Markovian Multiple Change-Point Modeling for Software Reliability Assessment

Abstract

We discuss software reliability modeling reflecting actual situation in a testing phase based on a Markovian software reliability modeling framework. Concretely, we discuss Markovian imperfect debugging modeling for software reliability assessment with multiple changes of testing environment. Testing-time changing the testing environment is called change-point. Taking into account the effect of change-point in software reliability growth modeling is expected to improve the accuracy of software reliability assessment because it is often observed that the stochastic characteristic of software failure-occurrence or fault-detection phenomenon is changed in an actual testing phase. Numerical examples for software reliability assessment based on our proposed approach are also shown by using actual software failure-occurrence time data. Further, we discuss the usefulness of considering the effect of the imperfect debugging and the multiple change-point into software reliability modeling by comparing the estimated behavior of the mean time between software failures based on our model and the existing related models.