Performance Evaluation of a Hyperparameter Tuned Random Forest Algorithm Based on Artificial Bee Colony for Improving Accuracy and Precision of Crime Prediction Model

Crime prediction is a crucial application of machine learning, enabling law enforcement to make proactive decisions. This research presents a novel crime prediction model that leverages a hybrid approach by tuning the hyperparameters of Random Forest (RF) classifier using the Artificial Bee Colony (ABC) optimization algorithm. The model was developed to improve the prediction accuracy and reliability of crime prediction systems by enhancing the performance of traditional machine learning classifiers. To validate the effectiveness of the proposed RF-ABC model, a comparative analysis against Decision Trees (DT), k-Nearest Neighbors (KNN), and the existing untuned Random Forest model was conducted. Experimental results demonstrate that the proposed RF-ABC model significantly outperforms the baseline models across multiple performance metrics. Specifically, the RF-ABC achieved an accuracy of 95%, precision of 90%, recall of 93%, and an F1-score of 90%. In comparison, the existing RF model yielded an accuracy of 81%, precision of 87%, recall of 84%, and an F1-score of 83%, while DT and KNN models recorded notably lower scores. DT obtained a PEI of 0.6900, PAI of 0.669 and RRI of 0.5200, while KNN has a PEI of 0.9647, PAI of 0.8670 and RRI of 0.5267, RF-ABC had the best result. PEI of 0.9800, PAI of 0.9000 and RRI of 0.7200. Crime prediction metrics show that These findings confirm that the integration of ABC with RF not only fine-tunes the hyperparameters efficiently but also enhances the model's predictive capabilities. The proposed hybrid approach shows promising potential for real-world crime analytic and decision support systems in law enforcement.