Methods of Data Analysis and Accident Rate Prediction on the Example of the City of Minsk
Year: 2025
Type: article
Abstract: The article discusses methods for predicting accidents on the roads in order to improve road safety. The methods were tested and the accident rate forecast was carried out using the example of Minsk, the estimate of the number of dead and injured was carried out using two models and two methods: the ARIMA model, the SARIMA model, the linear regression metric and the “Random Forest” method. Each method and each model is evaluated according to the accuracy and reliability of the forecasts. The analysis showed that linear regression and “Random Forest” methods most accurately predict the number of deaths, while the ARIMA and SARIMA models provide overestimated forecasts for both categories, and further refinement of the models is required to predict the number of injured. The article also discusses the possibility of using exogenous factors to improve the accuracy of the forecast. The results can be useful for developing effective measures to reduce accidents and improve the situation on the roads.
Authors: M. LEBEDEVA, A. LOBASHOV, S. SEMCHENKOV
Institution: Belarusian National Technical University
Language: English
Topic: Traffic and Road Safety
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4408206682