A diagnostic reasoning approach to defect prediction

  • Authors:

  • Rui Abreu;Alberto Gonzalez-Sanchez;Arjan J. C. Van Gemund

  • Affiliations:

  • Dept. of Informatics Engineering, University of Porto, Portugal and School of Computer Science, Carnegie Mellon University;Software Technology Dept., Delft University of Tech., The Netherlands;Software Technology Dept., Delft University of Tech., The Netherlands

  • Venue:
  • IEA/AIE'11 Proceedings of the 24th international conference on Industrial engineering and other applications of applied intelligent systems conference on Modern approaches in applied intelligence - Volume Part II
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

During software testing, defect prediction approaches measure current reliability status, forecasting future program failures, and provide information on how many defects need to be removed before shipping. Existing approaches often require faults to be detected and identified as a new one, before a model-based trend can be fitted. While during regression testing failures may frequently occur, it is not evident which are related to new faults. Consequently, reliability growth trending can only be performed in sync with fault identification and repair, which is often performed in between regression test cycles. In this paper we present a dynamic, reasoning approach to estimate the number of defects in the system early in the process of regression testing. Our approach, coined Dracon, is based on Bayesian fault diagnosis over abstractions of program traces (also known as program spectra). Experimental results show that Dracon systematically estimates the exact number of (injected) defects, provided sufficient tests cases are available. Furthermore, we also propose a simple, analytic performance model to assess the influence of failed test cases in the estimation. We observe that our empirical findings are in agreement with the model.