Analyzing the Severity of Freight Train Derailment Using Choquet Integral-Based Fuzzy Measure Weight Regression

Freight train derailments pose significant risks in terms of both financial and human costs. Understanding the factors contributing to derailment severity is imperative to develop efficacious risk management strategies. This study aimed to quantify the weights of various explanatory variables, including derailed speed, derailed cars, load factor, and residual train length, contributing to monetary damage while considering the dependency between these variables. We proposed a novel approach incorporating the Choquet integral-based regression multiple linear regression (CIB-MLR) model with the weights of the fuzzy measure of explanatory variables. This approach allows for modelling explanatory variables' non-linear and non-additive superposed effects on monetary damage, which is critical for understanding the complex nature of freight train derailment severity. The CIB-MLR model characterizes the importance of various explanatory variable sets contributing to derailment severity. Our analysis reveals that the most significant contributors to monetary loss in terms of the ternary and the binary fuzzy measure throughout all track classes are {Derailed Cars, Load Factor, Residual Length} (with the weight of 0.99) and {derailed cars, derailed speed} (with the weight of 0.69). In contrast, the importance of “derailed speed, derailed cars” and “derailed speed, load factor” is relatively minor. The proposed approach provides valuable insights for policymakers and industry stakeholders to improve the safety and reliability of freight train operations.