Software Reliability Assessment Using Machine Learning Technique

• Published in: Computational Science and Its Applications - ICCSA 2018 Vol. 10964; pp. 403 - 411
• Main Authors: Behera, Ranjan Kumar, Shukla, Suyash, Rath, Santanu Kumar, Misra, Sanjay
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2018; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Mean time to failure; Recurrent neural network; Software reliability; Support Vector Machine
• ISBN: 9783319951737; 3319951734
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-95174-4_32

Software reliability is one of the major attributes in software quality assurance system. A large number of research works have been attempted in order to improve the reliability of the software. Research directions in improving software reliability may be defined in a three-step process i.e., software modeling, software measurement and software improvement. Each of these phases is equally important in obtaining reliable software system. It is important to achieve better accuracy in estimating reliability in order to manage the software quality. A number of metrics have been proposed in the literature to evaluating the reliability of a software. Machine learning approaches are found to be suitable ways in evaluating different parameters of software reliability. Several machine learning techniques have been evolved in order to capture the different characteristics of a software system. The machine learning algorithms like naive bayes, support vector regression, decision tree and random forest algorithms are found to be successful in classifying the bug data from data where feature sets are dependent with each other. In this paper, deep learning approach has been proposed to estimate the reliability of software. The proposed approach uses recurrent neural network for predicting the number of bugs or failure in software. Effectiveness of deep learning is extensively compared with the standard machine learning algorithms by considering the dataset collected from the literature.