MODEL SELECTION IN THE NON-HOMOGENEOUS POISSON PROCESS FOR THE RELIABILITY OF A MACHINE

Model selection plays an important role, especially in critical areas such as reliability analysis. Non-homogeneous Poisson process (NHPP) is a statistical approach widely used to model the frequency of time-dependent events. In such processes, correct model selection is vital to effectively evaluate the performance and reliability of the system. In this study, two basic models frequently used in NHPP, the power law model and the log-linear model, are compared. The research focuses on the reliability of ATM machines of a bank. The reliability performance and expected failure numbers of both models are calculated using the data between the failure times of ATMs. In the study, the predictive power and accuracy of the models are compared; their suitability to real data is evaluated with statistical methods. The obtained results are compared and it is decided which model is a better choice.