Efficiency Comparison of Property Crimes in Thailand by using Statistical Analysis and Machine Learning

Abstract

The purpose of this research is to compare the most suitable methods for forecasting the number of criminal cases against property in the central region and Bangkok by using three techniques, regression analysis method, decision tree method, and random forest method. The data was collected by the National Statistical Office which is a 14-year historical data from 2007-2020 with seven relevant factors and 2,058 data. The data for analysis was divided into two sets, the number of criminal cases in the central region excluding Bangkok and the number of criminal cases in Bangkok. Statistical package was used to analyze related factors by using backward elimination method and testing the efficiency with the WEKA program. Algorithm, which was used in the decision tree is the REPTree algorithm including the Root Mean Square Error (RMSE), and the Mean Absolute Error (MAE) for comparing the prediction efficiency. The results showed that there were 3 factors such as inflation rate, average monthly household income (AMHI), and population which affected data in the central region and 2 factors such as inflation rate and unemployment rate which affected data in Bangkok. As for performance testing, the best method is regression analysis with RMSE value equaled to 0.3193 and 1032.4407 cases, and MAE value was 0.2405 and 858.1238 cases, respectively. This research can be used in forecasting the number of criminal cases in order to find procedures to deal with and solve situations that would occur in the future.