Estimating the fault content of software using the fix-on-fix model

Abstract

In statistical theory, the percentage of defects in a randomly drawn sample is an estimate of the percentage of defects in the entire population. When this concept is applied to the process of fixing faults during software development, a new fix-on-fix model results. Such a model can predict the number of software faults, thus providing a useful quality assessment. The model discussed in this paper implements the concepts of bad fix (BF) and fix on fix (FOF), which have been used in the 5ESS®-2000 switch project for several years. The FOF model is similar to error seeding models in which predetermined errors are planted in the code. The number of remaining errors can be predicted based on the number of original errors seeded and the number of both seeded and nonseeded errors found during testing. The model may initiate a new approach to software quality prediction, and it has the advantage of being independent of testing intensity, methodology, and environment. The FOF model is applicable to any software product in which BF and FOF rates can be measured from source-code management systems.