Software Defect Estimation using Data Mining Techniques: Experimental Study of Algorithms on “PROMISE” Repository

Nowadays, people are becoming self-employed software developers due to the growing demand for mobile apps and other forms of computerized software. However, Software Engineering is a broad field since it necessitates open lines of communication among all parties involved in the system's development and the timely and cost-effective delivery of the system itself. Acquiring excellent performance by reducing the size of the system is an integral part of meeting consumer expectations. But, the most expensive part of developing software is defects estimation. Metrics collected from software may provide some insight into the presence of defects. In the final stages of software testing, defect estimates may be produced. Furthermore, estimating defects in the early stages of the software development life cycle (SDLC) is among the critical fields of study only with the goal of cost-effectiveness and appropriate resource planning. In order to improve the efficiency of defect estimation, various data mining techniques have been applied, including Support Vector Machines, C4.5, Bagging, Naive Bayes, K-Nearest Neighbors, Neural Networks, Random Forest, Decision Trees, and Radial Bias. These techniques are applied to datasets from NASA dataset repositories and evaluated using various metrics for software defects estimation (SDE). The results of our study indicated that the Support Vector Machine algorithm had the highest performance among the data mining techniques evaluated. The algorithm achieved a precision of 43.74%, recall of 50.00%, accuracy of 87.48%, and an F-1 measure of 81.64%. It is essential to consider various evaluation metrics, including recall, F-measure, precision, and accuracy, when performing software defects estimation.