Design and development of efficient defect prediction using selective software metrics

Abstract

Software Defect Prediction is one of the active research areas in newlinesoftware engineering. Defect prediction approach identifies the defect prone newlinemodules before the testing phase starts. Metrics based defect prone modules newlineimprove the software quality, reduce the cost and leading to effective newlineallocation of resources. Also, rework can be avoided and high priority can be newlineassigned to the predicted defect prone modules. Instead of considering all the newlinemetrics, it would be more appropriate to find out a suitable set of metrics newlinewhich are relevant and significant for the prediction of defect prone modules. newlineThus, the objective of this work is to design and develop an efficient defect newlineprediction model using selected software metrics in order to improve the newlinedefect prediction.The Defect Prediction Method can be developed by combining the newlineclassification technique of Data Mining and the product metrics like Halstead newlinemetrics, McCabe s Metrics and LOC based Metrics of each module measured newlineand stored as a dataset. The classification Model for defect prediction is newlinedivided into training and testing phase. In the training phase, a Model is newlinecreated based on the input of the metrics dataset with the class label with newlineeither a defective module or a non defective module. In the testing phase, the newlinecreated Model in the training phase is evaluated based on the new metrics newlinedataset without the class label. The classification based defect prediction newlineModel can be evaluated by the performance measures such as Accuracy, newlineRecall, Precision and F-Measure. A number of researchers are working on newlinedefect prediction using metrics, but there is a scope to produce better newlineperformance of the defect prediction by improving the classifier s accuracy newline newline