Applying various learning curves to hyper-geometric distribution software reliability growth model

Abstract

The Hyper-Geometric Distribution software reliability growth Model (HGDM) has been shown to be able to estimate the number of faults initially resident in a program at the beginning of the test-and-debug phase.A key factor of the HGDM is the "sensitivity factar'', which represents the number of faults discovered and rediscovered at the application of Q test instance. The learning curve incorporated in the sensitivity factor is generally assumed to be linear in the literature. However, this assumption is apparently not realistic in many applications. We propose two new sensitivity factors based on the exponential learning curve and the S-shaped learning curve, respectively. Furthermore, the growth curves of the cumulative number of discovered faults for the HGDM with the proposed learning curves are investigated. Extensive experiments have been performed based on two real test/debug data sets, and the results show that the HGDM with the proposed learning curves estimates the number of initial faults better than previous approaches.