Comparative Analysis of Machine Learning Algorithms for Pedestrian Traffic Accident Risk Prediction

This This study presents a comparative analysis of two methodologies for predicting the risk of pedestrian-involved traffic accidents. The methodologies examined include one based on proportional risk distribution and another employing the machine learning algorithm Random Forest. The primary objective is to evaluate the applicability and accuracy of the proposed approaches through the processing and analysis of data derived from real-world cases of concluded legal proceedings. Linguistic variables, defined as risk factors, are classified and quantified based on expert evaluations. The results include interpolation models and graphical representations illustrating the risk severity according to the two methodologies. The analysis demonstrates that both methodologies are applicable for risk assessment, with the Random Forest algorithm providing superior accuracy and reliability in processing complex and heterogeneous data. Furthermore, a correlation analysis confirms a statistically significant linear relationship between the results of the two approaches. Visualization of the results through various graphical tools supports an objective comparison of the methodologies and their application in transport safety analysis.