On a Bayesian model for failures prediction in underground trains

摘要

We consider nonhomogeneous Poisson processes to analyse data about failures in underground trains over a eight-year period. A double scale is considered because both time and kilometers are recorded for each failure. Different choices to model the relation between operated time and run kilometers are presented, whereas a periodic component is introduced to deal with the seasonality in the data. Predictions over future failures are considered as well.