A Clustering Based Hotspot Identification Approach For Crime Prediction.

With the emergence in the field of crime prediction, researchers found that crime shows geographical patterns. These patterns can be useful to predict crime before it happens and allows police to take proactive measures. Crime prediction finds application in areas like predictive policing, Hotspot evaluation and geographic profiling. Each category of crime holds some relation with time, weather, location, census parameters like annual income, literacy rate of the area. All these serve as indicators for predicting crime. In this work, historic crime events are used as indicators to predict crime. In this paper, a spatiotemporal crime prediction technique based on machine learning coupled with 2-Dimensional Hotspot analysis is proposed. For performing 2-Dimensional Hotspot analysis clustering is used. Performance of the proposed model is compared when it used state of the art classification techniques without hotspot analysis and with hotspot analysis and it is found that model with hotspot analysis achieves better performance. [ABSTRACT FROM AUTHOR]