Neural network based software reliability prediction with the feed of testing process knowledge

Abstract

Software reliability is an important factor for evaluating software quality in the domain of safety-critical software. Traditional software reliability growth models (SRGMs) only uses the failure data produced in a testing process to evaluate the software reliability and its growth. However, the number and severity of the failures uncovered are determined by the effectiveness and efficiency of testing process. Therefore, an unbiased reliability prediction has to take test process knowledge into account. In this paper, we proposed a neural network based reliability prediction method to utilize the testing process metrics to correlate testing process knowledge with failure prediction. The metrics designed in this paper cover information from the system under test (SUT), design of testing, software failure and repair process. The method is validated through the testing process data collected from a real embedded operating system. And the results show a fairly accurate prediction capability. © 2013 Springer-Verlag.