Comparative study on artificial intelligence techniques in crime forecasting

Abstract

An application of efficient crime analysis is beneficial and helpful to understand the behavior of trend and pattern of crimes. Crime forecasting is an area of research that assists authorities in enforcing early crime prevention measures. Statistical technique has been widely applied in the past to develop crime forecasting models. However, it has been observed that researchers have begun to shift their research interests from statistical model to artificial intelligence model in crime forecasting. Thus, this study is conducted to observe the capabilities of artificial intelligence technique in improving crime forecasting. The main objective of this study is to conduct a comparative analysis on forecasting performance capabilities of four artificial intelligence techniques, namely, artificial neural network (ANN), support vector regression (SVR), random forest (RF), and gradient tree boosting (GTB) in forecasting crime rate. Forecasting capability of each technique was assessed in terms of measurement of errors. From the result obtained, GTB showed the highest performance capability where it scored the lowest measurement of errors compared to SVR, RF, and ANN.