This paper deals with the development of a model to predict the products’ terminal call rate (TCR) during the warranty period. TCR represents a key information for a quality management department to reserve the necessary funds for product repair during the warranty period. TCR prediction is often carried out by parametric models such as Poisson processes, ARIMA models and maximum likelihood estimation. Little research has been done with machine learning methods (MLM). Therefore, this paper addresses the utilization of machine learning methods (MLM), such as regression trees, ensembles of regression trees and neural networks in order to estimate the parameters of different models for TCR prediction. MLM were tested on exponential and logistic non-linear models, which best describe the shape of the cumulative density function of the failed products. The estimated cumulative density function was used to predict the TCR. The results have shown the ensembles of regression trees yield the smallest TCR prediction error among the tested MLM methods.