9 Modeling Of Fault Prediction Using Machine Learning Techniques

Abstract-- Predicting faults early in the software life cycle can be used to improve software process control and achieve high software reliability. Quality of software is increasingly important and testing related issues are becoming crucial for software. Methodologies and techniques for predicting the testing effort, monitoring process costs, and measuring results can help in increasing efficiency of software testing. Being able to measure the fault-proneness of software can be a key step towards steering the software testing and improving the effectiveness of the whole process. To estimate software faultiness before and during testing and analysis activities would greatly help software testing and analysis. Many systems are delivered to users with excessive faults. It has long been recognized that seeking out fault-prone parts of the system and targeting those parts for increased quality control and testing is an effective approach to fault reduction. Despite this it is difficult to identify a reliable approach to identifying fault-prone software components. Prediction models based on software metrics, can estimate number of faults in software modules. In this paper different machine learning algorithms and neural network techniques on two different real-time software defect datasets are evaluated. The results show that when all the prediction techniques are evaluated then best algorithm for classification of the software components into faulty/fault-free systems is found to be Generalized Regression Neural Networks