An ANN Based Approach for Software Fault Prediction Using Object Oriented Metrics

• Published in: Advanced Informatics for Computing Research Vol. 955; pp. 341 - 354
• Main Authors: Kaur, Rajdeep, Sharma, Sumit
• Format: Book Chapter
• Language: English
• Published: Singapore Springer 2019; Springer Singapore
• Series: Communications in Computer and Information Science
• Subjects: Artificial intelligence; Fault; Machine learning; Neural network; Software fault prediction
• ISBN: 9789811331398; 9811331391
• ISSN: 1865-0929; 1865-0937
• DOI: 10.1007/978-981-13-3140-4_31

During recent years, the enormous increase in demand for software products has been experienced. High quality software is the major demand of users. Predicting the faults in early stages will improve the quality of software and apparently reduce the development efforts or cost. Fault prediction is majorly based on the selection of technique and the metrics to predict the fault. Thus metrics selection is a critical part of software fault prediction. Currently techniques been evaluated based on traditional set of metrics. There is a need to identify the different techniques and evaluate them on the bases of appropriate metrics. In this research, Artificial neural network is used. For classification task, ANN is one of the most effective technique. Artificial neural network based SFP model is designed for classification in this study. Prediction is performed on the basis of object-oriented metrics. 5 object oriented metrics from CK and Martin metric sets are selected as input parameters. The experiments are performed on 18 public datasets from PROMISE repository. Receiver operating characteristíc curve, accuracy, and Mean squared error are taken as performance parameters for the prediction task. Results of the proposed systems signify that ANN provides significant results in terms of accuracy and error rate.