Evaluation of the level of responsibility in pedestrian crashes using machine learning algorithms
Alejandro Moreno-Sanfélix, F. Consuelo Gragera-Peña, Miguel A. Jaramillo-Morán
Traffic crashes involving pedestrians tend to result in the most casualties (minor, serious, or fatal).Therefore, accurately determining the level of responsibility in a pedestrian crash is crucial, as liabilitycan lead to civil, administrative, or criminal consequences. Despite its importance, the scientificliterature contains very few studies focused specially on the attribution of responsibility in trafficaccidents, and even fewer focus on pedestrian collisions. This study evaluated different supervisedclassification models using Machine Learning (ML) techniques to classify the levels of responsibilityof both drivers and pedestrians using real crash data. In this evaluation, 14 binary variables wereconsidered based on four subsystems: human, technological, structural, and normative. The goal is tohelp judicial and police authorities make more efficient and objective attributions of responsibility. Thisinvolves analyzing the most influential variables after the classification process. Then, policymakerswill be able to use these assessments to develop new strategies for improving road safety. The datasetconsists of 510 pedestrian crashes extracted from the reports by the Local Police of Badajoz (LPB) inSpain and judicial decisions of the Spanish Judiciary (SJ). Of the models analyzed, Decision Trees (DT),Naïve Bayes (NB), and Support Vector Machine (SVM) models produced the best initial performance.These three models were then compared, and the metrics showed that the DT model is the bestoption. Furthermore, the feature importance analysis of the 14 variables revealed that possessing adriver’s license is the most influential factor in determining responsibility (47.26%). The next mostinfluential factors were the pedestrian’s location (15.35%); driver under the influence of alcohol/drugs(7.24%); and distracted driving, e.g., using a mobile phone (7.04%).