Software Defect Detection Using Optimized Support Vector Machine Based on Grey Wolf Optimizer with Random Walk

An effective and efficient software defect detection process is an urgent necessity in modern software development. This study proposes the use of a Support Vector Machine (SVM) model and Principal Component Analysis (PCA) with hyperparameters optimized using the Grey Wolf Optimizer with Random Walk (GWO-RW) for software defect detection. The dataset used consists of 12 NASA MDP datasets containing static code metrics and defect labels. The proposed method involves five main stages: dataset collection, data balancing using SMOTE, feature dimensionality reduction using PCA, SVM model training with PCA and SVM hyperparameter optimization using GWO-RW, and model evaluation. Experimental findings indicate that the GWO-RW notably enhances the performance of SVM compared to other optimizers like the original GWO and random search, with accuracies ranging from 7 6. 2 6% to 9 8. 2 1% and an average accuracy of 87.03% across all datasets. These findings illustrate that the GWO-RW is potent in optimizing the hyperparameters of the PCA and SVM and enhancing software defect detection performance.