An exponential S-shaped software reliability growth model with two classes of error detectability

For software reliability assessment, it is of great importance to investigate software error detection and correction processes during the testing phase. The authors assume that there exists two classes of errors detected by software testing, i.e. some are easy to be detected early in the testing and the others more difficult to be detected later on. The former defined as Class 1 errors (the latter defined as Class 2 errors) can be modeled by and exponential (delayed S-shaped) software reliability growth model. In this paper, superposing two models above, they discuss realistic software reliability measurement, and assessment and present the numerical examples. [In Japanese.]