Analysis of Crime Rate Distribution Based on TPML-WMA

Crime distribution forecasting has a positive impact on social stability and has drew much attention in academia. Existing research methods are not applicable for specific research problems or specific data sets very well. So we build the Vector Motion Model and propose a new algorithm named as TPML-WMA (Transition Probability Matrix Learning and Weighted Moving Average algorithm) to predict a future robbery distribution and figure out how it transfers. According to the idea of machine learning algorithm, we let the transition probability matrix to learn by itself, and do the weighted moving processing on the matrices. Using data from 2001 to 2011 from a city in China, we set up the model, evaluate the TPML-WMA algorithm on brigandage prediction and discuss the performance of algorithms under different initial conditions. At the same time, we compare the proposed algorithm with the classical linear regression method based on the least square method. The results illustrate that the prediction performance of TPML-WMA is greatly improved compared with the linear regression method.