identifying factors influencing freight train derailment severity: a comparative study using machine learning algorithms

The aim of this study is to explore patterns and relationships between factors affecting the severity of freight train derailments, with a focus on utilizing machine learning techniques, particularly random forest, support vector machine (svm), and adaboost, to identify key features. the data for this research were obtained from the united states federal railroad administration (fra) database over the period from 2010 to 2022. in addition to identifying significant features using machine learning models, this study develops predictive models, evaluates their accuracy, and performs statistical model analysis. machine learning methods offer advantages in handling complex datasets and extracting nonlinear relationships, which can be effective in understanding the dynamics of rail incidents. the results indicate that the adaboost model achieved superior performance in predicting derailment severity, with an accuracy of 92.5%. key identified features include the number of cars, driver visibility conditions, and vehicle type. this study may contribute to a better understanding of risk patterns and play an important role in enhancing rail safety.