Another Look at Non-homogeneous Markovian Software Reliability Modeling
Nanxiang Qiu, Siqiao Li, Tadashi Dohi, Hiroyuki Okamura · 2023
This paper explores yet another software reliability modeling framework based on non-homogeneous Markov processes (NHMPs). For two subclasses of NHMPs; generalized binomial processes (GBPs) and generalized Polya processes (GPPs), we formulate 22 novel NHMP-based software reliability models (SRMs) with 11 kinds of baseline intensity functions, which are different from the existing NHMP-based SRMs by Li et al. (2023). Our evaluation of NHMP-based SRMs focuses on assessing their goodness-of-fit and predictive performances using 8 sets of software fault detection time-interval data (group data). The results are compared with the infinite-failure NHPP-based SRMs. Through comprehensive numerical experiments, we show that our new modeling framework could provide the strengths in the goodness-of-fit performance but have some limitations on the predictive performance in the early testing phase.