A Hybrid Approach to Identify the Maximum Likelihood Estimates of a Two Changepoint Goel-Okumoto Software Reliability Growth Model  
Author Vidhyashree Nagaraju
Co-Author(s) Lance Fiondella; Thierry Wandji
Abstract Non-homogeneous Poisson process software reliability growth models with changepoints can characterize a software testing process possessing factors that affect the fault detection process such as the testing strategy and environment, integration testing, and resource allocation. Most changepoint research focuses on single changepoint models due to their increased computational complexity and the instability of numerical algorithms for model fitting. To overcome these limitations, this paper proposes a hybrid approach that combines an initial parameter estimation strategy based on the EM algorithm, the expectation conditional maximization algorithm, and Newton’s method to identify the maximum likelihood estimates of the Goel-Okumoto software reliability model with two changepoints. Our results indicate that the 2-CP model achieves a better goodness of fit than the 1-CP model and that the proposed approach maximizes the log-likelihood more effectively on eight of 10 data sets considered.
Keywords Software reliability, changepoint, expectation conditional maximization algorithm, maximum likelihood estimation, Newton-Raphson method
   
    Article #:  23-059
 
Proceedings of the 23rd ISSAT International Conference on Reliability and Quality in Design
August 3-5, 2017 - Chicago, Illinois, U.S.A.