Neural-network techniques for software-quality evaluation

Software quality modeling involves identifying fault-prone modules and predicting the number of errors in the early stages of the software development life cycle. This paper investigates the viability of several neural network techniques for software quality evaluation (SQE). We have implemented a principal component analysis technique (used in SQE) with two different neural network training rules, and have classified software modules as fault-prone or nonfault-prone using software complexity metric data. Our results reveal that neural network techniques provide a good management tool in a software engineering environment.