Warranty Claim Forecasting Based On Weighted Maximum Likelihood Estimation

Warranty claims reported in recent months might carry more up-to-date information than those reported in earlier months. Using weighted maximum likelihood estimation for estimating model parameters might therefore lead to better performance of warranty forecasting models than maximum likelihood estimation. This paper examines this issue and also presents comparison of the forecasting performance of the parametric models such as Poisson processes and ARIMA models and non-parametric models such as artificial neural networks. It shows that mixed non-homogenous Poisson process models can lead to better forecasting results than other competing methods. The paper also shows that the models built with the weighted maximum likelihood estimation yield smaller error than those based on the maximum likelihood estimation.