Performance Evaluation of Machine learning Algorithms for Accident Detection and Prediction

Abstract

This paper presents a comprehensive performance evaluation of machine learning and deep learning models for accident detection and prediction to enhance the reliability of modern Accident Detection Systems (ADS). The study investigates widely used ML algorithms such as CART, Random Forest (RF), Support Vector Machine (SVM), and Multilayer Perceptron (MLP), along with optimization strategies to improve detection accuracy and prediction efficiency. Experimental analysis demonstrates that advanced neural architectures outperform traditional classifiers in handling complex, high-dimensional accident-related datasets. In particular, the GA-optimized MLP model achieves superior results with the highest accuracy, precision, recall, and F1-score, proving its robustness and scalability for real-time applications. Additionally, the proposed framework integrates accident detection, prediction, and alert mechanisms to ensure rapid emergency response and improved road safety. The findings confirm that optimized intelligent ML-based systems offer a dependable solution for reducing traffic risks and saving human lives through timely accident monitoring and prevention.