A Machine Learning-Based Predictive System for Road Accident Risk Management Using Cloud Architecture

This study aimed to address the issue of the high frequency of traffic accidents in Lima Metropolitana by providing a predictive system built using machine learning techniques. A Random Forest (RF) model was trained with a set of historical data on previous accidents, relevant climatic variables, and infrastructure characteristics. The method used involves basic stages of information capture and preprocessing, exhaustive feature engineering to maximize predictive capacity, and implementation in a cloud-based architecture. The results obtained show that the proposed model achieved an accuracy of 50.92%, a recall of 56.97% and an F1 score of 53.77%, demonstrating high efficiency in the detection of geographical areas with a high risk of accident occurrences. Feature importance analysis revealed that variables such as the district (37.2%), the time of the accident (10.8%), and the month (9.5%) are the most significant factors in the predictions, confirming the importance of spatio-temporal patterns in traffic incidents. This proposed system has a scalable and adaptable design that ensures its applicability in urban environments with similar characteristics. Such technologies can significantly contribute to the reduction of accidents and the improvement in traffic management in the capital of Peru.