Trade-off Assessment between Model Adequacy and Complexity in Nonhomogeneous Poisson Process Software Reliability Growth Models Incorporating a Changepoint  
Author Vidhyashree Nagaraju


Co-Author(s) Shadow Pritchard; Priscila Silva; Lance Fiondella


Abstract Traditional Non-homogeneous Poisson process (NHPP) software reliability growth models (SRGM) enable quantitative assessment of software systems based on failure data collected during testing. However,traditional models assume failure data is characterized by a single continuous curve without considering several factors that could significantly impact the fault detection rate. To address this, many studies have developed models to incorporate changepoints or imperfect debugging or both, yet existing studies develop models without careful consideration of the relationships between models and their complexity. Therefore, this paper presents a sequence of progressively more complex software reliability growth models according to their nesting relationships, including imperfect debugging and error generation as well as changepoint models with imperfect debugging and error generation. Model selection based on a novel multi-criteria approach is demonstrated. Our results indicate that the most complex models are not recommended and that simpler models exhibit the desirable attributes of simplicity and visual goodness of fit as well as information theoretic and other measures of goodness of fit.


Keywords Software reliability, software reliability growth model, non-homogeneous Poisson process, imperfect debugging, changepoint, error generation
    Article #:  RQD28-145

Proceedings of 28th ISSAT International Conference on Reliability & Quality in Design
August 3-5, 2023