Forecasting the Trends and Patterns of Crime in Bangladesh using Machine Learning Model

Abstract

Last few years in Bangladesh, the crime rate has increased rapidly. Hence it is an essential task to analyze and predict the crime so that the authority can minimize or prevent the crimes easily. In this situation, machine learning can perform a notable role to reveal the crime trends and patterns of Bangladesh. Here, various machine learning regression models i.e. linear regression, polynomial regression, and random forest regression are used to forecast the trends and patterns of crime in Bangladesh. Dataset used in this research is available for the public which is gathered from the Bangladesh police’s website. The dataset comprises record about various crime types i.e. dacoity, robbery, kidnapping, murder, women & child repression, theft, burglary, arms act, explosive, narcotics, and smuggling of Bangladesh. Firstly, training of regression models is done on the training dataset. After completion of the training, forecasting of crime is performed on the test data by the different regression models. Then we compare the forecasting results with the actual results and calculate the model evaluation metrics for the different applied regression models. After comparing the result, it is possible to find out the best-suited regression model for the crime-related data among all the applied regression models. Finally, it is observed that polynomial and random forest regression are better to predict the crime trends and patterns than the linear regression.