Predicting software defects is a critical component of software quality control. It is essential to plan for early defect detection and mitigation to enhance performance and reliability. Traditional machine learning and deep learning models often face challenges in managing missing values, extracting meaningful features, and effectively distinguishing between defective and non-defective software modules due to their reliance on linear classifiers and limited feature representation capabilities. To address these challenges, this study proposes an Entangling Quantum Generative Adversarial Network with Football Optimization Algorithm (EQGAN-FbOA) for efficient software defect prediction. The PROMISE dataset has been collected, and missing values are imputed during the pre-processing stage using Diffusion Models for Missing Value Imputation (DMVI), which employs a forward noising process to introduce controlled noise and a reverse denoising process to reconstruct the missing data. To enhance computational performance, a Spike-driven Transformer (S-DT) that incorporates a Leaky Integrate-and-Fire (LIF) spiking neuron is utilized for feature extraction. The EQGAN model improves defect prediction by generating quantum-enhanced feature representations. Additionally, the Football Optimization Algorithm (FbOA) is applied to balance exploration and exploitation through football-inspired search strategies, thereby preventing premature convergence and refining defect classification. Experimental findings from the PROMISE dataset demonstrate that the proposed method surpasses existing approaches, achieving a software defect prediction accuracy of 99.8%, precision of 99.7%, recall of 99.6%, Matthews correlation coefficient (MCC) of 99.5%, and F-measure of 99.4%.

Diffusion Models for Missing Value Imputation Entangling Quantum Generative Adversarial Networks Football Optimization Algorithm Spike-Driven Transformer Software Defect Prediction