Crime rate prediction using machine learning

Chee, Man Hang (2022) Crime rate prediction using machine learning. Final Year Project, UTAR.

Preview

PDF Download (3603Kb) | Preview

Abstract

As crime is a plague to society, every country has been actively trying to come up with solutions to reduce crimes. From things like campaigns to raise money for low-income household, crime watches, more frequent patrols, etc. However, even with these measures crime rates still remains at an all-time high. Therefore, with the implementation of this crime rate prediction system, the police can employ predictive policing whereby they can patrol the areas with a higher chance of crimes. With this, they can make a more informed decision on the areas to patrol. To develop this system, I used the San Francisco crime dataset. With this I have employed Feature engineering to aid the system in getting higher accuracies. I have also employed various ensemble learning methods such as XGBoost classifier, Decision tree, and Random Forest Classifier. After which I performed hyperparameter tuning with RandomSearchCV to aid in increasing the accuracies of the prediction of the system. One additional model was also used which was the SARIMAX model which was used to forecast future crime statistics for each Police District.

Actions (login required)