Enhancement of Defect Prediction Using Regression Learning Method for Quality Software Development

In software defect prediction, false alarms can lead to inefficient resource allocation and wasted time.Through balancing the accurate identification of defect-prone modules while minimizing false alarms is essential.This can be achieved by refining classification models through the use of advanced machine learning algorithms, finetuning parameters, and incorporating additional attributes that better capture the characteristics of defect-prone modules.Existing defect prediction classifiers often rely on past project data to learn patterns, but they face challenges when applied to new projects with different design parameters or contexts.Therefore, it is crucial to develop robust and adaptable models to improve defect prediction accuracy and efficiency.In this paper, we propose a Regression Learning Method for Defect Prediction (RLM-DP) in software development aimed at enhancing prediction accuracy.It analyze RLM-DP using both feature optimization (FO) and feature selection (FS) approaches to identify valuable attributes and understand the factors influencing software defects.The performance of RLM-DP is evaluated by comparing models without-FO and with-FO in terms of precision, recall, and accuracy on the CM1 dataset.It result outcome indicates that all ML classifiers with RLM-DP show improved outcomes, with the SVM classifier achieving the highest accuracy of 99.98%.Furthermore, a comparative analysis without-FS and with-FS is performed on four different PROMISE datasets.The results demonstrate that RLM-DP method achieved a better accuracy ranging from 3% to 8% with all other classifiers in comparison.