Heterogeneous redundancy in software quality prediction using a hybrid Bayesian approach

  • With the ever-increasing significance of software in our everyday lives, it is vital to afford reliable software quality estimates. Typically, quantitative software quality analyses rely on either statistical fault prediction methods (FPMs) or stochastic software reliability growth models (SRGMs). Adopting solely FPMs or SRGMs, though, may result in biased predictions that do not account for uncertainty in the distinct prediction methods; thus rendering the prediction less reliable. This paper identifies flaws of the individual prediction methods and suggests a hybrid prediction approach that combines FPMs and SRGMs. We adopt FPMs for initially estimating the expected number of failures for fi- nite failure SRGMs. Initial parameter estimates yield more accurate reliability predictions until sufficient failures are observed that enable stable parameter estimates in SRGMs. Being at the equilibrium level of FPM and SRGM pre- dictions we suggest combining the competing prediction methods with respect to the principle of heterogeneous redundancy. That is, we propose using the in- dividual methods separately and combining their predictions. In this paper we suggest Bayesian model averaging (BMA) for combining the different methods. The hybrid approach allows early reliability estimates and encourages higher confidence in software quality predictions.

Export metadata

  • Export Bibtex
  • Export RIS

Additional Services

Share in Twitter Search Google Scholar
Metadaten
Author:G. Hanselmann, A. Sarishvili
URN (permanent link):urn:nbn:de:hbz:386-kluedo-15476
Serie (Series number):Berichte des Fraunhofer-Instituts für Techno- und Wirtschaftsmathematik (ITWM Report) (125)
Document Type:Report
Language of publication:English
Year of Completion:2007
Year of Publication:2007
Publishing Institute:Fraunhofer-Institut für Techno- und Wirtschaftsmathematik
Tag:Bayesian Model Averaging; Fault Prediction ; Non-homogeneous Poisson Process ; Reliability Prediction
Faculties / Organisational entities:Fraunhofer (ITWM)
DDC-Cassification:510 Mathematik

$Rev: 12793 $