NOVEL CLUSTERED BAGGING MODEL USING COGNITIVE OBJECT ORIENTED METRICS FOR CROSS-PROJECT SOFTWARE DEFECT PREDICTION

Software defect prediction models focus on identifying modules that are prone to be defective. Identifying such fine-grained elements requires sufficient training data, which might not be available for all projects. This work presents a Novel Clustered Bagging (NCB) model for cross project defect prediction in software. The proposed work operates by identifying similar metrics between the training and the test data. High similarity levels and the cognitive nature of the features are used to select candidate features. Bagging ensemble model has been created, that implements effective bag creation mechanism has been used for the defect prediction process. Experiments were performed with the PROMISE dataset containing ten varied defect data. Cross project predictions has been performed, and results indicate high performances. Comparison with existing model indicates an average increase in True Positive Rate (TPR) level at 5% and a reduction inFalse Positive Rate (FPR) level at 10%, indicating effective performance.