A NEURAL-NETWORK APPROACH FOR EARLY DETECTION OF PROGRAM MODULES HAVING HIGH-RISK IN THE MAINTENANCE PHASE

A neural network model is developed to classify program modules as eit her high or low risk based on multiple criterion variables. The inputs to the model include a selection of software complexity metrics colle cted from a telecommunications system. Two criterion variables are use d for class determination: the number of changes to enhance the progra m modules, and the number of changes required to remove faults from th e modules. The data were deliberately biased to magnify differences in metrics values between the discriminant groups. The technique display ed a low classification error rate. This success, and the absence of t he data assumptions typical of statistical techniques, demonstrate the utility of neural networks in isolating high-risk modules where class determination is based on multiple quality metrics.