TITLE: Application of machine learning for prediction of road accidents based on indicators: A random forest approach
AUTHORS: Miloš Pljakić, Osman Lindov, Aleksandra Petrović, Predrag Stanojević, Nebojša Arsić
ABSTRACT: In this study, efforts were made to examine the extent of the relationship between specific indicators and traffic accidents resulting in fatalities and injuries at the municipal level in the Republic of Serbia through the application of machine learning algorithms. The indicators analyzed relate to the use of seat belts and mobile phones by drivers and passengers in cars, trucks, and buses. Data were observed for urban and rural areas at the municipal level. In this study, the Random Forest model was employed due to its ability to handle high-dimensional data and capture complex relationships between various factors. The results showed that the use of seat belts by drivers on rural roads has significant predictive power for accidents resulting in fatalities, while the use of seat belts by passengers in cars in urban areas has the highest predictive power for accidents resulting in injuries. Based on these findings and those of each individual indicator, efforts can be directed towards the implementation of measures focused on education, technological solutions, and legislative regulations aimed at increasing the use of safety systems in vehicles.
KEYWORDS: Тraffic accidents, Machine learning, Indicators, Random Forest
PAGES: 70-76
DOI: