AI Based Crime Prediction and Hotspot Alert System

Abstract

This paper presents a machine learning–based approach for predicting criminal activities and identifying areas with high crime rates. The system uses a Random Forest classifier to predict crime types based on spatial and temporal features, while K-Means clustering is applied to geographical coordinates to detect the top 10 crime-prone locations. The proposed system is developed as a web application with a Flask-based backend and a web-based frontend interface. The application provides real-time crime predictions, probability analysis, hotspot visualization, and risk level estimation for both the public and law enforcement agencies. Technologies such as HTML5, CSS3, JavaScript, Leaflet.js, and Chart.js are used to visualize the results through maps and charts. The system was evaluated using the Boston crime dataset, and the results indicate that the model can effectively predict crime types and identify hotspot areas. By combining supervised learning for crime classification and unsupervised learning for hotspot detection, the system provides a complete end-to-end solution for crime prediction and analysis. This system helps authorities and citizens make informed decisions for crime prevention and safety planning.