A Hybrid Model of Crime Prediction

Abstract

According to statistics, crimes are not random in spatial and temporal distribution. The key to predicting policing is predicting in advance when and where a crime may occur, and providing a reference for preventive measures of police officers. In this paper, a hybrid model of LSTM and STARMA is established. Crime data is complicated. It can be decomposed into trend components, seasonal components, and random components. The LSTM model is established for the trend components and the seasonal components . The STARMA model is established about random components. It solves the problem that the STARMA cannot be modeled on nonlinear or non-stationary data. Verification results show that the model is effective in crime prediction.