Comparison of Artificial Neural Network and Maximum Likelihood Models for Evaluating Reliability in Right-Censored Data Analysis

Abstract

In this paper, we conduct a comparative analysis of the maximum likelihood method and an artificial neural network model for estimating a new power function under Type-II right censoring. The primary objective of this study is to estimate various reliability measures, the mean residual life function, variance residual life function, mean inactivity time function, and variance inactivity time function, using both maximum likelihood estimation and artificial neural network models. Additionally, we explore the behavior of these measures for Type-II censored data through a numerical example. Therefore, we see that the artificial neural network can make accurate predictions and is a powerful approach which can be used in reliability analysis based on Type-II censoring.