A Complete Study of Different Models for Software Fault Prediction

Meetesh Nevendra

doi:10.37591/josettt.v5i2.1747

Abstract

Context- Software faults prediction improves software quality, software reliability, and software efficiency by early identification of faults. Numerous classification methods have been suggested for this task.Objective- The main objectives of the studies are (i) find the number of defects in AEEEM defect datasets, (ii) Challenges faced with Imbalanced Datasets.Method- We use AEEEM defect datasets for the prediction of faulty classes using four deferent machine learning technique. This particular paper present software fault prediction problem using individual and ensemble approach.Result- The results show that the balancing using SMOTE with random forest and AdaBoost as ensemble classifier has more predictive capability for predicting faults.Conclusion- The results ensure that the predictive capability of various machine learning techniques with regard to developing fault prediction modelsKeywords: SMOTE, AdaBoost, ADASYN, Transfer component analysis (TCA), random forest (RF), Fuzzy C Means (FCM)Cite this ArticleMeetesh Nevendra, Pradeep Singh. A Complete Study of Different Models forSoftware Fault Prediction. Journal of Software Engineering Tools & Technology Trends. 2018; 5(2): 1–10p.