Hybrid Ensemble White Shark Optimizer for Fine-Grained Software Bug Detection

Abstract

The increasing complexity of modern software systems has led to a rapid rise in software defects, making reliable bug detection and software classification essential for improving software quality and maintainability. Traditionally, software defect prediction relied on manual inspection, rule-based analysis, and conventional statistical methods. These approaches often required expert intervention, were time-consuming, and struggled to handle large-scale and high-dimensional software metrics data effectively. Machine learning techniques such as Bernoulli Naive Bayes (BNB), Multinomial Naive Bayes (MNB), and Linear Discriminant Analysis (LDA) were later introduced to automate the prediction process; however, these models exhibited limitations in capturing complex nonlinear relationships among software metrics and often produced suboptimal performance on diverse datasets. To address these challenges, this work proposes a hybrid intelligent system, Hybrid Ensemble White Shark Optimizer Convolutional Neural Network Extra Trees (HE-WSO-CNN-ET), designed for efficient bug detection and software classification. The proposed system integrates Convolutional Neural Network (CNN) for feature representation learning, White Shark Optimizer (WSO) for adaptive control, and Extra Trees (ET) for classification. The system processes software metrics such as Weighted Methods per Class (WMC), Depth of Inheritance Tree (DIT), Number of Children (NOC), Coupling Between Objects (CBO), Response For a Class (RFC), Lack of Cohesion in Methods (LCOM), Afferent Couplings (CA), Efferent Couplings (CE), Number of Public Methods (NPM), Lines of Code (LOC), and complexity-related attributes to identify defect patterns and classify software types. The proposed approach aims to enhance prediction accuracy, reduce manual effort, and improve scalability in handling real-world software datasets. By leveraging ensemble learning with intelligent feature mapping, the system provides a more robust and efficient solution compared to traditional methods. The experimental results demonstrate improved performance in terms of accuracy, precision, recall, and F1-score, highlighting the effectiveness of the proposed system in software defect prediction and classification tasks.