Application of Neural Networks for Software Quality Prediction Using Object-Oriented Metrics

This paper presents the application of neural networksin software quality estimation using object-orientedmetrics. Quality estimation includes estimating reliabilityas well as maintainability of a software. Reliability istypically measured as the number of defects.Maintenance effort can be measured as the number oflines changed per class. In this paper, two kinds ofinvestigation are performed. The first on predicting thenumber of defects in a class and the second on predictingthe number of lines change per class. Two neuralnetwork models are used, they are Ward neural networkand General Regression neural network (GRNN). Objectoriented design metrics concerning inheritance relatedmeasures, complexity measures, cohesion measures,coupling measures and memory allocation measures areused as the independent variables. GRNN network modelis found to predict more accurately than Ward networkmodel