Performance Improvements of Machine Learning-Based Crime Prediction, A Case Study in Bangladesh

Crime observation examines information on criminal episodes to identify trends in their types, locations, and timing. Forecasting probable criminal activity in some places exploits previous data and data analysis to predict future criminal activity. Crime prediction faces several challenges, and accuracy can be elusive, mainly when dealing with diverse locations. Here, we propose a machine learning-based method, namely the extra tree regressor model, for crime prediction, with specific emphasis on different crime categories like dacoity, robbery, and murder, as well as geographical areas, including metropolitan regions and divisions. We decided to focus our analysis on a particular case study, i.e., crime data about Bangladesh, thus exploiting the crime data taken from the Bangladesh police department website. Our results and comparisons with state-of-the-art methods, such as linear regression, randomforestregressor, XGBregressor, and gradientboosting regressor, demonstrate the superiority of the proposed approach. In particular, our method can reach an accuracy rate of 99.95%, thus proving to be very effective in predicting and analyzing Bangladesh's criminal activity and understanding crime patterns and trends.