Abstract

Crime is a perennial social problem to be controlled by the security forces. Over a period of time, several approaches have been proposed to understand and forecast the crime rates and are available in the literature. Once of the most widely used approach is to use a method called Compstat, developed by US police departments. However, due to its assumptions and simplistic model, Compstat will not effectively capture the patterns from the data. In this paper, a novel approach to forecasting and classification of the crime is proposed. The forecasting is based on regression using attributes derived from time series data. The forecasting is done using five different methods including average based methods (SMA, HW, ARIMA), linear models (MLR) and nonlinear models (GAM). In addition, the crime hotspots will be derived from the forecasted values. The proposed approach was tested on real world data set from the Chicago police department for three years. The results of the proposed systems are compared with the standard approach from Compstat. The results indicate the proposed system provides a much better fit of the data compared to the standard approach.

Presented and Published in: Proceedings of the Sixth International Conference on Business Analytics and Intelligence, December 2018, IISc, India.