An explainable artificial intelligence approach for mud pumping prediction in railway track based on GIS information and in-service train monitoring data

•GIS analysis is conducted to quantify the hydrological information for mud pumping prediction.•In-service train data and hydrological variables are combined to develop predictive model.•Incorporating hydrological variables can improve prediction performance.•Explainable deep learning can be used to reveal the effects of a single variable on the probability of mud pumping.•The most vulnerable sections for mud pumping in railway tracks can be identified.