EMPIRICAL ANALYSIS FOR CRIME PREDICTION USING MACHINE LEARNING AND DEEP LEARNING TECHNIQUES

The rising trend in criminal activity has underscored the urgent need for intelligent systems capable of predicting and analyzing crime patterns with high accuracy.This research presents an interactive Streamlit Web Application for crime prediction that integrates a Random Forest Classifier and Convolutional Neural Networks (CNNs).By transforming structured features such as crime type, location, latitude, longitude, weather, occurrence day, and demographic profiles of offenders and victims, the system classifies incidents as "Crime" or "Not Crime."While CNNs are traditionally applied in image and signal processing, this study demonstrates their effectiveness in structured crime data analysis.The application offers a secure, userfriendly interface with authentication, real-time forecasts, and context-aware recommendations.The backend leverages labeled datasets, stored models, and pre-trained encoders to ensure scalability and robust performance.Empirical evaluations show high accuracy rates of 95.2% (CNN) and 91.7% (Random Forest).Designed for researchers, law enforcement, and urban planners, this tool supports proactive crime prevention, resource allocation, and ethical, transparent decision-making, with future enhancements planned for clustering, spatial mapping, and live data integration.