Crime Forecasting Using Data Mining Techniques

Crime is classically "unpredictable". It is not necessarily random, but neither does it take place consistently in space or time. A better theoretical understanding is needed to facilitate practical crime prevention solutions that correspond to specific places and times. In this study, we discuss the preliminary results of a crime forecasting model developed in collaboration with the police department of a United States city in the Northeast. We first discuss our approach to architecting datasets from original crime records. The datasets contain aggregated counts of crime and crime-related events categorized by the police department. The location and time of these events is embedded in the data. Additional spatial and temporal features are harvested from the raw data set. Second, an ensemble of data mining classification techniques is employed to perform the crime forecasting. We analyze a variety of classification methods to determine which is best for predicting crime "hotspots". We also investigate classification on increase or emergence. Last, we propose the best forecasting approach to achieve the most stable outcomes. The result of our research is a model that takes advantage of implicit and explicit spatial and temporal data to make reliable crime predictions.