A Practical Software Quality Classification Model Using Genetic Programming

A software quality estimation model is an important tool for a given software quality assurance initiative. Software quality classification models can be used to indicate which program modules are fault-prone (FP) and not fault-prone (NFP). Such models assume that enough resources are available for quality improvement of all the modules predicted as FP. In conjunction with a software quality classification model, a quality-based ranking of program modules has practical benefits since priority can be given to modules that are more FP. However, such a ranking cannot be achieved by traditional classification techniques. We present a novel software quality classification model based on multi-objective optimization with genetic programming (GP). More specifically, the GP-based model provides both a classification (FP or NFP) and a quality-based ranking for the program modules. The quality factor used to rank the modules is typically the number of faults or defects associated with a module. Genetic programming is ideally suited for optimizing multiple criteria simultaneously. In our study, three performance criteria are used to evolve a GP-based software quality model: classification performance, module ranking, and size of the GP tree. The third criterion addresses a commonly observed phenomena in GP,that is, bloating. The proposed model is investigated with case studies of software measurement data obtained from two industrial software systems.