Analysing factors influencing railway accidents: A predictive approach using multinomial logistic regression and data mining

Abstract

Railway accidents, particularly suicides and suicide attempts, significantly disrupt operations, cause delays in passenger and freight services, and result in varying degrees of infrastructure damage. This study focuses on identifying the relationship between suicide-related railway incidents, as the most frequent type of railway accidents, and socio-economic factors, utilising data from 2015 to 2022 provided by the Railways of the Slovak Republic. Using a data mining approach, a logistic regression model was developed to predict the accident rate based on key socio-economic factors. This model demonstrates high prediction performance, with significant predictors including interest, marriage, and fertility rates. The data mining approach allows for the efficient extraction of relevant patterns and relationships and ensures that the model can be easily adjusted in response to significant changes in input factors or conditions. The findings contribute to understanding railway safety, offering practical insights for improving safety measures and aiding suicide prevention efforts. The high explanatory power of the predictive model underscores the critical role of societal influences in the dynamics of railway-related suicides and suicide attempts, providing valuable guidance for enhancing safety protocols and planning.