Fault-Tolerant Algorithm for Software Preduction Using Machine Learning Techniques

Abstract

Many software reliability algorithms have been used to predict and approximate the reliability of software. One general expectation of these traditional algorithms is to predict the fault and automatically delete the observed faults. This presumption will not be reasonable in practice and may not always exist. In this paper, the various algorithms have been used such as probabilistic neural network (PNN), generalized neural network (GRNN), linear regression, support vector machine (SVM), bagging, decision trees (DTs), and k-nearest neighbor (KNN) to measure the accuracy of various data and comparison has been done. The proposed algorithm has been used for predicting the reliability of software and the algorithms have been implemented to check the accuracy while using different machine learning (ML) techniques. Experimental studies based on actual failure evidence indicate that the proposed algorithm can more effectively explain the change in failure data and predict the software development behavior than conventional techniques.