AN EFFICIENT SAMPLING SCHEME FOR ESTIMATING SOFTWARE RELIABILITY WITH ASSOCIATED COST

Kamel Rekab (USA), Xin Yan (USA) and Herbert H. Thompson (USA)

Received September 7, 2007

Abstract

Statistical methods can be applied to testing software to create an estimate based on a sample of test cases. Often the risk of the software test is assessed based on a predefined loss function only. We introduce a cost  in addition to the loss function, which is associated with each software partition i and represents the cost to execute one test case from that domain. For a given fixed total number of test cases, we present method on how to allocate these test cases among partitions of the software so that it minimizes the expected loss and cost incurred by the Bayes estimator of the overall reliability. Method on how to determine stopping time for the iterative allocation sampling scheme is also proposed. A simulation results indicate significant improvement by proposed method over the standard industry practice where the total number of test cases is pre-determined.

Keywords and phrases: efficient sampling scheme, software reliability, testing, Bayes risk, loss function, cost.