More specifically, the article proposes DeepV-Net, a multimodal biometric authentication system that fuses EEG signals with handwritten signatures using V-net integrated with squeeze-excitation and attention modules. The model outperforms unimodal and state-of-the-art methods, demonstrating high accuracy, robustness, and significant contributions from its fusion and attention mechanisms.Ming-Lung Hsu, Yu-Wei Liu, and Sheng Tun Li from Taiwan address in their study the limitations of existing monotonic classification models in one-class classification by proposing a monotonicity-constrained support vector domain description – the MC-SVDD model, which integrates monotonicity constraints into the SVDD framework using quadratic programming and visualization techniques. Experimental results show that MC-SVDD outperforms conventional SVDD in prediction performance, contributing to the advancement of domain-driven data mining.Jafseer KT, Shailesh S, and Sreekumar A from India address in their research a feature evolution aware classification framework for streaming data using dynamic autoencoder and ensembled learning. The proposed research focuses on handling dynamically evolving features by introducing an enhanced solution that leverages a Dynamic Autoencoder DAE and ensemble learning. The ensemble technique used in the proposed classification framework demonstrates promising performances in diverse datasets, achieving accuracies of 86%, 94%, and 95% in the Weather, Electricity and Forest Cover Type datasets.Enjoy Reading!Kind regards,Christian Gütl, Managing Editor-in-Chief

HTML

PDF

DOI: 10.3897/jucs.140463

Authors: Daniel Spiekermann

Abstract: The evolution of modern data centers from traditional hardware-based infrastructures to highly virtualized environments has introduced new complexities in network traffic analysis. Virtual networks, characterized by dynamic changes in both overlay and underlay architectures, necessitate sophisticated methods for accurate anomaly detection and network analysis. This paper investigates the real-world behaviour of network traffic within virtualized environments, identifying the key factors that impact packet dynamics, including VM operations, multi-tenancy, user customization, and hardware adjustments. By defining the frequency and nature of these events, this research provides further details for the creation of accurate packet generation tools. These tools must simulate the dynamic characteristics of virtual networks, enabling reliable and realistic synthetic data generation. The research explains the need for enhanced packet generation methodologies to provide valid training and testing data for digital investigations, anomaly detection, and network simulations. As a result, this paper emphasizes the importance of developing accurate synthetic network traffic that mirrors real-world conditions and provides a valid basement for traffic analysis in virtual networks.

HTML

PDF

DOI: 10.3897/jucs.137544

Authors: Igor Tomičić, Petra Grd, Andrija Bernik

Abstract: This paper presents a comprehensive analysis of the integration of artificial intelligence (AI) into threat intelligence (TI) systems, focusing on its potential to enhance cybersecurity operations. The paper presents an extensive literature review, covering the current state of AI applications in TI, including machine learning, deep learning, and natural language processing for automating threat detection, classification, and analysis. It outlines the core functions of AI in TI, such as threat detection, correlation, prediction, and automated response, and introduces a detailed taxonomy categorizing AI techniques based on their roles in enhancing TI processes. Additionally, the paper proposes a conceptual workflow for AI-powered TI, illustrating how AI can streamline data collection, threat analysis, and incident response. A roadmap for future research is further provided, highlighting key areas for development, including explainable AI, federated learning, and edge computing. This work contributes to the field by offering a structured framework for integrating AI into TI and identifying critical challenges and future directions in AI-driven cybersecurity.

HTML

PDF

DOI: 10.3897/jucs.135198

Authors: Juan Morales-García, Fernando Terroso-Sáenz, Andrés Bueno-Crespo, José M. Cecilia

Abstract: Motivated by the large number of wearables offering geolocation, human mobility mining has emerged as an novel research field within AI. The study of mobility creates increasingly predictable models in which it is easy to find patterns of behaviour. However, this data is not publicly available and access to it is restricted to large telecommunications operators. In this context, this paper aims to solve one of the main problems of human mobility databases, i.e. the scarcity of data for the generation of human mobility models. For this purpose, Generative adversarial network (GANs) have been proposed to generate synthetic time-series mobility data. Moreover, several neural network models are proposed to assess the impact of synthetic data generation on the prediction of human mobility. Our results show that the use of synthetic data improves predictions of human mobility compared to models based on available measured data. Specifically, the reinforcement learning with synthetic data benchmark, when compared to using only ground truth data, achieved a 1.22% improvement in R2, a 0.70% reduction in RMSE, a 2.97% decrease in MAE, a 27.07% reduction in MAPE, and an 18.18% improvement in CVRMSE, demonstrating its effectiveness in enhancing predictive accuracy.

HTML

PDF

DOI: 10.3897/jucs.132502

Authors: Sorav Kumar Singh, Alak Roy, Rajneesh Raushan

Abstract: In the field of underwater exploration and research, Underwater Wireless Sensor Networks (UWSNs) play a vital role in understanding the marine environment, oceanography, and marine biology. A key strategy used in UWSNs to aggregate sensor nodes and improve network performance while extending battery life through lower energy usage is clustering. However, available clustering algorithms do not specifically address all the underwater problems, viz., communication is constrained by the limited bandwidth and high latency of acoustic signals, while energy consumption is critical due to the difficulty of recharging or replacing underwater batteries. The harsh underwater environment, with varying pressure, salinity, and movement, affects sensor performance and durability. Accurate localization is difficult without GPS and relies on less precise acoustic methods. So, this paper proposes a Residual Energy-Aware Fuzzy-Based Clustering Algorithm (REAFCA) for UWSNs which presents a novel framework intended to improve network performance and address issues with energy usage. For effective data routing, the REAFCA dynamically arranges clusters based on important factors such as node rank, radius, threshold, angular velocity, and residual energy. To maximize leadership inside the clusters, the adaptive threshold method makes sure that only superior cluster heads are chosen. The algorithm also incorporates dynamic range changes for communication to adapt to changing network circumstances. This algorithm mainly focuses on clustering in an underwater environment while improving the energy efficiency and network life of the nodes. Simulation results demonstrate the superiority of the proposed algorithm over K-means, K-meansA, LEACH, PEGASIS, HEED, DB-SCAN and HEER algorithms in terms of energy efficiency and throughput while achieving comparable average delay.

HTML

PDF

DOI: 10.3897/jucs.131773

Authors: Rodrigo Costa Camargos, Ismar Frango Silveira

Abstract: This study explores the application of Explainable Artificial Intelligence (XAI) techniques to mitigate cognitive biases in predicting student dropout. Focusing on the Explainable Boosting Machine (EBM), we compare its performance and explainability with Logistic Regression and XGBoost models. While EBM and Logistic Regression have inherent explainability, we employ SHAP and Morris Sensitivity Analysis for XGBoost to provide both local and global explanations. Our findings indicate that the inherently interpretable nature of EBM supports clear and actionable decision-making in educational settings. When integrated with additional XAI methods for comparative analysis with models like Logistic Regression and XGBoost, the approach can further enhance the understanding of key factors contributing to student dropout.

HTML

PDF

DOI: 10.3897/jucs.168512

Authors: Christian Gütl

Abstract: Dear Readers,I am very pleased to announce today the tenth J.UCS issue of 2025. In this issue, various topical aspects of computer science are covered in 5 articles by 13 authors from 5 countries (Brazil, Croatia, Germany, India, Spain). As always, I would like to thank all the authors for their sound research and the editorial board for their highly valuable review effort and suggestions for improvement. These contributions sustain the quality of our journal. I would also like to express my sincere thanks to the KOALA Initiative and its team for their financial support, without which the J.UCS team would not be able to publish our journal.In an ongoing effort to further strengthen our journal, I would like to expand the editorial board: If you are a tenured associate professor or above with a strong publication record, you are welcome to apply to join our editorial board. We are also interested in receiving high-quality proposals for special issues on new topics and trends. Please consider yourself and encourage your colleagues to submit high-quality articles or special issue proposals for our journal.In this regular issue, I am very pleased to introduce the following 5 accepted articles: Rodrigo Costa Camargos and Ismar Frango Silveira from Brazil explore in their research the application of Explainable Artificial Intelligence (XAI) techniques to mitigate cognitive biases in predicting student dropout comparing Explainable Boosting Machine (EBM), Logistic Regression and XGBoost models.Sorav Kumar Singh, Alak Roy and Rajneesh Raushan from India focus their research on underwater wireless sensor networks and propose a Residual Energy-Aware Fuzzy-Based Clustering Algorithm (REAFCA), which presents an enhanced framework to improve network performance and addresses issues with energy usage.Juan Morales-García, Fernando Terroso-Sáenz, Andrés Bueno-Crespo, and José M. Cecilia from Spain discuss in their research the analysis of synthetic timeseries as an enabler to improve region-based human mobility forecasting by applying Generative adversarial network (GANs) to generate synthetic time-series mobility data.Igor Tomičić, Petra Grd, and Andrija Bernik from Croatia present in their research a comprehensive analysis of the integration of artificial intelligence into threat intelligence (TI) systems focusing on its potential to enhance cybersecurity operations by an extensive literature review including machine learning, deep learning, and natural language processing for automating threat detection, classification, and analysis.And last but not least, Daniel Spiekermann from Germany investigates in his research the real-world behaviour of network traffic within virtualized environments to identify the key factors that impact packet dynamics, including VM operations, multi-tenancy, user customization, and hardware adjustments.Enjoy Reading!Best regards,Christian Gütl, Managing Editor-in-Chief

HTML

PDF

DOI: 10.3897/jucs.164745

Authors: Aik Tarkhanyan, Ashot Harutyunyan

Abstract: Several studies have shown that the Dempster–Shafer theory (DST) can be successfully applied to scenarios where model interpretability is essential. Although DST-based algorithms offer significant benefits, they face challenges in terms of efficiency. We present a method for the Dempster-Shafer Gradient Descent (DSGD) algorithm that significantly reduces training time—by a factor of 1.6—and also reduces the uncertainty of each rule (a condition on features leading to a class label) by a factor of 2.1, while preserving accuracy comparable to other statistical classification techniques. Our main contribution is the introduction of a ”confidence” level for each rule. Initially, we define the ”representativeness” of a data point as the distance from its class’s center. Afterward, each rule’s confidence is calculated based on representativeness of data points it covers. This confidence is incorporated into the initialization of the corresponding Mass Assignment Function (MAF), providing a better starting point for the DSGD’s optimizer and enabling faster, more effective convergence. The code is available at https://github.com/HaykTarkhanyan/DSGD-Enhanced.

HTML

PDF

DOI: 10.3897/jucs.164694

Authors: Aram Adamyan, Hovhannes Hovhannisyan, Daniel Radrigan, Nelson Baloian, Ashot Harutyunyan

Abstract: This study presents DSClustering, a novel algorithm that merges clustering validity with interpretability using the Dempster-Shafer theory. Traditional clustering methods like K-means, DBSCAN, and agglomerative clustering, while widely used for their efficiency and accuracy, often fall short in transparency, creating barriers in critical fields such as healthcare, finance, and consumer analytics where decision-making requires clear, interpretable insights. DSClustering aims to bridge this gap by assigning clusters based on belief functions from Dempster-Shafer theory, which allows it to generate rule-based explanations for each data point’s cluster assignment. Through detailed experiments on real-world datasets, including consumer behavior and airline satisfaction data, we evaluate DSClustering against traditional algorithms using key metrics such as Silhouette score, Rand index and Dunn’s index for clustering validity. The results indicate that DSClustering not only performs competitively but also offers a clear interpretative layer, making it suitable for applications where understanding model outputs is as essential as the accuracy of the outputs themselves. This work underscores the increasing importance of interpretability in machine learning, particularly in unsupervised learning, where transparency is typically challenging to achieve. DSClustering demonstrates a promising approach for balancing robust clustering with user-oriented interpretability, potentially encouraging broader adoption of interpretable clustering models in data-critical industries.

HTML

PDF

DOI: 10.3897/jucs.164737

Authors: Leo Buron, Andreas Erbslöh, Gregor Schiele

Abstract: For embedded devices, both memory and computational efficiency are essential due to their constrained resources. However, neural network training remains both computation and memory intensive. Although many existing studies apply quantization schemes to mitigate memory overhead, they often employ stochastic rounding for both inference and gradient computation. Notably, no prior work has explored its advantages exclusively in parameter updates. Here, we in-troduce Quantized Parameter Updates (QPU), which uses stochastic rounding (SQPU) to achieve improved and more stable training outcomes. Our fixed-point quantization scheme quantizes parameters (weights and biases) upon model initialization, conducts high-precision gradient com-putations during training, and applies stochastically quantized updates thereafter. This approach substantially lowers memory usage and enables mostly quantized inference, thereby accelerating calculations. Furthermore, storing quantized inputs for gradient computation reduces memory demands even more. When tested on the FASHION-MNIST dataset, our method matches the Straight-Through Estimator (STE) in performance, delivering 0.92% validation accuracy while consuming just 57% of the memory during training. Accepting a slight 1.5% drop in accuracy yields a 50% memory reduction. Additional techniques include stochastic rounding in inference, the use of higher precision for parameters than for layer outputs to limit overflow, L2 regularization via weight decay, and adaptive learning-rate scheduling for improved optimization across a range of batch sizes.

HTML

PDF

DOI: 10.3897/jucs.164692

Authors: Sergio Peñafiel, Esteban Ramírez, Nelson Baloian, Isabel Saffie, Paulo Luz, Inti Paredes

Abstract: Breast cancer is a significant global health problem, and HER2-positive breast cancer accounts for a substantial proportion of cases. The combination of Trastuzumab and Pertuzumab monoclonal antibodies with chemotherapy has demonstrated effectiveness in achieving pathologic complete response (pCR) among HER2-positive breast cancer patients. This study aims to develop an interpretable machine learning model to predict pCR in patients undergoing this neoadjuvant treatment. Previous studies have explored predictors of pCR and utilized statistical techniques, but no prior research has applied machine learning to this specific treatment. This work proposes a rule-based interpretable method based on Dempster-Shafer theory. The model is trained using a dataset of 390 patients, with 57% achieving pCR. The performance of the model is compared with other classification algorithms, demonstrating its moderate but promising results. This work highlights the importance of combining accuracy and interpretability in healthcare applications, providing insights into the factors influencing treatment response in HER2-positive breast cancer patients.

HTML

PDF

DOI: 10.3897/jucs.164671

Authors: Claudio Gutiérrez-Soto, Marco A. Palomino, Patricio Galdames

Abstract: In recent years, we have witnessed the proliferation of applications that generate thousands of terabytes of data per day, due to the explosive increase in storage capacity across various devices. As a consequence, a new concept called Data Deluge has emerged. Data deluge refers to the situation where the quantity of data generated exceeds the processing power available, and spatio-temporal data is no exception to this phenomenon. In this context, the efficient processing of spatio-temporal queries becomes crucial to address this challenge, as slow query processing can result in obsolete answers, which may lead to errors. Considering this dynamic context of storage and processing, we explore a new online workload algorithm in a distributed parallel environment using hybrid spatio-temporal indexes. This algorithm is able to update the indexes with the most appropriate data, aiming to achieve more efficient query processing. To measure the efficiency of this algorithm, we present its time complexity along with an empirical evaluation of its performance, considering processing time, number of accessed nodes, and communication costs. The empirical results show a significant reduction in processing time, communication costs, and number of accessed nodes.

HTML

PDF

DOI: 10.3897/jucs.165358

Authors: Mohini Sharma, Raghav Sandhane, Jaydeep Rajeshkumar Katariya

Abstract: The metaverse seamlessly integrates physical and digital spaces, enabling AI-driven innovations in virtual interactions, autonomous avatars, and real-time experiences. However, increased reliance on AI brings sweeping cybersecurity challenges, such as adversarial attacks, deep fake impersonation, and AI-driven phishing campaigns. The security of the metaverse is vital for the sustainability of user trust and system integrity. As AI assumes a larger role in virtual environments, proactive cybersecurity measures must be taken to counter emerging threats. This paper introduces DAI-TIRS, a holistic security framework designed to proactively secure the metaverse. DAI-TIRS is the integration of machine learning-based anomaly detection, dynamic honeypots, and predictive threat modeling that detect, classify, and mitigate AI-driven threats in real-time. By utilizing MITRE ATT&CK and the PyTM framework, it constantly learns new emerging threats through advanced behavioral analytics and keeps pace with the adversarial AI model’s evolution. The experimental results from a simulated metaverse environment demonstrate that DAI-TIRS achieves 93% accuracy in threat detection, 90% precision in classifying the severity, and a 36.9% faster threat mitigation response time than the average performance of baseline models, as detailed in the paper. These findings underscore the critical need for adaptive AI-based cybersecurity solutions that will enhance the resilience, trust, and integrity of metaverse ecosystems. This research establishes DAI-TIRS as an advanced cybersecurity framework that has demonstrated its adaptability and effectiveness in countering AI-driven threats across multiple sectors. The source code for DAI-TIRS is available on GitHub: [https://github.com/sharmamohini762/DAI-TIRS-Code]

HTML

PDF

DOI: 10.3897/jucs.167412

Authors: Yanling Chen

Abstract: In this paper, we show the problem of two-party privacy-preserving record linkage (PPRL) can be seen as an identification problem in Information Theory. We propose to apply the identification codes that are designed for identification via channels to the problem of PPRL, due to their advantage in the performance analysis, especially on a quantitative evaluation of the privacy. Note that for the PPRL, linkage quality is typically evaluated experimentally, whilst for privacy, there are so far no commonly accepted privacy measures available that allow an objective evaluation. Our approach of identification code provides an objective evaluation on both linkage quality and privacy based on parameters of employed identification codes.

HTML

PDF

DOI: 10.3897/jucs.164654

Authors: Wolfram Luther, A. J. Han Vinck

Abstract: This volume presents a conference paper selection from the 4th Workshop on Collaborative Technologies and Data Science in Smart City Applications (CODASSCA 2024): Data Science and Reliable Machine Learning, held in Yerevan, Armenia, October 3-6, 2024, https://codassca2024.aua.am/. The special issues guest editors invited five groups of authors from Armenia, Chile, Germany, the UK, and the USA to submit enlarged versions of their CODASSCA 2024 papers There was also a J.UCS open call so that any author could submit papers on the highlighted subjects. The invitation to review the 16 contributions received was accepted by 16 experts, and, after three rounds, seven articles were finally accepted for publication in the special issue.

HTML

PDF

DOI: 10.3897/jucs.135666

Authors: Davut Çulha

Abstract: Widespread adoption of blockchain technologies requires scalability. To achieve scalability, various methods are applied, including new consensus algorithms, directed acyclic graph solutions, sharding solutions, and off-chain solutions. Sharding solutions are particularly promising as they distribute workload across different parts of the blockchain network. Similarly, directed acyclic graphs use graph data structures to distribute workload effectively. In this work, a binary tree data structure is used to enhance blockchain scalability. Binary trees offer several advantages, such as the ability to address nodes with binary numbers, providing a straightforward and efficient method for identifying and locating nodes. Each node in the tree contains a block of transactions, which allows for transactions to be directed to specific paths within the tree. This directionality not only increases scalability by enabling parallel processing of transactions but also ensures that the blockchain can handle a higher volume of transactions without becoming congested. Moreover, transactions are decomposed into transaction elements, improving the immutability of the binary tree blockchain. This novel decomposition process helps to minimize the computational overhead required for calculating account balances, making the system more efficient. By breaking down transactions into their fundamental components, the system can process and verify transactions more rapidly and accurately. This approach effectively realizes implicit sharding using a binary tree structure, distributing the processing load more evenly and reducing bottlenecks. The proposed method is simulated to assess its performance. Experimental results demonstrate the method's scalability, showing that it can handle a significantly higher transaction throughput compared to traditional blockchain structures.

HTML

PDF

DOI: 10.3897/jucs.130590

Authors: Eugen Valentin Butilă, Mihai Burlacu, Răzvan Gabriel Boboc, Robertas Damaševičius

Abstract: As a relatively new technology that has gone through several iterations in the last decade, virtual reality (VR) applications have been used in a plethora of activities pertaining to various sciences, including anthropology. In this paper, we expound a bibliometric analysis of the reviews and research articles regarding the use of VR applications in anthropology between 2010 and 2023. The analysed publications were obtained from the Scopus database, and Microsoft Excel and VOSViewer were used to analyse the data. Utilizing bibliometric methods, the analysis encompasses a thorough examination of scholarly publications, identifying and scrutinizing prominent journals, prolific authors, affiliated institutions, and key research themes within the realm of VR applications in anthropology. The objective is to provide a systematic and insightful overview of the evolution, current state, and emerging trends in the integration of VR within the anthropo-logical discourse, shedding light on the interdisciplinary nature and impact of this innovative technology on anthropological research and practice.

HTML

PDF

DOI: 10.3897/jucs.121782

Authors: Muthukumaran N., Vignesh A.

Abstract: Artificial Intelligence (AI) chatbots have become indispensable for natural language interaction, with transformer-based models driving advances in conversational agent (CA) systems. While state-of-the-art models like RoBERTa, ALSI-Transformer, MEDN-Transformer, SG-Net Transformer, BART, and GPT-3 have achieved remarkable context understanding and response generation, they still face limitations. These include challenges with context retention over extended interactions, syntactic ambiguities, and bias propagation from training data, raising concerns for ethical and interpretable AI systems. This research proposes an advanced transformer model, the Improved T5 (IT5), designed to address these issues. IT5 integrates Aggregated Multi-Head Attention (AMHA) and Bidirectional Long Short-Term Memory (BiLSTM) into the T5 framework to improve context retention, response nuance, and bias reduction. Additionally, a retraining mechanism updates IT5’s knowledge base with every 50 new question-answer pairs, ensuring fairness and relevance in chatbot responses. The model's performance was rigorously tested on the NarrativeQA, SQuAD, MS MARCO, and InsuranceQA datasets, where IT5 achieved top BLEU scores of 0.7533, 0.7012, 0.7155, and 0.7373, respectively. It consistently demonstrated lower WER scores of 0.1957, 0.2106, 0.2254, and 0.1953, and higher ROUGE-L scores of 0.8875, 0.8991, 0.8731, and 0.8933 across these datasets. IT5 also exceeded in accuracy (0.98, 0.97, 0.96, 0.97), precision (0.96, 0.98, 0.97, 0.96), recall (0.95, 0.96, 0.97, 0.98), and F1 scores (0.95, 0.98, 0.96, 0.96), surpassing six state-of-the-art models, namely RoBERTa, ALSI-Transformer, MEDN-Transformer, SG-Net Transformer, BART, and GPT-3. The findings demonstrate IT5’s superior ability to generate meaningful, fair, and high-quality responses, establishing it as a frontrunner for robust and ethical conversational AI across various applications.

HTML

PDF

DOI: 10.3897/jucs.129515

Authors: Juan-Sebastián González-Sanabria, Cristian Pinto, Jhon Zuñiga, Hugo Ordoñez, Xiomara Blanco

Abstract: Gender-based violence remains a persistent social challenge in Colombia. Despite efforts to address it, statistics show a steady increase year after year. This study addresses the need for predictive solutions by introducing a Machine Learning model using XGBoost, chosen for its high performance in classification tasks with complex datasets. The model is trained on data collected from the department of Santander, Colombia, aiming to predict gender-based violence incidents based on specific socio-demographic and situational features. The motivation behind using XGBoost lies in its ability to handle diverse data types and produce accurate, interpretable results. Key influential features in the model’s predictions were identified, including the context of the incidents and the relationship between victim and the perpetrator, underscoring the importance of situational as well as individual factors. The model achieved promising results, with an accuracy, precision, recall, and F1 score exceeding 84% demonstrating its potential to effectively predict and contribute to preventing gender-based violence in the region. This approach not only represents a proactive response to a critical social challenge but also offers a framework that could be applied in similar contexts at the national and international levels.

HTML

PDF

DOI: 10.3897/jucs.165499

Authors: Christian Gütl

Abstract: Dear Readers,It gives me great pleasure to announce the seventh regular issue of 2025. In this issue, 4 papers by 12 authors from 6 countries – Colombia, India, Lithuania, Romania, Spain, Turkiye – cover various topical and novel aspects of computer science. As always, I would like to thank all the authors for their sound research and the editorial board and guest reviewers for their extremely valuable review effort and suggestions for improvement. I also want to thank the readers for their interest in our articles, which is reflected in the increasing number of accesses and PDF downloads. These contributions, together with the generous financial support of the KOALA initiative, sustain the quality of our journal. n a continuous effort to further strengthen our journal, I would like to expand the editorial board: If you are a tenured associate professor or above with a strong publication record, you are welcome to apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and trends. In the seventh regular issue, I am very pleased to introduce the following 5 accepted articles: In a collaboration between researchers from Colombia and Spain, Juan-Sebastián González-Sanabria, Cristian Pinto, Jhon Zuñiga, Hugo Ordoñez, and Xiomara Blanco focus on a XGBoost Classifier-Based Model to predict the nature of gender-based violence based on specific socio-demographic and situational features.Muthukumaran N and Vignesh A from India present enhancements of chatbot responses by addressing challenges such as context retention over extended interactions, syntactic ambiguities and bias propagation from training data. They propose an advanced transformer model, the Improved T5 (IT5), to solve these issues.In a collaboration between researchers from Romania and Lithuania, Eugen Valentin Butilă, Mihai Burlacu, Răzvan Gabriel Boboc, and Robertas Damaševičius discuss the findings of a bibliometric analysis of reviews and research articles on the use of VR applications in anthropology between 2010 and 2023.Last but not least, Davut Çulha from Turkiye addresses scalability aspects through a binary tree blockchain of decomposed transactions, which can reduce the computational overhead required to calculate account balances and make the system more efficient.Enjoy Reading!Best regards,Christian Gütl, Managing Editor-in-Chief

HTML

PDF

DOI: 10.3897/jucs.128450

Authors: Ji Woong Yoo, Kyoung Jun Lee, Arum Park

Abstract: The advancement of AI has heightened the significance of ethical concerns, particularly in managing negative user feedback like malicious comments, necessitating thoughtful deliberation. The focus of this research is to explore the potential of deep learning techniques in addressing these issues and enhancing the ethical nature of AI systems. Specifically, we investigated the collection and processing of news comment data using Long Short-Term Memory (LSTM) algorithm and Word2Vec model. The primary objective was to evaluate how deep learning techniques can improve the quality of data obtained from news comments. Our findings demonstrate that deep learning models surpass CleanBot in accuracy and block rates for handling negative user comments, including malicious ones, enabling organizations to effectively manage such comments in online communities using AI-based methods. This study adds to the existing research by showing how advanced deep learning techniques can effectively identify and classify harmful comments by analyzing complex language patterns.

HTML

PDF

DOI: 10.3897/jucs.129692

Authors: Pedro Henrique Dias Valle, Elisa Yumi Nakagawa

Abstract: Context: Software systems have become increasingly large and complex, and are required in several critical domains, including Industry 4.0, the military, smart cities, and transportation. Consequently, the architectural design of these systems becomes considerably complicated, in addition to requiring interoperability among diverse systems that sometimes comprise them. Problem: Although many interoperability architectural solutions exist, software architects have struggled to comprehend, analyze, and select the most suitable ones to solve their problems. Objective: This work provides a catalog of the main interoperability architectural solutions (patterns, styles, tactics, and approaches) for addressing the four levels of interoperability (namely, technical, syntactic, semantic, and organizational) to resolve interoperability issues in software systems. Method: 65 studies found systematically in the scientific literature were deeply examined and provided evidence to define our catalog, which comprises interoperability issues and architectural solutions to address these problems. Results: As a contribution, this catalog could help software architects better decide which architectural solutions could solve each interoperability issue in their integration projects.

HTML

PDF

DOI: 10.3897/jucs.124812

Authors: Roger Vieira, Kleinner Farias

Abstract: Wearable devices capable of capturing psychophysiological data have emerged as a tangible reality. Recent academic investigations emphasize the pivotal role of developers’ cognitive indicators, such as attention levels and cognitive load, in influencing their effectiveness in understanding and managing code-related tasks. However, existing Integrated Development Environments (IDEs) and code editors, such as Visual Studio (VS) Code, lack comprehensive contextual information on cognitive indicators alongside source code. This article, therefore, introduces CognIDE, a novel tool-supported methodology aimed at seamlessly integrating psy-chophysiological data linked to cognitive indicators into VS Code. Addressing this crucial gap, CognIDE enriches VS Code by offering actionable contextual cues alongside dynamic source code. The evaluation of CognIDE, involving a survey with six industry professionals and in-depth interviews, examined its perceived utility, ease of use, and real-world applicability. Encouragingly, professionals demonstrated high acceptance, indicating CognIDE’s potential to identify and prioritize code segments with specific cognitive indicators, notably related to bugs or code comprehension issues. This underscores CognIDE’s promise in improving code review processes.

HTML

PDF

DOI: 10.3897/jucs.124087

Authors: Michele Berti, Matheus Camilo da Silva, Sebastiano Saccani, Sylvio Barbon Junior

Abstract: Synthetic data generation is a promising alternative to traditional data anonymization, with Variational Autoencoders (VAEs) excelling at generating high-quality synthetic tabular datasets. However, VAE hyperparameter selection is often computationally expensive or suboptimal. We propose a meta-learning (MtL) method for hyperparameter recommendation, which achieves competitive performance to state-of-the-art Bayesian Optimization (BO) with median AUC values of 0.660 ± 0.038 (MtL) and 0.650 ± 0.041 (BO), showing no statistically significant difference. Notably, our approach reduces configuration time to under three minutes, compared to BO’s multi-hour requirement, while also enabling incremental improvements through new data integration. This combination of efficiency, adaptability, and performance establishes MtL as a practical solution for hyperparameter tuning in synthetic data generation.

HTML

PDF

DOI: 10.3897/jucs.162422

Authors: Christian Gütl

Abstract: Dear Readers,It gives me great pleasure to announce the sixth regular issue of 2025. In this issue, 4 papers by 11 authors from 3 countries - Brazil, Italy, Republic of Korea - cover various topical aspects of computer science. In a continuous effort to further strengthen our journal, I would like to expand the editorial board: If you are a tenured associate professor or above with a strong publication record, you are welcome to apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and trends.As always, I would like to thank all the authors for their sound research and the editorial board members and guest reviewers for their extremely valuable review effort and suggestions for improvement. I also want to thank the readers for their interest in our articles, which is reflected in the consistently high number of user accesses and PDF downloads. These contributions, together with the generous support of the KOALA initiative, maintain the quality of our journal.In the sixth regular issue, I am very pleased to present the following 4 accepted articles: Michele Berti, Matheus Camilo da Silva, Sebastiano Saccani, and Sylvio Barbon from Italy focus their research on synthetic data generation as an alternative to traditional data anonymization based on variational autoencoders to generate high-quality synthetic tabular datasets.Roger Vieira and Kleinner Farias from Brazil introduce in their research CognIDE, a tool-supported methodology that aims to seamlessly integrate psychophysiological data linked to cognitive indicators into VS Code by offering actionable contextual cues alongside dynamic source code.Pedro Henrique Dias Valle and Elisa Yumi Nakagawa from Brazil discuss in their research a catalog of the main interoperability architectural solutions for addressing the four levels of interoperability - namely technical, syntactic, semantic, and organizational – for solving interoperability issues in software systems by analyzing 65 studies from the scientific literature.Ji Woong Yoo, Kyoung Jun Lee and Arum Park from the Republic of Korea explore the potential of deep learning techniques - Long Short-Term Memory (LSTM) algorithm and Word2Vec model – for cleansing malicious comments from users, and enhancing the ethical nature of AI systems.Enjoy Reading!Christian Gütl, ManagingEditor-in-Chief

HTML

PDF

DOI: 10.3897/jucs.130186

Authors: Yang Zhang, Ziwen Wei, Zhihua Liu, Xiaolong Wu, Junchao Qian

Abstract: Purpose: Incorrect patient positioning during radiotherapy can significantly impact treatment efficacy and pose potential risks. This study aims to develop a model that can rapidly and effectively monitor the patient’s postures during radiotherapy sessions using real-time video. Methods: The neural network utilized in this research employed a two-stream architecture, consisting of spatial and temporal streams. For the spatial stream, RGB frames from the videos were directly used as input. In the temporal stream, representative frames were extracted from the video to construct stacked grayscale 3-channel images (SG3I) frames. This approach enabled capturing motion information through a large-scale dataset pre-trained 2D convolutional neural network (CNN), eliminating the need for computationally expensive optical flow calculations. Additionally, an improved lightweight network architecture was employed. The model was trained and tested using volunteer videos collected from a radiotherapy center in a hospital. Results: The results demonstrated that the proposed model outperforms existing methods in terms of detection accuracy while achieving higher efficiency in frame generation. Conclusion: In this study, we introduced a cost-effective and highly accurate method for recognizing patient’s postures during radiotherapy. This approach could be readily deployed in any radiotherapy facility, ensuring treatment precision and patient safety.

HTML

PDF

DOI: 10.3897/jucs.129624

Authors: Thuy Phuong Khuat, Trang Van, Hoang Thien Van

Abstract: Deep learning methods are increasingly used in automated plant species classification systems to support biodiversity conservation and ecological monitoring, particularly for medicinal plants. This study presents a novel approach to plant leaf recognition by integrating the Vision Transformer (ViT) model with the OSSGabor filter, termed the OGViT method. The OSSGabor filter is a leaf feature extraction technique that combines the responses of Gabor filters in 16 directions and optimizes their parameters using the Structural Similarity Index Measure (SSIM). These features capture intricate details such as leaf veins, texture, and frequency variations, which are essential for enabling ViT to fully leverage deep learning for leaf recognition. Experimental results on four public datasets—Swedish Leaf, Flavia, Folio, and UCI Leaf—demonstrate that the OGViT method outperforms state-of-the-art approaches, achieving accuracy scores of 100%, 100%, 100%, and 98.88%, respectively, with a 20% testing set and an 80% training set. This performance highlights the effectiveness of the proposed method for plant classification, offering a robust tool with potential applications in agriculture and biodiversity conservation.

HTML

PDF

DOI: 10.3897/jucs.129212

Authors: M. Priadarsini, J. Akilandeswari

Abstract: In the expanding field of sentiment analysis, the integration of personality prediction into aspect-based sentiment analysis (ABSA) represents a novel and promising approach to enhance the accuracy and depth of sentiment detection. This paper proposes a unique framework that leverages the Big Five personality traits (Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism) alongside Long Short-Term Memory (LSTM) networks, under a multitask learning paradigm, to improve the performance of ABSA. This is, to the best of knowledge, the first work considering the use of personality traits as auxiliary tasks in order to capture the manifold subtle ways in which personality would influence the expression of sentiment towards the different aspects of products or services. And then, model uses the LSTM component to model the sequential character of the text, which makes the extraction accurate in terms of the aspect terms and sentiment polarities. The proposed model designs a multitask learning strategy simultaneously to predict sentiments and personality traits. Such joint learning will allow enhancing the model's understanding of textual context and sentiment expression. Thorough experiments on many benchmark datasets show that the proposed approach is competitive with the state of the art for the aspect-based sentiment analysis and provides some of the deepest insights into personality predictions. Model has obtained F1-scores of 79.78%, 83.67%, and 88.80 % on the Twitter, Laptop, and Restaurant datasets, respectively. These results highlight a significant improvement over existing methods in the literature. For instance, our model outperformed traditional approaches like RAM, which achieved 69.36% on the Twitter dataset, and even advanced techniques such as DualGCN+Bert, which scored 77.4% on Twitter. It can be generally concluded that this research finally opens the way to a new and meaningful opportunity for sentiment analysis applications: integrated into ABSA models, personality prediction advances applications ranging from personalized recommendation systems to the nuance market analysis tools. As far as we know this study is the first attempt to utilise personality feature to enhance sentiment prediction tasks.

HTML

PDF

DOI: 10.3897/jucs.125999

Authors: Ryma Abassi

Abstract: As governments and organizations seek to strengthen cybersecurity measures, ethical considerations play a crucial role in shaping effective and responsible policies. This research article explores the ethical dimensions of cybersecurity policymaking, focusing on the balance between security imperatives and individual privacy rights. Drawing on principles of ethics, human rights, and legal frameworks, the article discusses challenges and dilemmas faced by policymakers in ensuring cybersecurity without compromising privacy and civil liberties. It proposes a set of ethical guidelines and best practices for designing and implementing cybersecurity policies that are both effective and respectful of fundamental rights and values.

HTML

PDF

DOI: 10.3897/jucs.127870

Authors: Ícaro Prado Fernandes, Luiz Eduardo Galvão Martins

Abstract: Building quality software, that is, suitable for use and meeting user needs, is one of the biggest challenges in the software industry. Although it is possible to guarantee the proper functioning of software through testing activities, such activities are exhaustive in nature, as it is impossible to test all inputs of a minimally complex program. This work proposes a method to prioritize test cases based on human knowledge using a combination of factors evaluated in an assessment answered by 29 software industry professionals and 5 academics. The assessment confirmed that the proposed factors are relevant. Finally, a practical example that prioritizes test cases for a banking application was carried out and it was observed that the proposed method works properly.

HTML

PDF

DOI: 10.3897/jucs.158922

Authors: Christian Gütl

Abstract: Dear Readers,It gives me great pleasure to announce the fifth regular issue of 2025. I would like to thank all the authors for their sound research papers and the editorial board and our guest reviewers for their extremely valuable reviews and suggestions for improvement. These contributions and the generous support of the KOALA consortium members enable us to run our journal and maintain its quality. I would also like to thank our broader community for reading and incorporating sound J.UCS papers into their research.Still, I would like to expand our editorial board: If you are a tenured associate professor or above with a good publication record, please apply to join our editorial board. We are also interested in receiving high-quality proposals for special issues on new topics and emerging trends. In this regular issue, I am very pleased to introduce 5 papers by 13 authors from 5 countries: Brazil, China, India, Tunisia, Vietnam. Icaro Prado Fernandes and Luiz Eduardo Galvão Martins from Brazil propose in their article a method to prioritize test cases based on human knowledge using a combination of factors evaluated in an assessment answered by 29 software industry professionals and 5 academics. Ryma Abassi from Tunisia builds on the principles of ethics, human rights and legal frameworks in his research to address the challenges and dilemmas faced by policymakers when it comes to ensuring cybersecurity without compromising privacy and civil liberties and proposes a set of ethical guidelines and best practices for designing and implementing cybersecurity policies. M. Priadarsini and J. Akilandeswari from India propose a unique framework in their research that leverages the big five personality traits alongside long short-term memory (LSTM) networks under a multitask learning paradigm to improve the performance of aspect-based sentiment analysis. Thuy Phuong Khuat, Trang Van and Hoang Thien Van from Vietnam discuss in their research an approach to plant leaf recognition by integrating the vision transformer (ViT) model with the OSSGabor filter, referred to as the OGViT method, and analyze the performance on four public datasets (Swedish Leaf, Flavia, Folio, and UCI Leaf) that outperforms state-of-the-art approaches. Yang Zhang, Ziwen Wei, Zhihua Liu, Xiaolong Wu and Junchao Qian from China introduce in their study a cost-effective and highly accurate method for recognizing patient postures during radiotherapy based on stacked grayscale 3-channel images. Enjoy Reading!Best regards,Christian Gütl, Managing Editor-in-Chief

HTML

PDF

DOI: 10.3897/jucs.137103

Authors: Hendrik Hendrik, Silmi Fauziati, Adhistya Erna Permanasari

Abstract: Knowledge graphs are a powerful way to represent and organize complex knowledge. They are used in many fields, like healthcare and finance. They allow for more insightful decision-making and discoveries. However, the quality of knowledge graphs depends heavily on their sources. Current methods for evaluating these sources are often slow and not scalable. They struggle to keep up with the large amount of online information. We created a new tool to address this problem. Our tool uses Large Language Models (LLMs) to assess online sources quickly. It evaluates websites based on credibility, relevance, content quality, coverage, comprehensiveness, and accessibility. We tested our tool on Halal tourism websites in Japan. We compared LLM evaluations with human expert judgments. Our comprehensive analysis revealed that certain LLM models, particularly GPT-3.5-turbo, GPT-4, and Mixtral-8x7B-Instruct-v0.1, showed strong correlation with human evaluations. Using a temperature setting of 0.4, these models demonstrated consistent and reliable performance across multiple evaluation runs. Our structured evaluation framework, incorporating weighted criteria validated through both expert input and statistical analysis, provides a robust foundation for automated source assessment. While some models showed varying performance across different criteria, our findings suggest that careful model selection and potential ensemble approaches could optimize evaluation accuracy. Our work contributes significantly to improving knowledge graph construction by demonstrating the viability of LLM-based source evaluation, while also identifying key areas for future research in scalability, cross-domain validation, and automated optimization.

HTML

PDF

DOI: 10.3897/jucs.131543

Authors: Sarah Khater, Magda B. Fayek, Mayada Hadhoud

Abstract: Human activity recognition (HAR) is a challenging computer vision problem that requires recognizing and categorizing human actions using spatiotemporal data. In recent years, ConvLSTM has shown distinctive advances in manipulating spatiotemporal data. ConvLSTM-based architectures, as any deep learning architecture, require deciding on many hyperparameters apart from trainable weights. State-of-the-art designs for general purpose datasets already exist, but specific purpose applications require architecture designs that perform well on application-dependent datasets. The design of such architectures requires either many trials and errors, which consume time and resources, or an experienced architect. Neural architecture search (NAS) meth-ods have been introduced to automate the design process and address the challenge of relying on expert knowledge when creating neural architectures. NAS enables rapid prototyping and experimentation, reducing the time spent on trial and error in manual design. One of the leading approaches in NAS is Genetic Algorithm (GA), which plays a significant role in optimizing neu-ral architectures. In this paper, a novel GA-based approach is proposed to automatically design ConvLSTM-based architectures from scratch for HAR applications. Our approach is based on multi-objective GA that maximizes recognition accuracy and minimizes the number of trainable parameters and overfitting measure. The experiments are held on KTH, Weizmann, and UCF Sports datasets. The best classification accuracies from the generated models are 97.92%, 96.77%, and 94.87% for KTH, Weizmann, and UCF Sports datasets, respectively. The experimental results show that the automatically generated models with the proposed approach outperform some of the state-of-the-art manually designed ConvLSTM-based architectures with percentages up to 9.92%, 5.77% and 23.64% for KTH, Weizmann, and UCF Sports, respectively. We also compared our approach with other NAS approaches. Our approach is found to outperform some of the introduced approaches with percentages approximately 2%, 11%, and 4% for KTH, Weizmann, and UCF Sports, respectively.

HTML

PDF

DOI: 10.3897/jucs.129859

Authors: Manpreet Singh, Jitender Kumar Chhabra

Abstract: Fault prediction of multimedia software is necessary to develop good quality multimedia software because integrating various multimedia heterogeneous components in a software system usually generates many faults. So, this research article proposes a new fault prediction model based on the decided threshold values of structural features. These features are captured using metrics specifically identified for multimedia software and weighted suitably based on the behavior of the components dealing with multimedia handling. The threshold values are optimized using the genetic algorithm (GA). This paper also proposes a GA-based technique to combine multiple features using conjunction (AND) and disjunction (OR) operators while finding threshold values. Finally, the proposed model is tested for cross-project software fault prediction on selected six multimedia software and validated on three other general software datasets. Results show that our identified thresholds-based model performs excellently for multimedia software and satisfactorily over other general software.

HTML

PDF

DOI: 10.3897/jucs.128779

Authors: Carlos Goetz, Rodrigo Simon Bavaresco, Wesllei Felipe Heckler, Gustavo Lazarotto Schroeder, Rafael Kunst, Jorge Luis Victória Barbosa

Abstract: Stress is a mental disorder responsible for impacting the industry through psychosomatic illnesses, loss of productivity, and accidents caused by stressful workplaces. Conversely, the literature indicates that fostering mental well-being among workers can boost motivation and performance while alleviating symptoms of stress. The fourth industrial revolution incorporated technologies into work that allowed the automation of processes and control of environments. The fifth revolution introduced the application of research and innovation aimed at a human-centered consciousness, enabling the advancement of mental health through sensors and wearables. Despite advancements in stress classification technologies, there remain opportunities for further research into identifying stress motivators within industrial work environments. In this sense, this paper proposes an ontology to identify stressors considering personal and environmental data, allowing knowledge generation related to work stressors for mitigating the problem. The methodology utilized in this ontology development consisted of seven stages and two evaluation phases. The findings addressed four key questions related to competency as outlined in the model. The results revealed potential stressful scenarios, including the timing of occurrence, shared locations, environmental factors, and identifying groups experiencing moments of stress. This study presents as a scientific contribution the first ontology to address the identification of work-related stressors in the industrial environment.

HTML

PDF

DOI: 10.3897/jucs.156450

Authors: Christian Gütl

Abstract: Dear Readers,I am very happy to announce the fourth regular issue of 2025. In this issue, 4 articles by 14 authors from 4 countries (Brazil, Egypt, India, Indonesia) cover a variety of topical research aspects in computer science. Allow me to express my appreciation to all the authors for their sound research work and to thank the editorial board and guest reviewers for their extremely valuable reviews and suggestions for improvement. This continuous stream of relevant and novel contributions, along with the generous support of the KOALA initiative, helps to maintain the quality of our journal.In the ongoing effort to further strengthen our journal, I would like to expand the editorial board: If you are a tenured associate professor or above with a strong publication record, you are welcome to apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and trends. Please consider yourself and encourage your colleagues to submit high-quality articles or special issues for our journal.In the fourth regular issue, I am very pleased to introduce the following four accepted articles: In their paper, Carlos Goetz, Rodrigo Simon Bavaresco, Wesllei Felipe Heckler, Gustavo Lazarotto Schroeder, Rafael Kunst, and Jorge Luis Victória Barbosa from Brazil propose an ontology to identify stressors, considering personal and environmental data, which makes it possible to generate knowledge about work stressors in order to mitigate the problem utilizing a methodology consisting of seven stages and two evaluation phases. In their research, Manpreet Singh and Jitender Kumar Chhabra from India deal with fault prediction of multimedia software which integrates various multimedia heterogeneous components by a GA-based technique to combine multiple features using conjunction (AND) and disjunction (OR) operators while finding threshold values. Sarah Khater, Magda B. Fayek, and Mayada Hadhoud from Egypt present their research on human activity recognition (HAR) and discuss a GA-based approach to automatically generate ConvLSTM architectures for human activity recognition. And last but not least, Hendrik Hendrik, Silmi Fauziati, and Adhistya Erna Permanasari from Indonesia introduce an enhancing knowledge graph construction with automated source evaluation utilizing large language models.Enjoy Reading!Best regards,hristian Gütl, Managing Editor-in-Chief

HTML

PDF

DOI: 10.3897/jucs.127703

Authors: Mahir Shadid, Mushfiqus Salehin Afnan, Rashed Mustafa, M. Jamshed Alam Patwary

Abstract: Early detection of anxiety disorders in a non-intrusive manner is crucial, as these conditions can profoundly impact an individual’s health and daily functioning. Traditional approaches relying solely on unimodal data often fall short, potentially introducing bias and inaccuracies. TI-Fusion is a novel late multimodal fusion technique that integrates text and image data for a unified reliable outcome, overcoming limitations in existing methods. The primary advantage of TI-Fusion is its non-intrusive nature, ensuring patient comfort by avoiding invasive methods while still delivering robust diagnostic capabilities. The study utilizes six advanced machine learning algorithms (Gaussian Naive Bayes, XGB Classifier, K-Neighbors, SVM, Decision-Tree, and RandomForest) for data classification, pattern recognition, and predictive accuracy. Concurrently, image data from the KDEF and CK+ datasets was processed through a Convolutional Neural Network (CNN) enhanced with a Real Gabor filter, which is particularly adept at capturing textures, edges, and complex visual patterns necessary for precise image analysis and recognition. By employing a late multimodal fusion approach, TI-Fusion integrates the outcomes of models trained on distinct data modalities, yielding a more comprehensive and accurate prediction than unimodal methods. This technique not only surpasses existing multimodal approaches but also achieves a commendable final accuracy rate of 92.38%, demonstrating its effectiveness in enhancing the early detection of anxiety disorders.

HTML

PDF

DOI: 10.3897/jucs.122380

Authors: Maroua Benleulmi, Ibtissem Gasmi, Nabiha Azizi, Nilanjan Dey

Abstract: Recommender systems have a pivotal function in delivering customized and pertinent suggestions to clients on the basis of their preferences and activities. The present paper presents a thorough overview of deep learning-based recommender systems, explores their application to enhance performance, and overcomes limitations. The survey encompasses fundamental models of recommender systems; moreover, it also delves into key deep learning models. This discussion focuses on the effective integration of deep learning techniques into recommender systems. Real-world applications highlight the effectiveness of these approaches in capturing complex and nonlinear patterns from large-scale data. This paper concludes by reflecting on challenges encountered in this research field and outlines potential future directions, offering valuable insights for academics and professionals in the field of recommender systems based on deep learning.

HTML

PDF

DOI: 10.3897/jucs.121757

Authors: Phan Minh Nhat, Ngo Le Huy Hien, Dinh Minh Toan, Le Viet Hung, Phan Binh, Phung Thi Anh, Nguyen Thi Hoang Phuong, Nguyen Van Hieu

Abstract: The examination of Near Infrared Reflectance Spectroscopy (NIR) in cattle and poultry fertilizers provides a viable solution for determining optimal fertilizer composition for crop growth while mitigating adverse impacts on soil and groundwater quality. In recent studies, conventional machine learning models combined with spectral analysis have been used to ascertain cattle and poultry fertilizer concentrations. However, these traditional machine learning models encounter challenges in achieving data generalization, resulting in suboptimal prediction accuracy. To address this issue, this study proposes a synthesized machine learning model named EBAR (Error Based Accumulation Regression), which exhibits a commendable coefficient of determination, with an average R2 = 0.865 across 7 chemical substances, surpassing the performance of existing traditional machine learning models. Additionally, a Backward Elimination technique is designed to identify crucial wavelength ranges for monitoring component concentrations. The research outcome is promising and acts as a novel benchmark for later models in determining component concentrations through NIR spectroscopy. Future research gears toward expanding datasets and increasing samples of fertilizers, extending examined wavelength, and improving the model’s efficiency to apply to various types of foods, including seafood, vegetables, and fruits.

HTML

PDF

DOI: 10.3897/jucs.121457

Authors: Liliana Favre

Abstract: Metamodeling is a central concept in Model Driven Engineering (MDE). An important consideration in metamodeling is that secure metamodels are a prerequisite for secure software, since errors in a metamodel lead to errors in its instances (models). Formal methods can help solve this problem by providing systematic and rigorous techniques for reducing ambiguities and inconsistencies in the specification of metamodels. The goal of this article is to present a unified formal framework for metamodeling in the context of MDE, essentially based on MOF, the metamodeling foundation of the OMG industry standards. It is based on the Nereus metamodeling language and includes transformers for translating both MOF metamodels to Nereus metamodels and Nereus metamodels to MOF metamodels, with some prospects for future industrial use of these results. The Nereus language can be seen as a concrete syntax for MOF, extended by additional properties expressed by axioms. Transformers are defined starting from systems of transformation rules that allow automation of processes. An original real-world case in the context of model-driven reverse engineering is described.

HTML

PDF

DOI: 10.3897/jucs.106159

Authors: Neha Kumari, Rajeev Kumar

Abstract: The two most significant yet complex elements of Java generics are wildcards and type argument inference. Both processes rely on the compiler. Even though type argument inference and wildcard execution are implicit processes, a programmer should be aware of them to make the most of the features. A compiler error message tells much about the code and the process mechanism. If the error message is unambiguous and sound, it is easy for the programmer to debug the code. However, in the context of wildcard-type argument inference, the current Javac compiler emits cryptic and imprecise error messages. A programmer may get confused about the inference outcome and failure, so it will be difficult to resolve the errors easily. In this paper, we propose a few additions to the current Wildcard-based type inference algorithm to get detailed and valuable error messages. We implement a Java compiler plugin tool based on the proposed algorithm. The plugin can be easily executed through the Java command line. It gives a comprehensive error message that aids programmers in resolving errors more effectively.

HTML

PDF

DOI: 10.3897/jucs.153315

Authors: Christian Gütl

Abstract: Dear Readers,It gives me great pleasure to announce the third regular issue of 2025. I would like to thank all the authors for their sound research and the editorial board and guest reviewers for the extremely valuable reviews and suggestions for improvement. These contributions together with the support of the community enable us to run our journal and maintain its quality.I would still like to expand our editorial board: If you are a tenured associate professor or above with a good publication record, please apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and emerging trends.In this regular issue, I am very pleased to present 5 accepted papers by 19 authors from 6 countries: Algeria, Argentina, Bangladesh, India, United Kingdom, and Vietnam.In the first paper Neha Kumari and Rajeev Kumar from India address improvements of compiler error messages in the context of wildcard-type argument inference in Java programs by additions to the current wildcard-based type inference algorithm to get detailed and valuable error messages. Liliana Favre from Argentina discusses research on a unified formal framework for metamodeling in the context of MDE, which is based on the Nereus metamodeling language and includes transformers for translating MOF metamodels to Nereus metamodels and Nereus metamodels to MOF metamodels. In their joint work between researchers from Vietnam and the UK, Phan Minh Nhat, Ngo Le Huy Hien, Dinh Minh Toan, Le Viet Hung, Phan Binh, Phung Thi Anh, Nguyen Thi Hoang Phuong, and Nguyen Van Hieu introduce their research on detecting concentrations by Near Infrared Reflectance Spectroscopy (NIR) in cattle and poultry fertilizers by a synthesized machine learning model named EBAR (Error Based Accumulation Regression) combined with a backward elimination technique designed to identify crucial wavelength ranges for monitoring component concentrations. In another collaborative research between Algeria and India, Maroua Benleulmi, Ibtissem Gasmi, Nabiha Azizi, and Nilanjan Dey present an overview of deep learning-based recommender systems, explore their application to enhance performance, and discuss their limitations. Last but not least, Mahir Shadid, Mushfiqus Salehin Afnan, Rashed Mustafa, and M. Jamshed Alam Patwary from Bangladesh highlight their research on a multimodal fusion algorithm for non-intrusive anxiety detection based on TI-Fusion, a multimodal fusion technique that integrates text and image data for a unified reliable outcome and overcomes the limitations of other existing methods. Enjoy Reading!Best regards, Christian Gütl, Managing Editor-in-ChiefGraz University of Technology, Graz, Austria

HTML

PDF

DOI: 10.3897/jucs.150245

Authors: Fernando Ornelas, Alejandro H. García, Alejandro Santiago, Salvador Ibarra Martínez, José Antonio Castán Rocha, Fausto Balderas, Julio Laria-Menchaca, Mayra Guadalupe Treviño-Berrones

Abstract: The Internet Shopping Optimization Problem with multiple item Units (ISHOP-U) is a recently proven NP-Hard variant of the classical ISHOP, which considers buying more than one unit of the same product. In this work, we propose a new set of instances where the prices of the products follow a Gaussian distribution, which is more realistic in a competitive market than the original instances with random uniform prices. We compute the optimal values of the previous uniform and new Gaussian instances using an Integer Programming formulation in CPLEX. In addition, we also propose two new low-complexity heuristics, the first not metaheuristics approaches proposed for the ISHOP-U, which use a linear representation instead of the original matrix candidate solution, achieving better results than the previous Evolutionary Algorithms for the ISHOP-U from the state-of-the-art.

HTML

PDF

DOI: 10.3897/jucs.150345

Authors: Francisco J. Quesada-Real, Francisco Moya-Pérez, Mercedes Rodriguez-Garcia, Bapi Dutta

Abstract: Tendering processes aim to provide transparency in the trade of services or goods but often fall short, leading to corruption and loss of trust. The emergence of Distributed Ledger Technologies (DLTs), such as blockchain, has prompted research into their application for enhancing transparency in tendering. However, adopting DLT usually incurs extra costs, network fees, and high carbon footprints. This paper conducts a Multi-Criteria Decision Making (MCDM) process to select the most suitable DLT for tendering processes. As a result, a novel tendering process based on IOTA is proposed, which improves transparency, ensures ecological sustainability, and avoids extra costs. The IOTA-based approach also fosters collaboration between human and computer capabilities in selecting the tender winner. Our method is compared with existing approaches, demonstrating the highest transparency.

HTML

PDF

DOI: 10.3897/jucs.150763

Authors: Thien Doanh Le, Duc Luan Dang, Thi Quynh Nhu Duong, Kha Tu Huynh

Abstract: Red-light Running increases the risk of collisions and traffic accidents. When a car runs a red light, it can cause a collision with other vehicles moving along the main road, causing serious accidents and even leading to casualties. In Vietnam, many traffic accidents are caused by red-light running. This research paper presents a novel approach for detecting red-light running violations for Vietnamese intersections by leveraging object detection techniques and the YOLO (You Only Look Once) algorithm, a deep neural learning model that uses convolutional neural network architecture (CNNs) for object detection in real-time. The proposed system utilizes CCTV video footage to capture video frames, which are then processed through a trained YOLOv8 model to identify red-light violators. The system’s performance is evaluated based on detection accuracy and processing speed and validated against a custom build dataset extracted from CCTV footages of Vietnamese streets. The experimental results demonstrate high accuracy and processing efficiency up to 93.4% mAP50, 89.2% precision and 92.6% recall, indicating that the proposed approach is suitable for deployment in the context of Vietnamese traffic conditions. The proposed system has significant potential to enhance road safety and mitigate the incidence of red-light running violations in Vietnam.

HTML

PDF

DOI: 10.3897/jucs.150833

Authors: Manuel J. Barranco, Raciel Yera, Francisco J. Martínez

Abstract: In recent years, food recommendation systems have garnered significant attention from internet users seeking diets that are both appealing and health-promoting. For individuals managing chronic conditions such as diabetes, personalized food recommendations that consider both individual preferences and nutritional requirements could potentially yield substantial benefits in maintaining an appropriate dietary regimen. Even though previous research works have been covered the problem of food recommendation for the diabetes domain, they suffer from an insufficient use of the corresponding domain knowledge, and from a deficient management of user preferences in this process. This study then presents a novel preference-based food recommendation framework specifically adapted for patients with diabetes, and that mitigates such previous gaps. Experimental findings suggest that within this context, a balance can be achieved between appealing and health promotion, resulting in nutritionally appropriate menus that simultaneously align with users’ preferences.

HTML

PDF

DOI: 10.3897/jucs.150597

Authors: Juan Carlos Leyva López, Omar Alejandro Reyna Gutiérrez

Abstract: Customer-oriented marketing is a strategy that aims to recommend the right product(s) to the right customer(s). It is low-cost, low-risk, and profit-driven. It typically involves two components: customers and products. One of the critical challenges of targeted marketing is identifying products with potential market value for customers. In this paper, we studied an instance of this general problem. This paper finds the k-most preferable products (k-MPP) from a set of products under study to be offered to a customer for targeted marketing. We model the k-MPP problem as a multicriteria ranking problem and propose an algorithmic framework for customer-oriented marketing. Our framework utilizes a multicriteria outranking approach to solve the k-MPP problem. The framework's effectiveness is demonstrated by conducting a case study to find the k-most preferable restaurants for a customer in a Mexican city.

HTML

PDF

DOI: 10.3897/jucs.150294

Authors: Rosa María Rodríguez Domínguez, Luis Martínez López, Kiril Tenekedjiev

Abstract: The recent advancements in Artificial Intelligence (AI) have profoundly transformed various aspects of our lives, from societal interactions to business operations and educational methodologies. As traditional AI systems grapple with challenges like transparency and the lack of human-centric adaptability, the concept of Hybrid-Augmented Intelligence emerges as a transformative approach. By integrating human cognitive capabilities with advanced computational systems, Hybrid-Augmented Intelligence aims to overcome such limitations, fostering collaboration and decision-making processes that leverage both human insight and machine precision. This special issue focuses on Hybrid-Augmented Intelligence methodologies and applications, exploring their potential to address modern challenges in Industry 4.0, healthcare, cybersecurity, robotics, and beyond.The goal of this special issue is to consolidate pioneering research and practical implementations in the realm of Hybrid-Augmented Intelligent Systems. With this collection of articles, we aim to provide a platform for advancing the understanding of Hybrid-Augmented Intelligence while inspiring future research. This issue offers theoretical insights, methodological advancements, and application-driven case studies, demonstrating how Hybrid-Augmented Intelligence can drive innovation and efficiency across diverse domains.This special issue encompasses five cutting-edge research articles in different related topics of the scope. Each paper is addressing a unique aspect of Hybrid-Augmented Intelligence. These contributions reflect the diversity and richness of current studies, focusing on topics ranging from decision-making frameworks and optimization techniques to applied case studies in real-world contexts. The articles included in this issue are described as follows.

HTML

PDF

DOI: 10.3897/jucs.125714

Authors: Ana Paula Allian, Frank Schnicke, Pablo Oliveira Antonino, Thomas Kuhn, Elisa Yumi Nakagawa

Abstract: The rapid evolution of Industry 4.0 has brought forth transformative changes in manufacturing, accentuating the need for seamless interoperability among heterogeneous systems. However, the geographically distributed and decentralized nature of Industry 4.0 ecosystems presents a pressing challenge: ensuring trustworthy interoperability within a complex web of entities and intermediaries. This paper delves into the pivotal role of blockchain technology in addressing this challenge, aiming to bridge the gap between theoretical promises and practical applications. By examining the feasibility and efficacy of blockchain solutions in fostering trust and enabling interoperability within Industry 4.0 environments, we confront the pressing issue of data security, integrity, and reliability. Through the lens of seven blockchain-based solutions, we navigate the intricate landscape of Industry 4.0, offering insights into the trade-offs, risks, and potentials associated with blockchain adoption. Real-world case studies and practical demonstrations underscore the urgency and relevance of our research, shedding light on pathways for industry stakeholders to navigate the complexities of interoperability. Our findings not only contribute to advancing the discourse on blockchain’s role in Industry 4.0 but also provide actionable strategies for addressing the overarching challenge of ensuring trustworthy interoperability in the digital age.

HTML

PDF

DOI: 10.3897/jucs.125224

Authors: Aneela Nargis, Muhammad Mobeen Movania, Shama Siddiqui

Abstract: Head and neck cancer presents a significant global health challenge, necessitating the development of robust computational models for accurate gene expression analysis. This study introduces Autoencoder-Integrated WideResNet with Dynamic Optimization (AIW-DynOpt), a novel hybrid deep learning framework specifically designed for analyzing head and neck cancer gene expression data. AIW-DynOpt integrates a Deep Undercomplete Autoencoder (DUAE) with a Wide Residual Network (WideResNet) architecture and employs the Successive Halving algorithm for optimal model selection. Utilizing the Cancer Genome Atlas (TCGA) HNSC dataset, which comprises 20,503 genes and 564 samples, our approach focuses on enhancing predictive performance and computational efficiency. A comprehensive evaluation of AIW-DynOpt was conducted, benchmarking it against alternative methods such as DUAE paired with Convolutional Neural Networks (CNN), Support Vector Machines (SVM), and Naive Bayes classifiers. The results demonstrate that AIW-DynOpt consistently outperforms the alternative methods across multiple performance metrics, including accuracy, recall, sensitivity, specificity, and Area under the Curve (AUC). Additionally, AIW-DynOpt exhibits superior computational efficiency, significantly reducing model training time while maintaining high predictive accuracy. This study underscores the potential of hybrid deep learning frameworks in advancing computational models for cancer research, positioning AIW-DynOpt as a promising tool for precise gene expression analysis in head and neck cancer and beyond.

HTML

PDF

DOI: 10.3897/jucs.123539

Authors: Kiran K A, Jaison Jacob

Abstract: Network-on-chip (NoC) architectures have emerged as a potential solution for facilitating communication between processing elements (PEs) in modern multi-core systems. The design and optimization of NoC architectures are critical for achieving efficient communication, reduced energy consumption, and improved overall system performance. In this study, we investigate and compare the performance of two prominent optimization algorithms, like Genetic Algorithm (GA) and CastNet Algorithm, for 2D and 3D mesh NoC architectures. The study’s objective is to estimate these algorithms’ effectiveness in optimizing communication cost, communication energy, and CPU run time in both 2D and 3D mesh NoC architectures. Performance metrics such as communication cost, communication energy consumption, and CPU run time are measured and compared between the two algorithms and carried out on real and custom benchmark applications like MWD, VOPD and MPEG4.

HTML

PDF

DOI: 10.3897/jucs.118220

Authors: Mehmet Ali Altuncu, Kaplan Kaplan, Melih Kuncan

Abstract: Skin cancer confirmation is critical in determining a patient’s treatment planning process after diagnosis. A proper confirmation process enables the determination of the type, stage, and other characteristics of skin cancer, helping to plan the appropriate treatment. These methods prevent the progression of the disease, thereby contributing to a better response to treatment and improving the patient's quality of life. Dermoscopic images are commonly used to confirm skin cancer types. To obtain meaningful results from these images, researchers often apply artificial intelligence techniques in such studies. Specifically, transfer learning models have been commonly used to enhance the features of these images due to the limited availability of medical image data and the difficulty in extracting meaningful information from such data. While most studies focus on classifying skin cancer types, this research aims to classify skin cancer confirmation types using dermoscopic dataset images. Dermoscopic HAM10000 dataset images were used for this purpose. The dataset includes four different confirmation methods: confocal, consensus, follow-up, and histopathology. Four distinct transfer learning models—Resnet-50, Resnet-101, VGG19, and InceptionResnetV2—were utilized. Additionally, ensemble learning was conducted based on the results of these models using the maximum voting approach. The highest success rate was achieved with Resnet-101 at 96.04%. Considering the comparative results, the accuracy of our promising model proved to be significantly high.

HTML

PDF

DOI: 10.3897/jucs.150728

Authors: Christian Gütl

Abstract: Dear Readers,Today we have sad news to share with you. We are deeply saddened by the passing of Arto Salomaa, one of the visionary founding members of our journal. Arto Salomaa, former professor of mathematics at the University of Turku, Finland, was a highly respected pioneer in the field of mathematical theory of computer science with a focus on formal languages and automata theory. Salomaa passed away peacefully on January 26 at the age of 90, surrounded by his family. Our thoughts are with his family, friends and all those who had the privilege of knowing him. His legacy will forever be a part of our journal.It gives me great pleasure to announce the second regular issue of 2025. In this issue, 4 papers by 13 authors from 5 countries - Brazil, Germany, India, Pakistan and Turkiye - cover a great variety of topical aspects of computer science.In a continuous effort to further strengthen our journal, I would like to expand the editorial board: If you are a tenured associate professor or above with a strong publication record, you are welcome to apply to join our editorial board. We are also interested in high-quality proposals for special issues on new topics and trends. As always, I would like to thank all the authors for their sound research and the editorial board and guest reviewers for their extremely valuable review effort and suggestions for improvement. I also want to thank the readers for their interest in our articles, which is reflected in the increasing number of accesses and PDF downloads.In the second regular issue, I am very pleased to introduce the following 4 accepted articles: Mehmet Ali Altuncu, Kaplan Kaplan, and Melih Kuncan from Turkiye discuss their comparative study of transfer learning models - Resnet-50, Resnet-101, VGG19, and InceptionResnetV2 - on skin cancer confirmation methods based on dermoscopic dataset images. Kiran K A and Jaison Jacob from India present their research on Energy-aware application mapping on 3D mesh-based network-on-chip using heuristic mapping algorithms where performance metrics such as communication cost, communication energy consumption, and CPU runtime were applied. In their study, Aneela Nargis, Muhammad Mobeen Movania, and Shama Siddiqui from Pakistan discuss an autoencoder-integrated WideResNet with dynamic optimization, which was designed specifically for the analysis of head and neck cancer gene expression data. In a joint research paper by researchers from Brazil and Germany, Ana Paula Allian, Frank Schnicke, Pablo Oliveira Antonino, Thomas Kuhn and Elisa Yumi Nakagawa look at the adoption of blockchain to trustworthy interoperability in Industry 4.0 systems and aim to highlight the challenges and close the gap between theoretical promises and practical applications.Enjoy Reading!Best wishes,Christian Gütl, Managing Editor-in-Chief

HTML

PDF

DOI: 10.3897/jucs.120860

Authors: Francisco Dominguez-Mateos, Vincent O’Brien, James Garland, Ryan Furlong, Daniel Palacios-Alonso

Abstract: In deep metric learning (DML) high-level input data are represented in a lower-level representation (embedding) space, such that samples from the same class are mapped close together, while samples from disparate classes are mapped further apart. In this lower-level representation, only a single inference sample from each known class is required to accurately discriminate between classes. To this end, embeddings trained for a specific task may contain additional feature information which can be used to go a level deeper into the discrimination task, i.e. allowing for feature sub-discrimination. This study takes an embedding trained to discriminate faces (identities) and uses the inherent feature information within the embedding to differentiate several attributes such as gender, age, and skin tone, without any additional training. This study is split into two cases; intra class discrimination where all the embeddings considered are from the same identity/in-dividual but with minor attributes such as beard/beardless, glasses/without glasses and emotions; and extra class discrimination where the embeddings represent different identities/people with more prominent attributes such as male/female, pale/dark tone, young/older. In the intra class sub-discriminant scenario, the inference process distinguishes common attributes and several artefacts of different identities, achieving 90.0% and 76.0% accuracy for beards and glasses, respectively. The system can also perform extra class sub-discrimination with a high accuracy rate, notably 99.3%, 99.3% and 94.1% for gender, skin tone, and age, respectively. To sum up, this work investigates the sub-discriminative capabilities of DML models by clustering discriminative features evident within the structure of DML embeddings.

HTML

PDF