Deep Learning Based Multilayer D_Highway Memory Neural Network for Software Defect Prediction Using Software Metrics

Abstract

As technology has progressed, new hardware and software needs have evolved. There has been a significant increase in demand for software across many different applications, which has coincided with the rise of the software industry. To expand the software business, producing and maintaining high-quality software is regarded to be the most crucial duty for every company. Software engineering is critical to the software industry's success in achieving this goal. In order to create software applications, computer code is used to accomplish the required goal. Software faults may cause defective software to be developed if these scripts include certain incorrect examples. In the realm of software engineering, predicting software defects is regarded as the most significant activity that can be utilized to ensure the quality of software. As a consequence of defect prediction findings, quality assurance teams may more efficiently allocate limited resources for testing software products by focusing their efforts on the source code that is most likely to have defects. Using defect prediction methods to help developers and speed up the time to market for more dependable software products will become more vital as software projects grow in size. The process of detecting and correcting flaws is one of the most time-consuming and expensive parts of embedded software development. Complex infrastructure, large scale, high costs and time constraints make it difficult to monitor and meet quality standards. A unique deep learning-based software fault prediction model has been shown in this study. A single spectrum stream flow filter was first used to remove data errors. The minkowski prewitt clustering approach was then utilized to group the software characteristics. Then, using the first-order binary - α sail fish optimization technique, we choose the best defective characteristics. Finally, a multilayer D highway memory neural network is used to identify different sorts of software error codes according on the severity of the problem. Using the PROMISE JM1 software defect prediction dataset in the python environment, the whole process was developed. Accuracy, AUC, precision, recall, and F1 score were used to evaluate the efficacy of the proposed methodology's performance.