Analysis of factors influencing road accident numbers in Poland

Read full text Summarise with VIRAT
Authors: Piotr Gorzelańczyk
Cited by: 0
References: 25

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Between 2010 and 2024, the number of road accidents in Poland decreased significantly – by 44.6%, from 38,832 to 21,519 incidents. Despite this improvement, road safety remains a significant social problem. This study analyzed the factors that had the greatest impact on the number of accidents during this period, combining machine learning methods with cause-and-effect analysis. The study covered a 15-year time series comprising 37 variables describing road infrastructure, demographic structure, and mobility. Causal relationships were analyzed using DAG models, which indicated that mobility patterns and infrastructure development play a key role in shaping the number of accidents.The impact of infrastructure and demographic factors was assessed using the XGBoost model, supported by PCA dimensionality reduction and validation taking into account temporal dependencies. The model achieved a very good fit to the data (MAE = 842, RMSE = 1,105, MAPE = 3.9%, R = 0.991). The strongest influences on the number of accidents were: passenger transport (an increase of approximately 3,210 accidents per million trips), the effect of the COVID-19 pandemic (a decrease of approximately 3,880 accidents per year), and the presence of expressways and motorways (a decrease of approximately 1,920 accidents per 1 km per 100 km). The causal analysis confirmed that road accidents are more a consequence of infrastructure and motorization development than a factor driving this development.The results obtained highlight the importance of coordinated development of public transport and expressway networks. They also indicate areas requiring special attention in road safety policy, such as demographic and mobility changes and the share of N2 category heavy goods vehicles. The proposed model can be a useful tool to support long-term infrastructure planning. A limitation of the study is the lack of behavioral data and detailed seasonal information, which may reduce the accuracy of short-term forecasts.

Fields of Study & Subjects

Traffic and Road Safety Road Transport and Highway Systems Wildlife-Road Interactions and Conservation

Publication details

Journal / source
Journal of Konbin
Volume
56
Issue
1
Pages
1–16
Publication date
30 March 2026
Language
EN