Crime Prediction Using Improved Hybrid Models Machine Learning Algorithm

Rising crime rates present a significant challenge to public safety and necessitate the development of reliable crime prediction models. This study investigates the application of machine learning techniques to real-world crime data for predicting future crime trends. Five regression models—Linear Regression, Gradient Boosting Regressor, Random Forest Regressor, Extra-Trees Regressor, and XGBRegressor—are evaluated and compared. To further improve predictive performance, a hybrid model is proposed by integrating the Extra-Trees Regressor and XGBRegressor through a voting-based ensemble strategy. Experimental findings show that the proposed hybrid model performs better than the individual models on the given dataset. The results demonstrate the potential of ensemblebased regression approaches for improving crime prediction and supporting data-driven decision-making in crime prevention.