DOI: 10.3897/jucs.161202

Authors: Yelda Fırat, Yılmaz Kılıçaslan, Hüseyin Ali Sarıkaya, Murat Kaan Yılmaz

Abstract: Autism Spectrum Disorder is a neurodevelopmental disorder with onset in early childhood and its diagnosis often requires clinical processes based on long, subjective observations. Although early diagnosis and intervention can significantly improve developmental outcomes, existing methods are limited in terms of scalability and objectivity. The aim of this study is to develop a hybrid deep learning model that detects Autism Spectrum Disorder with high accuracy by analyzing motor behaviors from videos of children recorded in their natural home environment. In this study, joint coordinates were extracted using the MediaPipe Pose model and spatial, temporal, frequency and coordination-based features were calculated from these data. The features were processed with a hybrid architecture integrating CNN, BiLSTM and attention mechanism. CNN captured spatial patterns, BiLSTM learned the dynamics over time, and the attention mechanism focused on critical movement segments. The model achieves over 97% accuracy on closed datasets and over 83% on public videos such as YouTube and TikTok. These results show that the method performs robustly under both controlled and real-world conditions. The study provides a scalable, objective and clinically applicable screening tool that overcomes the problems of artificial environments and limited data.

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.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.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.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.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.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.114357

Authors: Samit Bhanja, Banani Ghose, Abhishek Das

Abstract: Recently time series forecasting has become one of the prime application areas of climatology, economics and industries. Many research works are conducted to forecast the time series more accurately. But few of them are concentrated on predicting the time series over an extended future horizon, and there is also a scope to improve their forecasting accuracy. This work proposes a multi-step-ahead foresting method to produce a stable and accurate forecasting result for the extended future horizon. Firstly, a deep learning-based forecasting model is proposed to predict the time series. Secondly, a fuzzy time series-based error correction model is implemented to enhance the prediction performance of the deep learning model. Here to optimize all the fuzzy time series (FTS) parameters in an integrated way, an integrated butterfly optimization (FTS-IBO) algorithm is proposed. In this study, two different types of real-world multivariate time series datasets are used to analyze the forecasting performance of the proposed model. The performance of the proposed FTS-IBO algorithm is compared with the traditional butterfly optimization (FTS-BO) algorithm. The experimental results show that the FTS-IBO technique is superior to the FTS-BO technique. The forecasting performance of the proposed model has also compared the other state-of-the-art models, and the simulation results exhibit that the proposed model produces a more accurate prediction performance for multi-step-ahead time series forecasting problems compared to other models.

HTML

PDF

DOI: 10.3897/jucs.108073

Authors: Yunus Emre Avcı, Adem Tuncer

Abstract: Lane change detection is crucial for intelligent transportation systems, as it affects traffic flow on both macroscopic and microscopic levels. Lane change models are widely used in traffic and transportation studies, making it important to understand the factors that affect drivers’ lane changing behavior. In this context, we proposed a novel model for detecting lane changes by applying wavelet transform to high-resolution data from unmanned aerial vehicles. The model was trained and tested using empirical lane changing data from pNEUMA. Firstly, the azimuth angle was calculated on WGS-84 coordinates of each vehicle found in the specified road segment. Next, a multi-level wavelet transform was applied to the azimuth series using mother wavelets such as Haar, Daubechies, and Symlet for each vehicle. Machine learning method was applied to extracted features to detect lane changing. Additionally, the lane changing style of drivers was classified as sudden or normal using the same model. The results indicate that the proposed data-driven model is able to accurately detect lane changes and the type of lane change.

HTML

PDF

DOI: 10.3897/jucs.98733

Authors: Sait Can Yucebas, Sukran Yalpir, Levent Genc, Melike Dogan

Abstract: The use of machine learning in real estate is quite new. When the working area is large, the factors affecting the price may vary according to the geographical regions and socioeconomic factors. It is thought that the price prediction performance of a model that will reflect these differences will be more successful than a general model. Unsupervised learning methods can be used both to increase performance and to show the variation of different factors affecting the price according to regions. With this aim, a hybrid model of X-Means clustering and CART decision trees was established in this study. This model successfully learned the geographical and physical variables that affect the price. The prediction performance of the model was compared with the direct capitalization method, which is the gold standard in the domain. The hybrid model has a superior performance over direct capitalization in terms of mean square error, root mean square error and adjusted R-Squared metrics. The scores were 72.86, 0.0057 and 0.978, respectively. The effect of clustering was also examined. Clustering increased the prediction performance by 36%.

HTML

PDF

DOI: 10.3897/jucs.106543

Authors: Melih Kuncan, Kaplan Kaplan, Yılmaz Kaya, Mehmet Recep Minaz, H. Metin Ertunç

Abstract: In the contemporary landscape of industrial manufacturing, the concept of computer numerical control (CNC) has emerged due to the optimization of conventional machinery, distinguished by its remarkable precision and expeditious processing capabilities. These inherent advantages have seamlessly paved the way for the pervasive integration of CNC machines across a myriad of industrial manufacturing sectors. The present study embarks upon a comprehensive inquiry, delving into the intricate analysis of a specialized prototype CNC molding machine, encompassing a meticulous assessment of its structural rigidity, robustness, and propensity for vibrational occurrences. Moreover, an insightful exploration is undertaken to discern the intricate interplay between vibrational signals and intricate machining processes, particularly under diverse conditions such as the presence or absence of the cutting tool, and at varying rotational speeds denoted in revolutions per minute (RPM). The trajectory of this research voyage encompasses an extensive array of empirical experiments meticulously conducted on the prototype CNC machine, with synchronous real-time acquisition of vibrational data. This empirical journey starts by generating two distinct datasets, each meticulously designed to encompass an assemblage of seven distinct rotational speeds, spanning the spectrum from 18000 to 30000 RPM, thereby facilitating enhanced diversity within the dataset. In parallel, a secondary dataset is meticulously derived from the CNC machine operating in the absence of the cutting tool, thereby encapsulating an exhaustive range of 20 discrete RPM values. The extraction of pivotal features aimed at discerning between the vibrational signals arising from distinct conditions (i.e., those emanating from situations involving the presence or absence of the cutting tool) and the associated variance in CNC machine speeds is facilitated through an innovative framework grounded in co-occurrence matrices. The culmination of this methodological framework is the identification of discernible co-occurrence matrices, thereby facilitating the subsequent computation of Heralick features. The classification effort was performed systematically using 10-fold cross-validation analysis, covering a number of different machine learning models. The outcomes emanating from this intricate sequence of systematic methodologies underscore remarkable achievements. Specifically, the classification of vibrational signals corresponding to varying CNC machine speeds, contingent upon the presence or absence of the cutting tool, yields commendable accuracy rates of 94.27% and 94.16%, respectively. Notably, an exemplary accuracy rate of 100% is attained when classifying differing conditions (i.e., situations involving the presence or absence of the cutting tool) across specific RPM settings, prominently at 22000 24000 26000 28000 and 30000 RPM.

HTML

PDF

DOI: 10.3897/jucs.91309

Authors: Raimundo Osvaldo Vieira, Helyane Bronoski Borges

Abstract: Hierarchical multi-label classification problems typically deal with datasets with many attributes and labels, which can negatively impact the classifier performance. The application of dimensionality reduction methods can significantly improve the performance of classifiers. Dimensionality reduction can be performed by feature extraction or feature selection, according to the problem domain and datasets characteristics. This work carried out a systematic literature mapping to identify the approaches and techniques of dimensionality reduction that have been used in hierarchical multi-label classification tasks. Searches were performed on 7 important databases for the Computer Science field. From a list of 184 retrieved papers, 12 were selected for analysis, from which it was possible to determine a general overview of studies conducted from 2010 to 2022. It was identified that feature selection was the most frequent reduction method, with filter approach standing out. In addition, it was detected that most of the works used tree hierarchical structure. As its main outcome, this paper presents the state of the art of dimensionality reduction problem for hierarchical multi-label classification, indicating trends and research issues in the field.

HTML

XML

PDF

DOI: 10.3897/jucs.112513

Authors: Arnak Poghosyan, Ashot Harutyunyan, Naira Grigoryan, Clement Pang

Abstract: Diagnosing IT issues is a challenging problem for large-scale distributed cloud environments due to complex and non-deterministic interrelations between the system components. Modern monitoring tools rely on AI-empowered data analytics for detection, root cause analysis, and rapid resolution of performance degradation. However, the successful adoption of AI solutions is anchored on trust. System administrators will not unthinkingly follow the recommendations without sufficient interpretability of solutions. Explainable AI is gaining popularity by enabling improved confidence and trust in intelligent solutions. For many industrial applications, explainable models with moderate accuracy are preferable to highly precise black-box ones. This paper shows the benefits of rule-induction classification methods, particularly RIPPER, for the root cause analysis of performance degradations. RIPPER reveals the causes of problems in a set of rules system administrators can use in remediation processes. Native cloud applications are based on the microservices architecture to consume the benefits of distributed computing. Monitoring such applications can be accomplished via distributed tracing, which inspects the passage of requests through different microservices. We discuss the application of rule-learning approaches to trace traffic passing through a malfunctioning microservice for the explanations of the problem. Experiments performed on datasets from cloud environments proved the applicability of such approaches and unveiled the benefits.

HTML

XML

PDF

DOI: 10.3897/jucs.94067

Authors: Luis P. Prieto, Gerti Pishtari, Yannis Dimitriadis, María Jesús Rodríguez-Triana, Tobias Ley, Paula Odriozola-González

Abstract: Recent advances in machine learning and natural language processing have the potential to transform human activity in many domains. The field of learning analytics has applied these techniques successfully to many areas of education but has not been able to permeate others, such as doctoral education. Indeed, doctoral education remains an under-researched area with widespread problems (high dropout rates, low mental well-being) and lacks technological support beyond very specialized tasks. The inherent uniqueness of the doctoral journey may help explain the lack of generalized solutions (technological or otherwise) to these challenges. We propose a novel approach to apply the aforementioned advances in computation to support doctoral education. Single-case learning analytics defines a process in which doctoral students, researchers, and computational elements collaborate to extract insights about a single (doctoral) learner's experience and learning process. The feasibility and added value of this approach are demonstrated using an authentic dataset collected by nine doctoral students over a period of at least two months. The insights from this exploratory proof-of-concept serve to spark a research agenda for future technological support of doctoral education, which is aligned with recent calls for more human-centred approaches to designing and implementing learning analytics technologies.

HTML

XML

PDF

DOI: 10.3897/jucs.89923

Authors: Stefan Strydom, Andrei Michael Dreyer, Brink van der Merwe

Abstract: International Classification of Disease (ICD) coding plays a significant role in classify-ing morbidity and mortality rates. Currently, ICD codes are assigned to a patient’s medical record by hand by medical practitioners or specialist clinical coders. This practice is prone to errors, and training skilled clinical coders requires time and human resources. Automatic prediction of ICD codes can help alleviate this burden. In this paper, we propose a transformer-based architecture with label-wise attention for predicting ICD codes on a medical dataset. The transformer model is first pre-trained from scratch on a medical dataset. Once this is done, the pre-trained model is used to generate representations of the tokens in the clinical documents, which are fed into the label-wise attention layer. Finally, the outputs from the label-wise attention layer are fed into a feed-forward neural network to predict appropriate ICD codes for the input document. We evaluate our model using hospital discharge summaries and their corresponding ICD-9 codes from the MIMIC-III dataset. Our experimental results show that our transformer model outperforms all previous models in terms of micro-F1 for the full label set from the MIMIC-III dataset. This is also the first successful application of a pre-trained transformer architecture to the auto-coding problem on the full MIMIC-III dataset.

HTML

XML

PDF

DOI: 10.3897/jucs.89254

Authors: Fahimeh Ramazankhani, Mahdi Yazdian-Dehkord, Mehdi Rezaeian

Abstract: Features extracted from facial images are used in various fields such as kinship verification. The kinship verification system determines the kin or non-kin relation between a pair of facial images by analysing their facial features. In this research, different texture and color features have been used along with the metric learning method, to verify the kinship for the four kinship relations of father-son, father-daughter, mother-son and mother-daughter. First, by fusing effective features, NRML metric learning used to generate the discriminative feature vector, then SVM classifier used to verify to kinship relations. To measure the accuracy of the proposed method, KinFaceW-I and KinFaceW-II databases have been used. The results of the evaluations show that the feature fusion and NRML metric learning methods have been able to improve the performance of the kinship verification system. In addition to the proposed approach, the effect of feature extraction from the image blocks or the whole image is investigated and the results are presented. The results indicate that feature extraction in block form, can be effective in improving the final accuracy of kinship verification.

HTML

XML

PDF

DOI: 10.3897/jucs.79905

Authors: Fatemeh Farnaghi-Zadeh, Mohsen Rahmani, Maryam Amiri

Abstract: Feature selection plays an important role as a preprocessing step for pattern recognition and machine learning. The goal of feature selection is to determine an optimal subset of relevant features out of a large number of features. The neighborhood discrimination index (NDI) is one of the newest and the most efficient measures to determine distinguishing ability of a feature subset. NDI is computed based on a neighborhood radius (E). Due to the significant impact of E on NDI, selecting an appropriate value of E for each data set might be challenging and very time-consuming. This paper proposes a new approach based on targEt PointS To computE neIgh- borhood relatioNs (EPSTEIN). At first, all the data points are sorted in the descending order of their density. Then, the highest density data points are selected as many as the number of classes. To determine the neighborhood relations, the circles centered on the target points are drawn and the points inside or on the circles are considered to be neighbors. In the next step, the significance of each feature is computed and a greedy algorithm selects appropriate features. The performance of the proposed approach is compared to both the commonest and newest methods of feature selection. The experimental results show that EPSTEIN could select more efficient subsets of features and improve the prediction accuracy of classifiers in comparison to the other state-of-the-art methods such as Correlation-based Feature Selection (CFS), Fast Correlation-Based Filter (FCBF), Heuris- tic Algorithm Based on Neighborhood Discrimination Index (HANDI), Ranking Based Feature Inclusion for Optimal Feature Subset (KNFI), Ranking Based Feature Elimination (KNFE) and Principal Component Analysis and Information Gain (PCA-IG).

HTML

XML

PDF

DOI: 10.3897/jucs.70484

Authors: Canan Tastimur, Erhan Akin

Abstract: Deep Learning has been widely used in image-based applications such as object classification, object detection, and object recognition in recent years. Classifying highly similar objects is a very difficult problem. It is difficult to classify datasets in this situation where object similarity between classes and differences between classes are high. In this study, Siamese Convolution Neural Network, which is a similarity measurement-based network, has been practiced to classify 6 types of screws, 5 types of nuts, and 7 types of bolts that are very similar to each other. In addition, this neural network formed with the One-Shot Learning technique is trained. Thanks to the OSL technique, there is no need to use large data sets. Also, there is no need to use large amounts of data from each class. Adding a new class to be classified is also made easier by the use of the OSL technique. The performance results of the proposed method are manifested in detail in the article.

HTML

XML

PDF

DOI: 10.3897/jucs.77029

Authors: Ani Vanyan, Hrant Khachatrian

Abstract: Semi-supervised learning is a branch of machine learning focused on improving the performance of models when the labeled data is scarce, but there is access to large number of unlabeled examples. Over the past five years there has been a remarkable progress in designing algorithms which are able to get reasonable image classification accuracy having access to the labels for only 0.1% of the samples. In this survey, we describe most of the recently proposed deep semi-supervised learning algorithms for image classification and identify the main trends of research in the field. Next, we compare several components of the algorithms, discuss the challenges of reproducing the results in this area, and highlight recently proposed applications of the methods originally developed for semi-supervised learning.

HTML

XML

PDF

DOI: 10.3897/jucs.68185

Authors: Murat Firat, Derya Yiltas-Kaplan, Ruya Samli

Abstract: Over the past decades, air transportation has expanded and big data for transportation era has emerged. Accurate travel demand information is an important issue for the transportation systems, especially for airline industry. So, “optimal seat capacity problem between origin and destination pairs” which is related to the load factor must be solved. In this study, a method for determining optimal seat capacity that can supply the highest load factor for the flight operation between any two countries has been introduced. The machine learning methods of Artificial Neural Network (ANN), Linear Regression (LR), Gradient Boosting (GB), and Random Forest (RF) have been applied and a software has been developed to solve the problem. The data set generated from The World Bank Database, which consists of thousands of features for all countries, has been used and a case study has been done for the period of 2014-2019 with Turkish Airlines. To the best of our knowledge, this is the first time that 1983 features have been used to forecast air travel demand in the literature within a model that covers all countries while previous studies cover only a few countries using far fewer features. Another valuable point of this study is the usage of the last regular data about the air transportation before COVID-19 pandemic. In other words, since many airline companies have experienced a decline in the air travel operation in 2020 due to COVID-19 pandemic, this study covers the most recent period (2014-2019) when flight operation performed on a regular basis. As a result, it has been observed that the developed model has forecasted the passenger load factor by an average error rate of 6.741% with GB, 6.763% with RF, 8.161% with ANN, and 9.619 % with LR.

HTML

XML

PDF

DOI: 10.3897/jucs.2020.058

Authors: Daniel Baulé, Christiane Gresse von Wangenheim, Aldo Wangenheim, Jean Carlo Rossa Hauck

Abstract: The manual transformation of a user interface design into code is a costly and time-consuming process. A solution can be the automation of the generation of code based on sketches or GUI design images. Recently, Machine Learning approaches have shown promising results in detecting GUI elements for such automation. Thus, to provide an overview of existing approaches, we performed a systematic mapping study. As a result, we identified and compared 20 approaches, that demonstrate good performance results being considered useful. These results can be used by researchers and practitioners in order to improve the efficiency of the GUI design process as well as continue to evolve and improve approaches for its support.

HTML

XML

PDF

DOI: 10.3897/jucs.2020.037

Authors: Oscar Camacho-Nieto, Cornelio Yáñez-Márquez, Yenny Villuendas-Rey

Abstract: This paper proposes a novel undersampling method, for dealing with imbalanced datasets. The proposal is based on a novel instance importance measure (also introduced in this paper), and is able to balance hybrid and incomplete data. The numerical experiments carried out show the proposed undersampling algorithm outperforms others algorithms of the state of art, in well-known imbalanced datasets.

HTML

XML

PDF

DOI: 10.3897/jucs.2020.023

Authors: Milan Sečujski, Darko Pekar, Siniša Suzić, Anton Smirnov, Tijana Nosek

Abstract: The paper presents a novel architecture and method for training neural networks to produce synthesized speech in a particular voice and speaking style, based on a small quantity of target speaker/style training data. The method is based on neural network embedding, i.e. mapping of discrete variables into continuous vectors in a low-dimensional space, which has been shown to be a very successful universal deep learning technique. In this particular case, different speaker/style combinations are mapped into different points in a low-dimensional space, which enables the network to capture the similarities and differences between speakers and speaking styles more efficiently. The initial model from which speaker/style adaptation was carried out was a multi-speaker/multi-style model based on 8.5 hours of American English speech data which corresponds to 16 different speaker/style combinations. The results of the experiments show that both versions of the obtained system, one using 10 minutes and the other as little as 30 seconds of target data, outperform the state of the art in parametric speaker/style-dependent speech synthesis. This opens a wide range of application of speaker/style dependent speech synthesis based on small quantities of training data, in domains ranging from customer interaction in call centers to robot-assisted medical therapy.

HTML

XML

PDF

DOI: 10.3897/jucs.2020.014

Authors: Damir Krstinić, Maja Braović, Dunja Božić-Štulic

Abstract: Natural landscape image classification is a difficult problem in computer vision. Many classes that can be found in such images are often ambiguous and can easily be confused with each other (e.g. smoke and fog), and not just by a computer algorithm, but by a human as well. Since natural landscape video surveillance became relatively pervasive in recent years, in this paper we focus on the classification of natural landscape images taken mostly from forest fire monitoring towers. Since these images usually suffer from the lack of the usual low and middle level features (e.g. sharp edges and corners), and since their quality is degraded by atmospheric conditions, this makes the already difficult problem of natural landscape classification even more challenging. In this paper we tackle the problem of automatic natural landscape classiffication by proposing and evaluating a classifier based on a pretrained deep convolutional neural network and transfer learning.

HTML

XML

PDF

DOI: 10.3897/jucs.2020.005

Authors: Smail Sellah, Vincent Hilaire

Abstract: Document classification is a large body of search, many approaches were proposed for single label and multi-label classification. We focus on the multi-label classification more precisely those methods that transformation multi-label classification into single label classification. In this paper, we propose a novel problem transformation that leverage label dependency. We used Reuters-21578 corpus that is among the most used for text categorization and classification research. Results show that our approach improves the document classification at least by 8% regarding one-vs-all classification.

HTML

XML

PDF

DOI: 10.3217/jucs-025-06-0667

Authors: Sonia Contreras, Miguel Manzanedo, Álvaro Herrero

Abstract: Workplace accident rates are always a constant source of concern in different scenarios. On the other hand, different economic cycles modify employment indicators, thereby affecting both the health and the wellbeing of workers. In this study, a Hybrid Neural System is proposed to support the analysis of both workplace accidents and different macroeconomic variables. The application of exploratory projection and nonlinear autoregressive models allow us to recognize patterns and subsequently study the interplay between the two fields, providing valuable information, so that directors can take more informed management decisions.

HTML

XML

PDF

DOI: 10.3217/jucs-025-04-0418

Authors: Zeynep Ozger, Bulent Bolat, Banu Diri

Abstract: Microarray technology is widely used to report gene expression data. The inclusion of many features and few samples is one of the characteristic features of this platform. In order to define significant genes for a particular disease, the problem of high-dimensionality microarray data should be overcome. The Artificial Bee Colony (ABC) Algorithm is a successful meta-heuristic algorithm that solves optimization problems effectively. In this paper, we propose a hybrid gene selection method for discriminatively selecting genes. We propose a new probabilistic binary Artificial Bee Colony Algorithm, namely PrBABC, that is hybridized with three different filter methods. The proposed method is applied to nine microarray datasets in order to detect distinctive genes for classifying cancer data. Results are compared with other wellknown meta-heuristic algorithms: Binary Differential Evolution Algorithm (BinDE), Binary Particle Swarm Optimization Algorithm (BinPSO), and Genetic Algorithm (GA), as well as with other methods in the literature. Experimental results show that the probabilistic self-adaptive learning strategy integrated into the employed-bee phase can boost classification accuracy with a minimal number of genes.

HTML

XML

PDF

DOI: 10.3217/jucs-024-11-1604

Authors: Ramón Cabada, María Lucía Barrón Estrada, Raúl Bustillos

Abstract: This paper describes the process of creating an opinion-mining module that uses deep learning techniques to detect the positive or negative polarity of students' opinions regarding the exercises they solve in an intelligent learning environment (ILE) for the Java language, as well as the detection of learning-centered emotions such as engagement, boredom, and frustration. The information serves as the basis for administrators and teachers who use the ILE to analyze the opinions in order to improve the pedagogy of the ILE exercises. To determine the effectiveness of the deep learning model, we carried out experiments with ten different architectures using the Yelp dataset and one of its own named SentiText containing 147,672 and 10,834 balanced sentences, respectively. We obtained encouraging results with a model that combines a Convolutional Neural Network and a Long Short-Term Memory with an accuracy of 84.32% and an error rate of 0.24 for Yelp and 88.26% and an error rate of 0.33% for SentiText.

HTML

XML

PDF

DOI: 10.3217/jucs-024-11-1582

Authors: Shadi Abudalfa, Moataz Ahmed

Abstract: Microblogging services have been significantly increased nowadays and enabled people to share conveniently their sentiments (opinions) with regard to matters of concerns. Such sentiments have shown an impact on many fields such as economics and politics. Different sentiment analysis approaches have been proposed in the literature to predict automatically sentiments shared in micro-blogs (e.g., tweets). A class of such approaches predicts opinion towards specific target (entity); this class is referred to as target-dependent sentiment classification. Another class, called open domain targeted sentiment classification, extracts targets from the micro-blog and predicts sentiment towards them. In this research work, we propose a new semi-supervised learning technique for developing open domain targeted sentiment classification by using fewer amounts of labelled data. To the best of our knowledge, our model represents the first semi-supervised technique that is proposed for open domain targeted sentiment classification. Additionally, we propose a new supervised learning model for improving accuracy of open domain targeted sentiment classification. Moreover, we show for the first time that SVM HMM is able to improve accuracy of open domain targeted sentiment classification. Experimental results show that our proposed technique outperforms other prominent techniques available in the literature.

HTML

XML

PDF

DOI: 10.3217/jucs-024-10-1378

Authors: Francois Meyer, Brink van der Merwe, Dirko Coetsee

Abstract: Word embedding techniques can be used to learn vector representations of concepts from temporal datasets. Previous attempts to do this amounted to appling word embedding techniques to event sequences. We propose a concept embedding model that extends existing word embedding techniques to take time into account by explicitly modelling the time between concept occurrences. The model is implemented and evaluated using medical temporal data. It is found that incorporating time into the learning algorithm can improve the quality of the resulting embeddings, as measured by an existing methodological framework for evaluating medical concept embeddings.

HTML

XML

PDF

DOI: 10.3217/jucs-024-06-0775

Authors: Yenny Villuendas-Rey, Mariana-D. Alanis-Tamez, Carmen-F. Benguría, Cornelio Yáñez-Márquez, Oscar Camacho-Nieto

Abstract: Computational Intelligence techniques in medicine have become an increasing area of research worldwide. Among them, the application and development of new models and algorithms for disease diagnosis and prediction have been an active research topic. The research contribution of the current paper is the proposal of a novel classification model, and its application to the diagnosis of chronic diseases. One of the main characteristics of the new model is that it is designed to deal with imbalanced data. With the purpose of making experimental comparisons to demonstrate the benefits of the proposed model, we tested five classification models, over medical data. The application of the supervised classification algorithms is done over the Knowledge Extraction based on Evolutionary Learning (KEEL) environment, using a distributed optimally balanced stratified 5-fold cross validation scheme. In addition, the experimental results obtained were validated in order to identify significant differences in performance by mean of a non-parametric statistical test (the Friedman test), and a post-hoc test (the Holm test). The hypothesis testing analysis of the experimental results indicates that the proposed model outperforms other supervised classifiers for medical diagnosis.

HTML

XML

PDF

DOI: 10.3217/jucs-024-06-0725

Authors: Xiucai Ye, Tetsuya Sakurai

Abstract: The analysis of microarray gene expression data to obtain useful information is a challenging problem in bioinformatics. Feature selection is an efficient computational technique in processing the analysis of high-dimensional microarray data. Due to the lack of label information in practice, unsupervised feature selection is considered to be more practically important and correspondingly more difficult. In this paper, we propose a novel unsupervised feature selection method, which utilizes local regression and discriminant analysis for structure learning on microarray gene expression data. By imposing row sparsity on the weight matrix through l2,1-norm regularization, the proposed method optimizes for selecting the discriminative genes which are more informative and better capture the interesting natural classes of samples. We develop an effective algorithm to solve the l2,1-norm-based optimization problem in our method and present the convergence analysis. Finally, we evaluate the proposed method on real microarray gene expression datasets. The experimental results demonstrate that the proposed method not only achieves good performance, but also outperforms other state-of-the-art unsupervised feature selection methods.

HTML

XML

PDF

DOI: 10.3217/jucs-024-06-0711

Authors: Quan Zou, Chi-Wei Chen, Hao-Chen Chang, Yen-Wei Chu

Abstract: Gelatinases proteases with the ability to cleave the extracellular matrix (ECM). Two types of gelatinases exist: Gelatinase A, also referred to as matrix metalloproteinase-2 (MMP-2), and gelatinase B, also referred to as matrix metalloproteinase-9 (MMP-9). MMP-2 and MMP-9 degrade ECM, which is highly expressed during tumor metastasis. The poor therapeutic effects of inhibitors can be attributed to the high structural homology shared by members of the matrix metalloproteinase family. The highly similar structures of these proteases preclude the specific binding of inhibitor drugs. Moreover, the regulatory pathways of MMP-2 and MMP-9 remain poorly understood. An accurate model for the prediction of substrates and the cleavage sites of gelatinases should be developed to enable screening and exploring the physiological and pathological mechanisms of these enzymes. Prediction is based on various types of information on binary integration, physical-chemical properties, protein stability, solvent accessibility, and protein secondary structure. In this study, the first level of the prediction model was constructed on the basis of intergroup differences and support vector machine. Predictive probability was then taken as the characteristic of the second level of the prediction model, which was constructed using different machine-learning methods. The Mathews correlation coefficients of the MMP-2 and MMP-9 prediction models were 89.4% and 64.4%, respectively. The physical-chemical properties of the active sites of MMP-2 and MMP-4 were selected for analysis. The completion of this prediction system will aid the discovery of regulatory paths and novel applications of MMP-2 and MMP-9, as well as provide references for drug design.

HTML

XML

PDF

DOI: 10.3217/jucs-024-05-0577

Authors: S. Prabavathy, K. Sundarakantham, S. Mercy Shalinie

Abstract: The rapid growth of Internet of Things(IoT) has reached all the facets of life including critical infrastructures. It has become the foundation for most of the critical infrastructures. The increased connectivity and the heterogeneity in IoT have widened the attack surface of critical infrastructures for attackers to exploit. Certain cyberattacks in critical infrastructures can lead to catastrophe and hence the attack has to be identified as early as possible to stop or reduce its impact by activating suitable responses. Therefore, the critical infrastructures require an intelligent security mechanism which can intelligently interpret the attacks from the IoT traffic and efficiently handle the attack scenario by activating appropriate response at faster rate. In this work, an autonomic security system with intelligent self-protect mechanism has been proposed for critical infrastructures. The autonomic security system can autonomously detect known attacks using Extreme Learning Machine, predict the unknown attacks using Gaussian process regression, and select suitable response for handling the attack using fuzzy logic. This intelligence of self-protect mechanism is incorporated in the distributed fog nodes to handle the attack scenario at faster rate and protect the critical infrastructures with minimal human intervention. The experimental analysis of the proposed autonomic security system proves to be efficient in detecting and defending the cyber-attacks with high accuracy and success rate. The results on network load and response time indicates the effectiveness of fog computing in proposed system.

HTML

XML

PDF

DOI: 10.3217/jucs-023-12-1200

Authors: Teresa Magal-Royo, Jesus Garcia Laborda, Sara Price

Abstract: Computer adaptive language testing offers the possibility to research and practice m-learning using ubiquitous technology. Virtual education in m-learning uses conventional Learning Objects (LO) to enable the possibility of development of several tasks oriented towards language learning including the assessment and verification of skill improvements in second language acquisition. So far, there are few research papers on the impact of using new multimodal digital resources as LO in the design process of foreign language assessment tests through mobile devices because many English certification tests still continue to use traditional testing techniques [face-to-face and pen-and-paper assessments] combined with conventional digital environments oriented to virtual education. Learning languages requires not only new m-learning scenarios for assessment but also multimodal interactive environments to improve the user's experience during proficiency tests or language certification. Multimodal object learning such as augmented environments, learning games, spatial sound, etc. can be integrated into an assessment process to enhance the user´s experience by simulating natural communicative scenarios. The present article defines an innovative new m-learning scenario for listening comprehension assessment in an on-line test by implementing a multimodal audio learning source named binaural sound. Use of this technology will enable demonstration of other possibilities of human interaction to improve the user's experience in language learning through sound perception and its cognition from the user in an especific learning task.

HTML

XML

PDF

DOI: 10.3217/jucs-022-09-1203

Authors: Jernej Tonejc, Sabrina Güttes, Alexandra Kobekova, Jaspreet Kaur

Abstract: In recent years, the volume and the complexity of data in Building Automation System networks have increased exponentially. As a result, a manual analysis of network traffic data has become nearly impossible. Even automated but supervised methods are problematic in practice since the large amount of data makes manual labeling, required to train the algorithms to differentiate between normal traffic and anomalies, impractical. This paper introduces a framework which allows the characterization of BACnet network traffic data by means of unsupervised machine learning techniques. Specifically, we use clustering, random forests, one-class support vector machines and support vector classifier, after a pre-processing step that includes principal components analysis for dimensionality reduction. We compare the effectiveness of the methods in detecting anomalies by performing experiments on BACnet network traffic data from various sources. We describe which of these unsupervised methods work best in specific scenarios since each method has its distinct advantages and disadvantages. In particular, we discuss which method is best suited to detect new types of anomalies (novelty detection), or which method most reliably and efficiently finds new attacks of a type that has been captured in the data previously.

HTML

XML

PDF

DOI: 10.3217/jucs-022-05-0671

Authors: Yolanda Baca-Gomez, Alicia Martinez, Paolo Rosso, Hugo Estrada, Delia Irazu Hernandez Farias

Abstract: The increasing use of social networks and online sites where people can express their opinions has created a growing interest in Opinion Mining. One of the main tasks of Opinion Mining is to determine whether an opinion is positive or negative. Therefore, the role of the feelings expressed on the web has become crucial, mainly due to the concern of businesses and government to automatically identify the semantic orientation of the views of customers or citizens. This is also a concern, in the area of health to identify psychological disorders. This research focuses on the development of a web application called SWePT (Web Service for Polarity detection in Spanish Texts), which implements the Sequential Minimal Optimization (SMO) algorithm, extracting its features from an affective lexicon in Mexican Spanish. For this purpose, a corpus and an affective lexicon in Mexican Spanish were created. The experiments using three (positive, neutral, negative) and five categories (very positive, positive, neutral, negative, and very negative) allow us to demonstrate the effectiveness of the presented method. SWePT has also been implemented in the Emotion-bracelet interface, which shows the opinion of a user graphically.

HTML

XML

PDF

DOI: 10.3217/jucs-021-11-1454

Authors: Ahmed Al-Zahrani, Hassan Mathkour, Hassan Abdalla

Abstract: Feature-based approaches play an important role and are widely applied in extractive summarization. In this paper, we use particle swarm optimization (PSO) to evaluate the effectiveness of different state-of-the-art features used to summarize Arabic text. The PSO is trained on the Essex Arabic summaries corpus data to determine the best particle that represents the most appropriate simple/combination of eight informative/structure features used regularly by Arab summarizers. Based on the elected features and their relevant weights in each PSO iteration, the input text sentences are scored and ranked to extract the top ranking sentences in the form of an output summary. The output summary is then compared with a reference summary using the cosine similarity function as the fitness function. The experimental results illustrate that Arabs summarize texts simply, focusing on the first sentence of each paragraph.

HTML

XML

PDF

DOI: 10.3217/jucs-021-09-1143

Authors: Gerald De Grancy

Abstract: Distribution planning in urban areas faces a lack of available parking space at customer sites. One approach to mitigate the issue is to cluster nearby customers around known parking locations. Deliveries from each parking location to its assigned customers occur by a second mode of transport (for example by foot). These lead to long service times at each of the clusters. However, long service times in conjunction with time windows can lead to inefficient routes as nearby customer clusters with overlapping service times may not be connected. As a consequence, assigning additional service workers to each vehicle is a strategy to reduce service times. The additional workers can do the last mile deliveries in parallel to reduce the service time of a cluster and hence permit more efficient routing. The trade-off between paying additional workers to reduce costs for vehicles and driving creates a new decision problem called the vehicle routing problem with time windows and multiple service workers (VRPTWMS).

HTML

XML

PDF

DOI: 10.3217/jucs-021-03-0369

Authors: Junho Choi, Chang Choi, Ilsun You, Pankoo Kim

Abstract: The majority of existing malware detection techniques detects malicious codes by identifying malicious behavior patterns. However, they have difficulty identifying new or modified malicious behaviors; consequently, new techniques that can effectively and accurately detect new malicious behaviors are crucial. This paper proposes a method that defines the malicious behaviors of malware using conceptual graphs that are able to describe their concepts and the relationships among them and, consequently, infer their malicious behavior patterns. The inferred patterns are then learned by a Support Vector Machine (SVM) classifier that compares and classifies the behaviors as either normal or malicious. The results of experiments conducted verify that the proposed method detects malicious codes more efficiently than conventional methods. In the experimental results, it exhibits a better detection rate than that of malicious code detection methods that rely solely on the signature based approach. This suggests that the proposed method is not only suitable for detection of malicious codes, but is also more efficient than other detection methods as it combines the advantages of more than two malicious code detection methods.

HTML

XML

PDF

DOI: 10.3217/jucs-020-15-2032

Authors: Antonio Cabot, Eva Garcia-Lopez, Luis De-Marcos, Javier Abraham-Curto

Abstract: E-learning has been a revolution in recent years in training field. This, combined with the increased use of mobile devices has caused the emergence of m-learning. Hence new problems have appeared in the training field, such as displaying correctly learning contents in a mobile device that has restricted features or taking into account the learner's context in the learning process, who could be anywhere. For this reason the adaptation concept is used, in order to personalize or adapt the learning contents to each student. This paper presents a case study in a real course using a multi-agent system for adapting the learning contents to the learner's context and to his/her mobile device. The results of the experiment show that the students who used the adaptive system (experimental group) obtained better grades than the students who did not (control group).

HTML

XML

PDF

DOI: 10.3217/jucs-020-09-1174

Authors: Antonina Danylenko, Jonas Lundberg, Welf Löwe

Abstract: Classification is a constitutive part in many different fields of Computer Science. There exist several approaches that capture and manipulate classification information in order to construct a specific classification model. These approaches are oftentightly coupled to certain learning strategies, special data structures for capturing the models, and to how common problems, e.g. fragmentation, replication and model over-fitting, are addressed. In order to unify these different classification approaches, we define a Decision Algebrawhich defines models for classification as higher order decision functions abstracting from their implementations using decision trees (or similar), decision rules, decisiontables, etc. Decision Algebra defines operations for learning, applying, storing, merging, approximating, and manipulating models for classification, along with some generalalgebraic laws regardless of the implementation used. The Decision Algebra abstraction has several advantages. First, several useful DecisionAlgebra operations (e.g., learning and deciding) can be derived based on the implementation of a few core operations (including merging and approximating). Second,applications using classification can be defined regardless of the different approaches.Third, certain properties of Decision Algebra operations can be proved regardless of the actual implementation. For instance, we show that the merger of a series of probablyaccurate decision functions is even more accurate, which can be exploited for efficientand general online learning. As a proof of the Decision Algebra concept, we compare decision trees with decisiongraphs, an efficient implementation of the Decision Algebra core operations, which cap-ture classification models in a non-redundant way. Compared to classical decision tree implementations, decision graphs are 20% faster in learning and classification withoutaccuracy loss and reduce memory consumption by 44%. This is the result of experiments on a number of standard benchmark data sets comparing accuracy, access time, and sizeof decision graphs and trees as constructed by the standard C4.5 algorithm. Finally, in order to test our hypothesis about increased accuracy when merging decisionfunctions, we merged a series of decision graphs constructed over the data sets. The result shows that on each step the accuracy of the merged decision graph increases withthe final accuracy growth of up to 16%.

HTML

XML

PDF

DOI: 10.3217/jucs-020-04-0507

Authors: Ingrid Christensen, Silvia Schiaffino

Abstract: Recommendation is a significant paradigm for information exploring, which focuses on the recovery of items of potential interest to users. Some activities tend to be social rather than individual, which puts forward the need to offer recommendations to groups of users. Group recommender systems present a whole set of new challenges within the field of recommender systems. In this paper, we present a hybrid approach based on group profiling for homogeneous and non-homogenous groups containing a few distant individual profiles among their members. This approach combines three familiar individual recommendation approaches: collaborative filtering, content-based filtering and demographic information. This hybrid approach allows the detection of those implicit similarities in the user rating profile, so as to include members with divergent profiles. We also describe the promising results obtained when evaluating the approach proposed in the movie and music domain.

HTML

XML

PDF

DOI: 10.3217/jucs-019-12-1834

Authors: Balazs Hidasi, Domonkos Tikk

Abstract: The implicit feedback based recommendation problem--when only the user history is available but there are no ratings--is a much harder task than the explicit feedback based recommendation problem, due to the inherent uncertainty of the interpretation of such user feedbacks. Recently, implicit feedback problem is being received more attention, as application oriented research gets more attractive within the field. This paper focuses on a common matrix factorization method for the implicit problem and investigates if recommendation performance can be improved by appropriate initialization of the feature vectors before training. We present a general initialization framework that preserves the similarity between entities (users/items) when creating the initial feature vectors, where similarity is defined using e.g. context or metadata information. We demonstrate how the proposed initialization framework can be coupled with MF algorithms. We experiment with various similarity functions, different context and metadata based similarity concepts. The evaluation is performed on two implicit variants of the MovieLens 10M dataset and four real life implicit databases. We show that the initialization significantly improves the performance of the MF algorithms by most ranking measures.

HTML

XML

PDF

DOI: 10.3217/jucs-019-09-1295

Authors: Davide Anguita, Alessandro Ghio, Luca Oneto, Xavier Parra, Jorge Reyes-Ortiz

Abstract: In this paper we propose a novel energy efficient approach for the recog-nition of human activities using smartphones as wearable sensing devices, targeting assisted living applications such as remote patient activity monitoring for the disabledand the elderly. The method exploits fixed-point arithmetic to propose a modified multiclass Support Vector Machine (SVM) learning algorithm, allowing to better pre-serve the smartphone battery lifetime with respect to the conventional floating-point based formulation while maintaining comparable system accuracy levels. Experimentsshow comparative results between this approach and the traditional SVM in terms of recognition performance and battery consumption, highlighting the advantages of theproposed method.

HTML

XML

PDF

DOI: 10.3217/jucs-019-07-0965

Authors: Juan Garrido, Victor Penichet, María Lozano

Abstract: Current technology allows educational environments to offer teachers and students the functionality and the information required at any time, whatever the place and circumstance. Concretely, these environments mix three remarkable features: ubiquity, context-awareness and collaboration. Accordingly, a system which is developed with these three features can avoid oversights when performing tasks. Additionally, many aspects of learning fundamentals can be improved, such as collaboration and cooperative learning or students' behaviour. In this paper, we present the definition of a system architecture, which is the first step in obtaining our proposed environment, as the adequate support is not found in any other related works. The architecture presents both a software architecture and a hardware architecture. The software architecture shows the layers in which the system distributes functionality and information. The hardware architecture shows the hardware components to be used, such as smartphones, server, communication elements, etc.

HTML

XML

PDF

DOI: 10.3217/jucs-019-04-0539

Authors: Bogdan Trawiński

Abstract: An approach to predict from a data stream of real estate sales transactions based on ensembles of genetic fuzzy systems was presented. The proposed method relies on incremental expanding an ensemble by models built over successive chunks of a data stream. The output of aged component models produced for current data is updated according to a trend function reflecting the changes of premises prices since the moment of individual model generation or the beginning of the data stream. The impact of different trend functions on the accuracy of single and ensemble fuzzy models was investigated in the paper. Intensive experiments were conducted to evaluate the proposed method using real-world data taken from a dynamically changing real estate market. The statistical analysis of experimental output was made employing the nonparametric methodology designed especially for multiple comparisons including Friedman tests followed by Nemenyi's, Holm's, Shaffer's, and Bergmann-Hommel's post-hoc procedures. The results proved the usefulness of ensemble approach incorporating the correction of individual component model output.

HTML

XML

PDF

DOI: 10.3217/jucs-019-04-0502

Authors: Tomasz Kajdanowicz, Przemyslaw Kazienko

Abstract: Multi-label classification is a machine learning task that assumes that a data instance may be assigned with multiple number of class labels at the same time. Modelling of this problem has become an important research topic recently. This paper revokes AdaBoostSeq multi-label classification algorithm and examines it in order to check its robustness properties. It can be stated that AdaBoostSeq is able to result with quite stable Hamming Loss evaluation measure regardless of the size of input and output space.

HTML

XML

PDF

DOI: 10.3217/jucs-019-04-0462

Authors: Piotr Sobolewski, Michal Woźniak

Abstract: The paper presents a concept drift detection method for unsupervised learning which takes into consideration the prior knowledge to select the most appropriate classification model. The prior knowledge carries information about the data distribution patterns that reflect different concepts, which may occur in the data stream. The presented method serves as a temporary solution for a classification system after a virtual concept drift and also provides additional information about the concept data distribution for adapting the classification model. Presented detector uses a developed method called simulated recurrence and detector ensembles based on statistical tests. Evaluation is performed on benchmark datasets.

HTML

XML

PDF

DOI: 10.3217/jucs-019-03-0325

Authors: Pekka Kekolahti, Juuso Karikoski

Abstract: The purpose of this paper is to identify probabilistic relationships of mobile service usage behaviour, and especially to understand the probabilistic relationship between overall service usage diversity and average daily service usage intensity. These are topical themes due to the high number of services available in application stores which may or may not lead to high usage diversity of mobile services. Four analytical methods are used in the study, all are based on Bayesian Networks; 1) Visual analysis of Bayesian Networks to find initially interesting patterns, variables and their relationships, 2) user segmentation analysis, 3) node force analysis and 4) a combination of expert-based service clustering and machine learning for usage diversity vs. intensity analysis. All the analyses were conducted with handset-based data collected from university students and staff. The analysis indicates that services exist, which mediate usage of other services. In other words, usage of these services increases the probability of using also other services. A service called Installer is an example of this kind of a service. In addition, probabilistic relationships can be found within certain service cluster pairs in their usage diversity and intensity values. Based on these relationships, similar mediation type of behaviour can be found for service clusters as for individual services. This is most visible in the relation between System/Utilities and Business/Productivity service clusters. They do not have a direct relationship but usage diversity is a mediator between them. Furthermore, segmentation analysis shows that the user segment called "experimentalists" uses more mediator services than other user segments. Furthermore, "experimentalists" use a much broader set of services daily, than the other segments. This study demonstrates that a Bayesian Network is a straightforward way to express model characteristics on high level. Moreover, Node Force, Direct and Total effect are useful metrics to measure the mediation effects. The clustering implemented as a hybrid of machine learning and expert-based clustering process is also a useful way to calculate relationships between clusters of more than a hundred individual services.

HTML

XML

PDF

DOI: 10.3217/jucs-018-20-2748

Authors: Lutz Prechelt

Abstract: In September 1997, J.UCS published an article titled "Why we Need an Explicit Forum for Negative Results" [Prechelt, 1997]. It argued that when a plausible approach for solving a computer science or software engineering problem had failed to work out, it was silly for the scientific system not to publish the attempt iff a useful insight had been gained along the way nevertheless. Due to the strong bias of essentially all Computer Science publication venues towards "successful" research results, it was thus required to call for such negative results explicitly in order to avoid that those results would either be misleadingly disguised as successes or disappear in some closet. The article declared that J.UCS had thus agreed to create the "Forum for Negative Results (FNR)" as a permanent special section of J.UCS.

HTML

XML

PDF

DOI: 10.3217/jucs-018-07-0901

Authors: Freddy Bolanos, Fredy Rivera, Nader Bagherzadeh

Abstract: Network-on-Chip (NoC) is a growing and promising communication paradigm for Multiprocessor-System-On-Chip (MPSoC) design, because of its scalability and performance features. In designing such systems, mapping and scheduling are becoming critical stages, because of the increase of both size of the network and application's complexity. Some reported solutions solve each issue independently. However, a conjoint approach for solving mapping and scheduling allows to take into account both computation and communication objectives simultaneously. This paper shows a mapping and scheduling solution, which is based on a Population-Based Incremental Learning (PBIL) algorithm. The simulation results suggest that our PBIL approach is able to find optimal mapping and scheduling, in a multi-objective fashion. A 2-D heterogeneous mesh was used as target architecture for implementation, although the PBIL representation is suited to deal with more complex architectures, such as 3-D meshes.

HTML

XML

PDF

DOI: 10.3217/jucs-017-06-0944

Authors: Sang Lee, Dong Kim, Jong Park

Abstract: E-mail spam is no more garbage but risk since it recently includes virus attachments and spyware agents which make the recipients' system ruined, therefore, there is an emerging need for spam detection. Many spam detection techniques based on machine learning techniques have been proposed. As the amount of spam has been increased tremendously using bulk mailing tools, spam detection techniques should counteract with it. To cope with this, parameters optimization and feature selection have been used to reduce processing overheads while guaranteeing high detection rates. However, previous approaches have not taken into account feature variable importance and optimal number of features. Moreover, to the best of our knowledge, there is no approach which uses both parameters optimization and feature selection together for spam detection. In this paper, we propose a spam detection model enabling both parameters optimization and optimal feature selection; we optimize two parameters of detection models using Random Forests (RF) so as to maximize the detection rates. We provide the variable importance of each feature so that it is easy to eliminate the irrelevant features. Furthermore, we decide an optimal number of selected features using two methods; (i) only one parameters optimization during overall feature selection and (ii) parameters optimization in every feature elimination phase. Finally, we evaluate our spam detection model with cost-sensitive measures to avoid misclassification of legitimate messages, since the cost of classifying a legitimate message as a spam far outweighs the cost of classifying a spam as a legitimate message. We perform experiments on Spambase dataset and show the feasibility of our approaches.

HTML

XML

PDF

DOI: 10.3217/jucs-017-01-0107

Authors: Satyadhyan Chickerur, Aswatha M

Abstract: Neural network learning approach for color image restoration has been discussed in this paper and one of the possible solutions for restoring images has been presented. Here neural network weights are considered as regularization parameter values instead of explicitly specifying them. The weights are modified during the training through the supply of training set data. The desired response of the network is in the form of estimated value of the current pixel. This estimated value is used to modify the network weights such that the restored value produced by the network for a pixel is as close as to this desired response. One of the advantages of the proposed approach is that, once the neural network is trained, images can be restored without having prior information about the model of noise/blurring with which the image is corrupted.

HTML

XML

PDF

DOI: 10.3217/jucs-016-21-3245

Authors: João Luis Garcia Rosa, Juan Adan-Coello

Abstract: In Natural Language Processing (NLP) symbolic systems, several linguistic phenomena, for instance, the thematic role relationships between sentence constituents, such as AGENT, PATIENT, and LOCATION, can be accounted for by the employment of a rule-based grammar. Another approach to NLP concerns the use of the connectionist model, which has the benefits of learning, generalization and fault tolerance, among others. A third option merges the two previous approaches into a hybrid one: a symbolic thematic theory is used to supply the connectionist network with initial knowledge. Inspired on neuroscience, it is proposed a symbolic-connectionist hybrid system called BIOθPRED (BIOlogically plausible thematic (θ) symbolic-connectionist PREdictor), designed to reveal the thematic grid assigned to a sentence. Its connectionist architecture comprises, as input, a featural representation of the words (based on the verb/noun WordNet classification and on the classical semantic microfeature representation), and, as output, the thematic grid assigned to the sentence. BIOθPRED is designed to "predict" thematic (semantic) roles assigned to words in a sentence context, employing biologically inspired training algorithm and architecture, and adopting a psycholinguistic view of thematic theory.

HTML

XML

PDF

DOI: 10.3217/jucs-016-12-1500

Authors: Juan García-Vázquez, Marcela Rodríguez, Monica Tentori, Diana Saldaña, Ángel Andrade, Adán Espinoza

Abstract: Ageing is a global phenomenon which has motivated many research and development projects with the aim of providing computing services that support the active and independent living of the elderly. To integrate the ambient intelligence (AmI) vision into the home environment to allow elders to "age in place", it has been identified the necessity of providing high-level software support for creating ambient assisted living (AAL) environments. We propose activity-aware computing to allow smart environments to provide continuous activity awareness and opportunistically offer assistance aimed at supporting the elders’ current activity. This new paradigm calls for novel tools to help developers mirror human activities in the digital domain, and adapt smart environments based on the activities executed by the users. This paper proposes the use of autonomous agents to cope with the design issues for developing activity-aware systems. We specialized the SALSA agent architecture by incorporating customizable activity-aware mechanisms to infer and represent activities. We illustrate the capabilities offered by SALSA autonomous agents through a design of an activity-aware application for helping elders to manage their medication activity.

HTML

XML

PDF

DOI: 10.3217/jucs-016-09-1190

Authors: Matjaž Juršič, Igor Mozetič, Tomaž Erjavec, Nada Lavrač

Abstract: Lemmatisation is the process of finding the normalised forms of words appearing in text. It is a useful preprocessing step for a number of language engineering and text mining tasks, and especially important for languages with rich inflectional morphology. This paper presents a new lemmatisation system, LemmaGen, which was trained to generate accurate and efficient lemmatisers for twelve different languages. Its evaluation on the corresponding lexicons shows that LemmaGen outperforms the lemmatisers generated by two alternative approaches, RDR and CST, both in terms of accuracy and efficiency. To our knowledge, LemmaGen is the most efficient publicly available lemmatiser trained on large lexicons of multiple languages, whose learning engine can be retrained to effectively generate lemmatisers of other languages.

HTML

XML

PDF

DOI: 10.3217/jucs-016-05-0604

Authors: Johannes Aßfalg, Jing Gong, Hans-Peter Kriegel, Alexey Pryakhin, Tiandi Wei, Arthur Zimek

Abstract: When considering the prediction of a structural class for a protein as a classificationproblem, usually a classifier is based on a feature vector x ∊ ℝn, where the features represent certain attributes of the primary sequence or derived properties (e.g., the predicted secondary structure) of a given protein. Since the structure of a protein (i.e., its native conformation) is stable only under specific environmental conditions, it is commonly accepted to assume proteins being evolutionarily adapted to specific subcellular localizations and according to their physicochemical environment. Our statistical evaluation shows a strong correlation between the subcellular localization of proteins and their structural class. The correlation is strong enough to allow fora classification of proteins into their structural class solely based on information regarding the subcellular localization. We conclude that knowledge regarding the subcellular localization ofproteins can be useful as a feature for the structural classification of proteins.

HTML

XML

PDF

DOI: 10.3217/jucs-015-13-2547

Authors: Yi-Chung Chen, Jeen-Shing Wang

Abstract: This paper presents a Hammerstein-Wiener recurrent neural network (HWRNN) with a systematic identification algorithm for identifying unknown dynamic nonlinear systems. The proposed HWRNN resembles the conventional Hammerstein-Wiener model that consists of a linear dynamic subsystem that is sandwiched in between two nonlinear static subsystems. The static nonlinear parts are constituted by feedforward neural networks with nonlinear functions and the dynamic linear part is approximated by a recurrent network with linear activation functions. The novelties of our network include: 1) the structure of the proposed recurrent neural network can be mapped into a state-space equation; and 2) the state-space equation can be used to analyze the characteristics of the identified network. To efficiently identify an unknown system from its input-output measurements, we have developed a systematic identification algorithm that consists of parameter initialization and online learning procedures. Computer simulations and comparisons with some existing models have been conducted to demonstrate the effectiveness of the proposed network and its identification algorithm.

HTML

XML

PDF

DOI: 10.3217/jucs-015-13-2528

Authors: Yifei Chen, Feng Liu, Bram Vanschoenwinkel, Bernard Manderick

Abstract: This paper focuses on the use of support vector machines on a typical context-dependent classification task, splice site prediction. For this type of problems, it has been shown that a context-based approach should be preferred over a transformation approach because the former approach can easily incorporate statistical measures or directly plug sensitivity information into distance functions. In this paper, we designed three types of context-sensitive kernel functions: polynomial-based, radial basis function-based and negative distance-based kernels. From the experimental results it becomes clear that the radial basis function-based kernel with information gain weighting gets the best accuracies and can always outperform their simple non-sensitive counterparts both in accuracy and in model complexity. And with well designed features and carefully chosen context sizes, our system can predict splice sites with fairly high accuracy, which can achieve the F P 95% rate, 3.94 for donor sites and 5.98 for acceptor sites, an approximate state of the art performance for the moment.

HTML

XML

PDF

DOI: 10.3217/jucs-015-12-2409

Authors: Ratchadaporn Amornchewin, Worapoj Kreesuradej

Abstract: In dynamic databases, new transactions are appended as time advances. This paper is concerned with applying an incremental association rule mining to extract interesting information from a dynamic database. An incremental association rule discovery can create an intelligent environment such that new information or knowledge such as changing customer preferences or new seasonal trends can be discovered in a dynamic environment. In this paper, probability-based incremental association rule discovery algorithm is proposed to deal with this problem. The proposed algorithm uses the principle of Bernoulli trials to find expected frequent itemsets. This can reduce a number of times to scan an original database. This paper also proposes a new updating and pruning algorithm that guarantee to find all frequent itemsets of an updated database efficiently. The simulation results show that the proposed algorithm has better performance than that of previous work.

HTML

XML

PDF

DOI: 10.3217/jucs-015-09-1924

Authors: Yong Xiang, Shaohua Zhang, Yuzhu Shen, Meilin Shi

Abstract: Automatic workflow generation is becoming an active research area for dealing with the dynamics of grid infrastructure, because it has a pervasive impact on system usability, flexibility and robustness. Artificial intelligence technology and explicit knowledge have been exploited in some research for workflow construction or composition. With the increasing use of knowledge, its quality has growing impact on system performance. In this report, we present the process pattern as a vehicle for knowledge representation to capture process expertise at the business level. A pattern-based planning approach is proposed for automated workflow generation. Our pattern-oriented approach decreases user-visible complexity and makes systems more scalable and flexible by utilizing explicit knowledge support. Then we propose a hybrid method of pattern knowledge optimization for pattern-based workflow generation planning; experts define the primary model, and subsequent classifier training adjusts and improves the pattern knowledge settings. Experiments with a prototype application demonstrated that this approach can substantially reduce modelling difficulties and effectively improve pattern knowledge quality.

HTML

XML

PDF

DOI: 10.3217/jucs-015-05-1119

Authors: Naixue Xiong, Jing He, Jong Park, Donald Cooley, Yingshu Li

Abstract: Pervasive smart computing environments make people get accustomed to convenient and secure services. The overall goal of this research is to classify vehicles along the I215 freeway in Salt Lake City, USA. This information will be used to predict future roadway needs and the expected life of a roadway. The classification of vehicles will be performed by a synthesis of multiple sets of features. All feature sets have not yet been determined; however, one such set will be the reduced wavelet transform of the image of a vehicle. In order to use such a feature, it is necessary that the image be normalized with respect to size, position, and so on. For example, a car in the right most lane in an image will appear smaller than one in the left most lane, because the right most lane is closest to the camera. Likewise, a vehicle’s size will vary depending on where in a lane its image is captured. In our case, the image capture area for each lane is approximately 100 feet of roadway. A goal of this paper is to normalize the image of a vehicle so that regardless of its lane or position in a lane, the features will be approximately the same. The wavelet transform itself will not be used directly for recognition. Instead, it will be input to a neural network and the output of the neural network will be one element of the feature set used for recognition.

HTML

XML

PDF

DOI: 10.3217/jucs-015-04-0742

Authors: Sing-Wu Liou, Chia-Ming Wang, Yin-Fu Huang

Abstract: For de novo pattern mining in genomic sequences, the main issues are constructingpattern definition model (PDM) and mining sequence patterns (MSP). The representations of PDMs and the discovery of patterns are functionally dependent; the performances thus dependon the adopted PDMs. The popular PDMs provide only descriptive patterns; they lack multifaceted considerations. Many of existing MSP methods are tied up with the exclusively devisedPDMs, and the specialized and sophisticated models make the mined results hard to be reused. In this research, an integrative pattern mining system is proposed, which consists of a computation-oriented PDM (CO-PDM) and general-purpose MSP (GP-MSP) methods. The CO-PDM defines four computational concerns (CCs) as facets of MSP: expression (E), location (L), range (R)and weight (W), which are integrated into a frame-relayed pattern model (FRPM). The GP-MSP develops a frame-relayed search strategy to resolve the ELR-CCs firstly, with the aids of critical-parameter automating (CPA) procedure; and then the W-CC is determined by hybridizing particle swarm optimization (PSO) and artificial neural network (ANN). The proposed FRPM andGP-MSP had been implemented and applied to 22,448 human introns; from the results, all the well-known patterns were recovered and some new ones were also discovered. Furthermore, theeffectiveness of identified patterns were verified by a two-layered k-nearest neighbor (k-NN) classifier; the average precision and recall are 0.88 and 0.92, respectively. By the case study, theintegrative PDM-MSP system is believed to be effective and reliable; it is optimistic the proposed CO-PDM and GP-MSP are both widely applicable and reusable for mining sequence patterns inthe eukaryotic protein-coding genes.

HTML

XML

PDF

DOI: 10.3217/jucs-015-04-0705

Authors: Jeen-Shing Wang, Jen-Chieh Chiang

Abstract: This paper presents an efficient data preprocessing procedure for the of support vector clustering (SVC) to reduce the size of a training dataset. Solving the optimization problem and labeling the data points with cluster labels are time-consuming in the SVC training procedure. This makes using SVC to process large datasets inefficient. We proposed a data preprocessing procedure to solve the problem. The procedure contains a shared nearest neighbor (SNN) algorithm, and utilizes the concept of unit vectors for eliminating insignificant data points from the dataset. Computer simulations have been conducted on artificial and benchmark datasets to demonstrate the effectiveness of the proposed method.

HTML

XML

PDF

DOI: 10.3217/jucs-014-17-2777

Authors: César Vialardi, Javier Bravo, Alvaro Ortigosa

Abstract: Authoring in adaptive educational hypermedia environment is complex activity. In order to promote a wider application of this technology, the teachers and course designers need specific methods and tools for supporting their work. In that sense, data mining is a promising technology. In fact, data mining techniques have already been used in E-learning systems, but most of the times their application is oriented to provide better support to students; little work has been done for assisting adaptive hypermedia authors through data mining. In this paper we present a proposal for using data mining for improving an adaptive hypermedia system. A tool implementing the proposed approach is also presented, along with examples of how data mining technology can assist teachers.

HTML

XML

PDF

DOI: 10.3217/jucs-014-15-2582

Authors: Yudel Gómez, Rafael Bello, Amilkar Puris, María García, Ann Nowe

Abstract: In this paper we propose a new approach to Swarm Intelligence called Two-Step Swarm Intelligence. The basic idea is to split the heuristic search performed by agents into two stages. In the first step the agents build partial solutions which, are used as initial states in the second step. We have studied the performance of this new approach for the Feature Selection Problem by using Ant Colony Optimization and Particle Swarm Optimization. The feature selection is based on the reduct concept of the Rough Set Theory. Experimental results obtained show that Two-step approach improves the performance of ACO and PSO metaheuristics when calculating reducts in terms of computation time cost and the quality of reducts.

HTML

XML

PDF

DOI: 10.3217/jucs-014-15-2456

Authors: Marcos Cintra, Heloisa Camargo, Estevam Hruschka, Maria do Carmo Nicoletti

Abstract: The definition of the Fuzzy Rule Base is one of the most important and difficult tasks when designing Fuzzy Systems. This paper discusses the results of two different hybrid methods, previously investigated, for the automatic generation of fuzzy rules from numerical data. One of the methods, named DoC-based, proposes the creation of Fuzzy Rule Bases using genetic algorithms in association with a heuristic for preselecting candidate rules based on the degree of coverage. The other, named BayesFuzzy, induces a Bayesian Classifier using a dataset previously granulated by fuzzy partitions and then translates it into a Fuzzy Rule Base. A comparative analysis between both approaches focusing on their main characteristics, strengths/weaknesses and easiness of use is carried out. The reliability of both methods is also compared by analyzing their results in a few knowledge domains.

HTML

XML

PDF

DOI: 10.3217/jucs-014-14-2288

Authors: Longbing Cao, Yuming Ou

Abstract: Market Surveillance plays important mechanism roles in constructing market models. From data analysis perspective, we view it valuable for smart trading in designing legal and profitable trading strategies and smart regulation in maintaining market integrity, transparency and fairness. The existing trading pattern analysis only focuses on interday data which discloses explicit and high-level market dynamics. In the mean time, the existing market surveillance systems available from large exchanges are facing crucial challenges of diversified, dynamic, distributed and cyber-based misuse, mis-disclosure and misdealing of information, announcement and orders in one market or crossing multiple markets. Therefore, there is a crucial need to develop innovative and workable methods for smart trading and surveillance. To deal with such issues, we propose the innovative concept microstructure pattern analysis and corresponding approaches in this paper. Microstructure pattern analysis studies trading behaviour patterns of traders in market microstructure data by utilizing market microstructure knowledge. The identified market microstructure patterns are then used for powering market trading and surveillance agents for automatically detecting/designing profitable and legal trading strategies or monitoring abnormal market dynamics and trader’s behaviour. Such trading/surveillance agent-driven market trading/surveillance systems can greatly enhance the analytical, discovery and decision-support capability of market trading/surveillance than the current predefined rule/alert-based systems.

HTML

XML

PDF

DOI: 10.3217/jucs-014-14-2263

Authors: Damian Dudek

Abstract: Adaptive behavior and learning are required of software agents in many application domains. At the same time agents are often supposed to be resource-bounded systems, which do not consume much CPU time, memory or disk space. In attempt to satisfy both requirements, we propose a novel framework, called APS (standing for Analysis of Past States), which provides agent with learning capabilities with respect to saving system resources. The new solution is based on incremental association rule mining and maintenance. The APS process runs periodically in a cycle, in which phases of agent's normal performance intertwine with learning phases. During the former ones an agent stores observations in a history. After a learning phase has been triggered, the history facts are analyzed to yield new association rules, which are added to the knowledge base by the maintenance algorithm. Then the old observations are removed from the history, so that in the next learning runs only recent facts are processed in search of new association rules. Keeping the history small can save both processing time and disk space as compared to batch learning approaches.

HTML

XML

PDF

DOI: 10.3217/jucs-014-11-1877

Authors: Lorenzo Blanco, Valter Crescenzi, Paolo Merialdo

Abstract: In data-intensive web sites pages are generated by scripts that embed data from a backend database into HTML templates. There is usually a relationship between the semantics of the data in a page and its corresponding template. For example, in a web site about sports events, it is likely that pages with data about athletes are associated with a template that differs from the template used to generate pages about coaches or referees. This article presents a method to classify web pages according to the associated template. Given a web page, the goal of our method is to accurately find the pages that are about the same topic. Our method leverages on a simple, yet effective model to abstract some structural features of a web page. We present the results of an extensive experimental analysis that show the performance of our methods in terms of both recall and precision regarding a large number of real-world web pages.

HTML

XML

PDF

DOI: 10.3217/jucs-014-10-1562

Authors: Janez Brank, Dunja Mladenić, Marko Grobelnik, Nataša Milić-Frayling

Abstract: Feature selection methods are often applied in the context of document classification. They are particularly important for processing large data sets that may contain millions of documents and are typically represented by a large number, possibly tens of thousands of features. Processing large data sets thus raises the issue of computational resources and we often have to find the right trade-off between the size of the feature set and the number of training data that we can taken into account. Furthermore, depending on the selected classification technique, different feature selection methods require different optimization approaches, raising the issue of compatibility between the two. We demonstrate an effective classifier training and feature selection method that is suitable for large data collections. We explore feature selection based on the weights obtained from linear classifiers themselves, trained on a subset of training documents. While most feature weighting schemes score individual features independently from each other, the weights of linear classifiers incorporate the relative importance of a feature for classification as observed for a given subset of documents thus taking the feature dependence into account. We investigate how these feature selection methods combine with various learning algorithms. Our experiments include a comparative analysis of three learning algorithms: Naïve Bayes, Perceptron, and Support Vector Machines (SVM) in combination with three feature weighting methods: Odds ratio, Information Gain, and weights from the linear SVM and Perceptron. We show that by regulating the size of the feature space (and thus the sparsity of the resulting vector representation of the documents) using an effective feature scoring, like linear SVM, we need only a half or even a quarter of the computer memory to train a classifier of almost the same quality as the one obtained from the complete data set. Feature selection using weights from the linear SVMs yields a better classification performance than other feature weighting methods when combined with the three learning algorithms. The results support the conjecture that it is the sophistication of the feature weighting method rather than its compatibility with the learning algorithm that improves the classification performance.

HTML

XML

PDF

DOI: 10.3217/jucs-014-07-1136

Authors: Koichi Moriyama, Mitsuhiro Matsumoto, Ken-ichi Fukui, Satoshi Kurihara, Masayuki Numao

Abstract: In recent years, market forecasting by machine learning methods has been flourishing.Most existing works use a past market data set, because they assume that each trader's individual decisions do not affect market prices at all. Meanwhile, there have been attempts to analyzeeconomic phenomena by constructing virtual market simulators, in which human and artificial traders really make trades. Since prices in a market are, in fact, determined by every trader'sdecisions, a virtual market is more realistic, and the above assumption does not apply. In this work, we design several reinforcement learners on the futures market simulator U-Mart (UnrealMarket as an Artificial Research Testbed) and compare our learners with the previous champions of U-Mart competitions empirically.

HTML

XML

PDF

DOI: 10.3217/jucs-014-05-0745

Authors: Dominique Barth, Loubna Echabbi, Chahinez Hamlaoui

Abstract: We present a distributed learning algorithm for optimizing transit prices in the inter-domain routing framework. We present a combined game theoretical and distributed algorithmic analysis, where the notion of Nash equilibrium with the first approach meets the notion of stability in the second. We show that providers can learn how to strategically set their prices according to a Nash equilibrium; even when assuming incomplete information. We validate our theoretical model by simulations confirming the expected outcome. Moreover, we observe that some unilateral deviations from the proposed rule do not seem to affect the dynamic of the system.

HTML

XML

PDF

DOI: 10.3217/jucs-013-10-1484

Authors: Tao Zeng, Changjie Tang, Yong Xiang, Peng Chen, Yintian Liu

Abstract: Rule mining is an important issue in data mining. To address it, a novel Immune Gene Expression Programming (IGEP) model was proposed. Concepts of rule, gene, immune cell, and antibody were formalized. The dynamic evolution models and the corresponding recursive equations of immune cell, self, immune-tolerance were built. The novel key techniques of IGEP were presented. Experiment results showed that the new method has good stability, scalability and flexibility. It can discover traditional association rule, non-traditional rule including connective "OR" or "NOT", and meta-rule of strong rule. Furthermore, it can perform well in constrained pattern mining.

HTML

XML

PDF

DOI: 10.3217/jucs-013-09-1287

Authors: José Troyano, Fernando Enríquez, Fermín Cruz, José Cañete-Valdeón, F. Ortega

Abstract: In this paper we present an experience in the extraction of named entities from Spanish texts using stacking. Named Entity Extraction (NEE) is a subtask of Information Extraction that involves the identification of groups of words that make up the name of an entity, and the classification of these names into a set of predefined categories. Our approach is corpus-based, we use a re-trainable tagger generator to obtain a named entity extractor from a set of tagged examples. The main contribution of our work is that we obtain the systems needed in a stacking scheme without making use of any additional training material or tagger generators. Instead of it, we have generated the variability needed in stacking by applying corpus transformation to the original training corpus. Once we have several versions of the training corpus we generate several extractors and combine them by means of a machine learning algorithm. Experiments show that the combination of corpus transformation and stacking improve the performance of the tagger generator in this kind of natural language processing applications. The best of our experiments achieves an improvement of more than six percentual points respect to the predefined baseline.

HTML

XML

PDF

DOI: 10.3217/jucs-013-09-1246

Authors: Lihong Li, Vadim Bulitko, Russell Greiner

Abstract: Classification-based reinforcement learning (RL) methods have recently been pro-posed as an alternative to the traditional value-function based methods. These methods use a classifier to represent a policy, where the input (features) to the classifier is the state and theoutput (class label) for that state is the desired action. The reinforcement-learning community knows that focusing on more important states can lead to improved performance. In this paper,we investigate the idea of focused learning in the context of classification-based RL. Specifically, we define a useful notation of state importance, which we use to prove rigorous bounds on policyloss. Furthermore, we show that a classification-based RL agent may behave arbitrarily poorly if it treats all states as equally important.

HTML

XML

PDF

DOI: 10.3217/jucs-013-02-0209

Authors: Cesar Sanín, Edward Szczerbicki, Carlos Toro

Abstract: Collecting, distributing and sharing knowledge in a knowledge-explicit way is a significant task for any company. However, collecting decisional knowledge in the form of formal decision events as the fingerprints of a company is an utmost advance. Such decisional fingerprint is called decisional DNA. Set of experience knowledge structure can assist on accomplishing this purpose. In addition, Ontology-based technology applied to set of experience knowledge structure would facilitate distributing and sharing companies' decisional DNA. Such possibility would assist in the development of an e-decisional community, which will support decision-makers on their overwhelming job. The purpose of this paper is to explain the development of .an OWL decisional Ontology built upon set of experience, which would make decisional DNA, that is, explicit knowledge of formal decision events, a useful element in multiple systems and technologies, as well as in the construction of the e-decisional community.

HTML

XML

PDF

DOI: 10.3217/jucs-012-12-1731

Authors: Narayanan Kulathuramaiyer, Wolf-Tilo Balke

Abstract: Everyone realizes how powerful the few big Web search engine companies have become, both in terms of financial resources due to soaring stock quotes and in terms of the still hidden value of the wealth of information available to them. Following the common belief that "information is power" the implications of what the data collection of a de-facto monopolist in the field like Google could be used for should be obvious. However, user studies show that the real implications of what a company like Google can do, is already doing, and might do in a not too distant future, are not explicitly clear to most people. Based on billions of daily queries and an estimated share of about 49% of the total Web queries [Colburn, 2007], allows predicting with astonishing accuracy what is going to happen in a number of areas of economic importance. Hence, based on a broad information base and having the means to shift public awareness such a company could for instance predict and influence the success of products in the market place beyond conventional advertising or play the stock market in an unprecedented way far beyond mere time series analysis. But not only the mining of information is an interesting feature; with additional services such as Google Mail and on-line communities, user behavior can be analyzed on a very personal level. Thus, individual persons can be targeted for scrutiny and manipulation with high accuracy resulting in severe privacy concerns. All this is compounded by two facts: First, Google's initial strategy of ranking documents in a fair and objective way (depending on IR techniques and link structures) has been replaced by deliberatively supporting or ignoring sites as economic or political issues are demanding [Google Policy: Censor, 2007]. Second, Google's acquisition of technologies and communities together with its massive digitization projects such as [Burright, 2006] [Google Books Library, Project, 2006] enable it to combine information on issues and persons in a still more dramatic way. Note that search engines companies are not breaking any laws, but are just acting on the powers they have to increase shareholder value. The reason for this is that there are currently no laws to constrain data mining in any way. We contend that suitable internationally accepted laws are necessary. In their absence, mechanisms are necessary to explicitly ensure web content neutrality (which goes beyond the net neutrality of [Berners-Lee, 2006]) and a balanced distribution of symbolic power [see Couldry, 2003]. In this paper we point to a few of the most sensitive issues and present concrete case studies to support our point. We need to raise awareness to the threat that a Web search engine monopoly poses and as a community start to discuss the implications and possible remedies to the complex problem.

HTML

XML

PDF

DOI: 10.3217/jucs-012-06-0627

Authors: Einoshin Suzuki

Abstract: In this paper, we survey efforts devoted to discovering interesting exceptions from data in data mining. An exception differs from the rest of data and thus is interesting and can be a clue for further discoveries. We classify methods into exception instance discovery, exception rule discovery, and exception structured-rules discovery and give a condensed and comprehensive introduction.

HTML

XML

PDF

DOI: 10.3217/jucs-012-06-0581

Authors: Benjamin Blankertz, Guido Dornhege, Steven Lemm, Matthias Krauledat, Gabriel Curio, Klaus-Robert Müller

Abstract: We outline the Berlin Brain-Computer Interface (BBCI), a system which enables us to translate brain signals from movements or movement intentions into control commands. The main contribution of the BBCI, which is a non-invasive EEG-based BCI system, is the use of advanced machine learning techniques that allow to adapt to the specific brain signatures of each user with literally no training. In BBCI a calibration session of about 20min is necessary to provide a data basis from which the individualized brain signatures are inferred. This is very much in contrast to conventional BCI approaches that rely on operand conditioning and need extensive subject training of the order 50-100 hours. Our machine learning concept thus allows to achieve high quality feedback already after the very first session. This work reviews a broad range of investigations and experiments that have been performed within the BBCI project. In addition to these general paradigmatic BCI results, this work provides a condensed outline of the underlying machine learning and signal processing techniques that make the BBCI succeed. In the first experimental paradgm we analyze the predictability of limb movement long before the actual movement takes place using only the movement intention measured from the pre-movement (readiness) EEG potentials. The experiments include both off-line studies and an online feedback paradigm. The limits with respect to the spatial resolution of the somatotopy are explored by contrasting brain patterns of movements of left vs. right hand rsp. foot. In a second conplementary paradigm voluntary modulations of sensorimotor rhythms caused by motor imagery (left hand vs. right hand vs. foot) are translated into a continuous feedback signal. Here we report results of a recent feedback study with 6 healthy subjects with no or very little experience with BCI control: half of the subjects achieved an information transfer rate above 35 bits per minute (bmp). Furthermore one subject used the BBCI to operate a mental typewriter in free spelling mode. The overall spelling speed was 4.5-8 letters per minute including the time needed for the correction errors.

HTML

XML

PDF

DOI: 10.3217/jucs-012-04-0408

Authors: Ajith Abraham, Crina Grosan

Abstract: This paper presents three variants of Genetic Programming (GP) approaches for intelligent online performance monitoring of electronic circuits and systems. Reliability modeling of electronic circuits can be best performed by the stressor — susceptibility interaction model. A circuit or a system is considered to be failed once the stressor has exceeded the susceptibility limits. For on-line prediction, validated stressor vectors may be obtained by direct measurements or sensors, which after pre-processing and standardization are fed into the GP models. Empirical results are compared with artificial neural networks trained using backpropagation algorithm and classification and regression trees. The performance of the proposed method is evaluated by comparing the experiment results with the actual failure model values. The developed model reveals that GP could play an important role for future fault monitoring systems.

HTML

XML

PDF

DOI: 10.3217/jucs-012-02-0127

Authors: Ahmad Almonayyes

Abstract: Exploratory data analysis over foreign language text presents virtually untapped opportunity. This work incorporates Naïve Bayes classifier with Case-Based Reasoning in order to classify and analyze Arabic texts related to fanaticism. The Arabic vocabularies are converted to equivalent English words using conceptual hierarchy structure. The understanding process operates at two phases. At the first phase, a discrimination network of multiple questions is used to retrieve explanatory knowledge structures each of which gives an interpretation of a text according to a particular aspect of fanaticism. Explanation structures organize past documents of fanatic content. Similar documents are retrieved to generate additional valuable information about the new document. In the second phase, the document classification process based on Naïve Bayes is used to classify documents into their fanatic class. The results show that the classification accuracy is improved by incorporating the explanation patterns with the Naïve Bayes classifier.

HTML

XML

PDF

DOI: 10.3217/jucs-012-01-0015

Authors: Diane Cook

Abstract: To many people, home is a sanctuary. For those people who need special medical care, they may need to be pulled out of their home to meet their medical needs. As the population ages, the percentage of people in this group is increasing and the effects are expensive as well as unsatisfying. We hypothesize that many people with disabilities can lead independent lives in their own homes with the aid of at-home automated assistance and health monitoring. In order to accomplish this, robust methods must be developed to collect relevant data and process it dynamically and adaptively to detect and/or predict threatening long-term trends or immediate crises. The main objective of this paper is to investigate techniques for using agent-based smart home technologies to provide this at-home health monitoring and assistance. To this end, we have developed novel inhabitant modeling and automation algorithms that provide remote health monitoring for caregivers. Specifically, we address the following technological challenges: 1) identifying lifestyle trends, 2) detecting anomalies in current data, and 3) designing a reminder assistance system. Our solution approaches are being tested in simulation and with volunteers at the UTA's MavHome site, an agent-based smart home project.

HTML

XML

PDF

DOI: 10.3217/jucs-011-11-1752

Authors: Martin Atzmueller, Frank Puppe

Abstract: Visual mining methods enable the direct integration of the user to overcome major problems of automatic data mining methods, e.g., the presentation of uninteresting results, lack of acceptance of the discovered findings, or limited confidence in these. We present a novel subgroup mining approach for explorative and descriptive data mining implemented in the VIKAMINE system. We propose several integrated visualization methods to support subgroup mining. Furthermore, we describe three case studies using data from fielded systems in the medical domain.

HTML

XML

PDF

DOI: 10.3217/jucs-011-09-1494

Authors: José Bravo, Ramón Hervás, Gabriel Chavira

Abstract: In recent years, there have been many efforts at research towards obtaining the simple and natural use of computers, with interfaces closer to the user. New visions such as that of the Ubiquitous Computing paradigm emerge. In Ubiquitous Computing the computer is distributed in a series of devices with reduced functionality, spread over the user's environment and communicating wirelessly. With these, context-aware applications are obtained. In this paper we present an approach to the classroom context by identification process using RFID technology, as an implicit input to the system. The main goal is to acquire natural interaction, because the only requirement for the user (teacher or student) is to carry a device (smart label), identifying and obtaining context services. Some of these services and the mechanisms that make them available are described here, together with a scenario of their use in the classroom.

HTML

XML

PDF

DOI: 10.3217/jucs-011-08-1426

Authors: Francisco Ferrer-Troyano, Jesús Aguilar-Ruiz, José Riquelme

Abstract: Mining data streams is a challenging task that requires online systems based on incremental learning approaches. This paper describes a classification system based on decision rules that may store up-to-date border examples to avoid unnecessary revisions when virtual drifts are present in data. Consistent rules classify new test examples by covering and inconsistent rules classify them by distance as the nearest neighbour algorithm. In addition, the system provides an implicit forgetting heuristic so that positive and negative examples are removed from a rule when they are not near one another.

HTML

XML

PDF

DOI: 10.3217/jucs-011-08-1353

Authors: João Gama, Pedro Medas

Abstract: This paper presents a system for induction of forest of functional trees from data streams able to detect concept drift. The Ultra Fast Forest of Trees (UFFT) is an incremental algorithm, which works online, processing each example in constant time, and performing a single scan over the training examples. It uses analytical techniques to choose the splitting criteria, and the information gain to estimate the merit of each possible splitting-test. For multi-class problems the algorithm builds a binary tree for each possible pair of classes, leading to a forest of trees. Decision nodes and leaves contain naive-Bayes classifiers playing different roles during the induction process. Naive-Bayes in leaves are used to classify test examples. Naive-Bayes in inner nodes play two different roles. They can be used as multivariate splitting-tests if chosen by the splitting criteria, and used to detect changes in the class-distribution of the examples that traverse the node. When a change in the class-distribution is detected, all the sub-tree rooted at that node will be pruned. The use of naive­Bayes classifiers at leaves to classify test examples, the use of splitting-tests based on the outcome of naive-Bayes, and the use of naive-Bayes classifiers at decision nodes to detect changes in the distribution of the examples are directly obtained from the sufficient statistics required to compute the splitting criteria, without no additional computations. This aspect is a main advantage in the context of high-speed data streams. This methodology was tested with artificial and real-world data sets. The experimental results show a very good performance in comparison to a batch decision tree learner, and high capacity to detect drift in the distribution of the examples.

HTML

XML

PDF

DOI: 10.3217/jucs-011-04-0495

Authors: Tomaso Forzi, Meikel Peters

Abstract: A well-functioning Knowledge Management is a competitive advantage for enterprises that act in co-operative and distributed networks with knowledge intensive production processes. A Knowledge Management approach that integrates both, hard factors (e.g. Information Technology) and soft factors (e.g., cultural aspects) for distributed and dynamic entrepreneurial (inter-organisational) networks is currently missing. This paper presents research findings of a project that is developing a methodology as well as an appropriate toolkit to support a service provider responsible for the KM within distributed entrepreneurial networks. The project integrates explicitly both new Information and Communication Technology driven organisational concepts, human-oriented approaches and existing KM methodologies and instruments.

HTML

XML

PDF

DOI: 10.3217/jucs-011-04-0452

Authors: Norbert Gronau, Claudia Müller, Roman Korf

Abstract: Existing approaches in the area of knowledge-intensive processes focus on integrated knowledge and process management systems, the support of processes with KM systems, or the analysis of knowledge-intensive activities. For capturing knowledge-intensive business processes well known and established methods do not meet the requirements of a comprehensive and integrated approach of process-oriented knowledge management. These approaches are not able to visualise the decisions, actions and measures which are causing the sequence of the processes in an adequate manner. Parallel to conventional processes knowledge-intensive processes exist. These processes are based on conversions of knowledge within these processes. To fill these gaps in modelling knowledge-intensive business processes the Knowledge Modelling and Description Language (KMDL) got developed. The KMDL is able to represent the development, use, offer and demand of knowledge along business processes. Further it is possible to show the existing knowledge conversions which take place additionally to the normal business processes. The KMDL can be used to formalise knowledge-intensive processes with a focus on certain knowledge-specific characteristics and to identify process improvements in these processes. The KMDL modelling tool K-Modeler is introduced for a computer-aided modelling and analysing. The technical framework and the most important functionalities to support the analysis of the captured processes are introduced in the following contribution.

HTML

XML

PDF

DOI: 10.3217/jucs-010-12-1597

Authors: Rajat Arora, Michael Hsiao

Abstract: We propose a novel technique to improve SAT-based Combinational Equivalence Checking (CEC). The idea is to perform a low-cost preprocessing that will statically induce global signal relationships into the original CNF formula of the miter circuit under verification, and hence reduce the complexity of the SAT instance. This efficient and effective preprocessing quickly builds up the implication graph for the miter circuit under verification, yielding a large set of direct, indirect and extended backward implications. These two-node implications spanning the entire circuit are converted into binary clauses, and they are added to the miter CNF formula. The added clauses constrain the search space of the SAT solver and provide correlation among the different variables, which enhances the Boolean Constraint Propagation (BCP). Experimental results on large and difficult ISCAS'85, ISC AS'89 (full scan) and ITC'99 (full scan) CEC instances show that our approach is independent of the state-of-the-art SAT solver used, and that the added clauses help to achieve not eworthy speedup for each of the cases. Also, comparison with Hyper-Resolution (Hypre), Non-Increasing Variable Elimination Resolution (NIVER) and the propositional formula checker HeerHugo, suggests that our technique is more powerful, yielding non-trivial clauses that significantly simplify the SAT instance complexity.

HTML

XML

PDF

DOI: 10.3217/jucs-010-03-0252

Authors: Koichi Hori, Kumiyo Nakakoji, Yasuhiro Yamamoto, Jonathan Ostwald

Abstract: Our research on knowledge management is rooted in the community perspective. We believe that knowledge systems should serve primarily to help people create and share new knowledge. But we also acknowledge the role of stable, structured and reliable information, both as a component of our systems and as a component of the organizations within which we work. The contribution of the paper is a framework for integrating organizational and community perspectives on knowledge management and its computational support. Our basic idea is that knowledge is not a static chunk of information, but rather, knowledge evolves in a cycle of knowledge liquidization and crystallization. The evolving process takes place through the interactions among conceptual worlds, representational worlds, and the real world. This paper first describes the knowledge liquidization and crystallization framework. We then illustrate the approach with three systems, Knowledge Nebula Crystallizer, livingOM, and ART-SHTA.

HTML

XML

PDF

DOI: 10.3217/jucs-010-03-0205

Authors: Dimitris Apostolou, Grigoris Mentzas, Kostas Baraboutis, Soumi Papadopoulou

Abstract: The new knowledge-based economy necessitates increasingly the collaboration between different organisations. Despite the recent upsurge in knowledge management and decision support systems, the vast majority of these systems focus on individual organisations. This article introduces the concept of the Learning Network - inter-organisational structures, formally established to increase the participants' knowledge and innovative capability - and examines the main functions and roles of a Learning Network. It presents an integrated toolkit for supporting knowledge sharing and decision making in Learning Networks that consists of a software system and a methodology. It also briefly presents how the toolkit has been piloted in an automotive cluster. Finally it provides a constructive set of recommendations for using IT to support learning and knowledge sharing in Learning Networks.

HTML

XML

PDF

DOI: 10.3217/jucs-010-01-0073

Authors: Maja Pivec, Konrad Baumann

Abstract: The paper focuses on adaptability, knowledge mediation and knowledge flows in face-to-face classes compared to computer-based or Internet-based classes. The paper gives an overview of features of on-line learning systems that facilitate the learning process and gives some aspects on adaptation and personalisation issues within those systems. Some recent developments of intelligent tutors capable of expressing emotions are presented. Application examples of adaptable multimedia e-learning solutions for different user groups are described. An outlook on possible future developments and constraints is provided. The paper starts an important discussion about how to design effective human-computer interaction.

HTML

XML

PDF

DOI: 10.3217/jucs-009-12-1372

Authors: Kai Reinhardt, Klaus North

Abstract: The present state of research on competence management does not provide any suitable model that can be used in practice. Neither results from organizational nor from cognitive and social sciences meet the requirements for an application-oriented competence management completely as yet. An integrative competence management must be able to synchronise individual with organisational competencies. This linking is still neglected in research. A convenient solution has not been described yet. This article presents a model for an integrated competence management model, which gives approaches from both cognitive science and organizational science a practical framework of action.

HTML

XML

PDF

DOI: 10.3217/jucs-009-06-0530

Authors: Christian Biemann, Uwe Quasthoff, Karsten Böhm, Christian Wolff

Abstract: Automatic acquisition of information structures like Topic Maps or semantic networks from large document collections is an important issue in knowledge management. An inherent problem with automatic approaches is the treatment of multiword terms as single semantic entities. Taking company names as an example, we present a method for learning multiword terms from large text corpora exploiting their internal structure. Through the iteration of a search step and a verification step the single words typically forming company names are learnt. These name elements are used for recognizing compounds in order to use them for further processing. We give some evaluation of experiments on company name extraction and discuss some applications.

HTML

XML

PDF

DOI: 10.3217/jucs-007-09-0794

Authors: Sandra Fontani

Abstract: We study the problem of efficient identification of particular classes of p-time languages, called uniform. We require the learner to identify each language of such a class by constantly guessing, after a small number of examples, the same index for it. We present three identification paradigms based on different kind of examples: identification on informant (positive and negative information), measure identification (positive information in a probabilistic setting), identification with probability (positive and negative information in a probabilistic setting). In each case we introduce two efficient identification paradigms, called efficient and very efficient identification respectively. We characterize efficient identification on informant and with probability and, as a corollary, we show that the two identification paradigms are equivalent. A necessary condition is shown for very efficient identification on informant, which becomes sufficient if and only if P = NP. The same condition is sufficient for very efficient identiffication with probability if and only if NP=RP. We show that (very) efficient identification on informant and with probability are strictly stronger than (very) efficient measure identiffication.

HTML

XML

PDF

DOI: 10.3217/jucs-006-08-0759

Authors: Sandra Fontani

Abstract: We consider the problem of identifying a class of p­time functions in efficient time. We restrict our attention to particular classes of p-time functions, called uniform and we try to identify each function of such a class by guessing, after a small number of examples, some index for it or its next value. In both cases we introduce two efficient identification paradigms, called efficient and very efficient identification respectively. We find a characterization for efficient identification and, as a corollary, we show that the entire class P is not efficiently identifiable. A necessary condition is shown for very efficient identification, which becomes sufficient if and only if P = NP. We give some examples of well-known uniform classes which are very efficiently identifiable in both identification paradigms.

HTML

XML

PDF

DOI: 10.3217/jucs-003-09-1074

Authors: Lutz Prechelt

Abstract: Current Computer Science (CS) research is primarily focused on solving engineering problems. Often though, promising attempts for solving a particular problem fail for non-avoidable reasons. This is what I call a negative result: something that should have worked does not. Due to the current CS publication climate such negative results today are usually camouflaged as positive results by non-evaluating or mis-evaluating the research or by redefining the problem to fit the solution. Such publication behavior hampers progress in CS by suppressing some valuable insights, producing spurious understanding, and misleading further research efforts. Specific examples given below illustrate and back up these claims. This paper is the announcement of a (partial) remedy: a permanent publication forum explicitly for negative CS research results, called the Forum for Negative Results, FNR. FNR will be a regular part of J.UCS.

HTML

XML

PDF