Predicted of Software Fault Based on Random Forest and K-Nearest Neighbor

Abstract

Software systems have gotten increasingly complicated and adaptable in today's computer world. As a result, it's critical to track down and fix software design flaws on a regular basis. Software fault prediction in early phase is useful for enhancing software quality and for reducing software testing time and expense; it's a technique for predicting problems using historical data. To anticipate software flaws from historical databases, several machine learning approaches are applied. This paper focuses on creating a predictor to predict software defects, Based on previous data. For this purpose, a supervised machine learning techniques was utilized to forecast future software failures, K-Nearest Neighbor (KNN) and Random Forest (RF) applied technique applied to the defective data set belonging to the NASA's PROMISE repository. Also, a set of performance measures such as accuracy, precision, recall and f1 measure were used to evaluate the performance of the models. This paper showed a good performance of the RF model compared to the KNN model resulting in a maximum and minimum accuracy are 99%,88% on the MC1 and KC1 responsibly. In general, the study's findings suggest that software defect metrics may be used to determine the problematic module, and that the RF model can be used to anticipate software errors.