An Isolation Forest Algorithm for the Detection of Effective Software Defects Estimation

Abstract

Software metrics are affected by many factors, including reliability, usability, integrity, and maintainability. The metrics are important for measuring software performance, planning work items, measuring productivity, debugging, and estimating cost. To estimate the expected delivered quality and maintenance effort, several industries seek to investigate the number of bugs in software modules before they are delivered. Hence, automated defect estimation has been a crucial and fundamental task in the field of software development. To assist with this endeavour, a huge number of software metrics and statistical methods have been proposed by researchers, and an equally extensive body of written material has been produced. Current software frameworks are generally huge and complicated; they have many associated metrics that capture various elements of the software modules. This study reviewed all the prediction techniques and discussed various projects that have been studied in recent years. Furthermore, using the results of this study, we have proposed an Isolation Forest Defect Estimation (IFDE-Framework) for identifying defects in software as it can detect previously unknown and abnormal behaviours. Moreover, we can provide difficult challenges for the successive stage in the process of software defect prediction.