Dynamic pattern analysis refers to analyzing the relationship of spatial patterns at different time points. Traditional spatial pattern analysis such as data clustering can find the spatial patterns extant at a geographical location at a particular time point but failing to identify spatial dynamics, or changes that occur over time in a particular place. In this paper, we present a dynamic pattern analysis framework, the DPA framework. This framework allows user to identify three types of dynamic patterns in spatial-temporal data: 1) similar spatial patterns at different time points, 2) interactive relationship between two geographical locations as a result of a specific reason and 3) frequent association rules related to particular types of events, geographical locations, and time points. To evaluate the proposed framework, we used it to analyze a set of reported crime data for a district of Hong Kong and compared the identified patterns with some expectations of field experts and prior empirical studies for this kind of data and patterns. In line with expert predictions, we found strong correlations between school holidays and crime clusters. On the contrary, in our data set, we could not find obvious seasonal dependency. These findings are corroborated by related empirical crime studies.