ABSTRACT
An essential attribute of evaluating a software product’s quality is its reliability, which is an integral part of software quality. It is challenging for the software industry to develop highly reliable software due to various challenges. In the field of software reliability prediction, machine learning (ML) techniques have demonstrated meticulous and remarkable results. Here, we propose the use of machine learning techniques for predicting software reliability and evaluating their performance. In addition to predicting better results than statistical methods, machine learning can also be used to accurately and precisely predict software failures. For modeling complex software with complex phenomena, these techniques consider past failure data as input and require relatively little assumption. A machine learning methodology is based on learning automatically and allows computers to predict system behavior based on historical failure data and current failures. In this study, we propose novel machine learning architecture transformers that are more powerful parallel processing models and that are the most suitable and powerful solutions for better software reliability; they can be very helpful in defect detection with higher-end performance.
