Software Reliability Modeling with Software Metrics Data via Gaussian Processes Articles uri icon

publication date

  • August 2013

start page

  • 1179

end page

  • 1186

issue

  • 8

volume

  • 39

International Standard Serial Number (ISSN)

  • 0098-5589

Electronic International Standard Serial Number (EISSN)

  • 1939-3520

abstract

  • In this paper, we describe statistical inference and prediction for software reliability models in the presence of covariate information. Specifically, we develop a semiparametric, Bayesian model using Gaussian processes to estimate the numbers of software failures over various time periods when it is assumed that the software is changed after each time period and that software metrics information is available after each update. Model comparison is also carried out using the deviance information criterion, and predictive inferences on future failures are shown. Real-life examples are presented to illustrate the approach.

keywords

  • bayes methods; gaussian processes; inference mechanisms; software metrics; software reliability; statistical analysis; system recovery; covariate information; deviance information criterion; future failure; predictive inference; semiparametric bayesian model; software failure; software metrics information; software reliability modeling; statistical inference