Defects prediction on software visualization and hybrid optimization by deep neural network

Newly developed software often encounters challenges in defect prediction using deep learning due to the absence of historical data. This paper introduces a software classification method based on visualization and residual neural networks to select the most suitable data for new software, thereby reducing the adverse effects of inappropriate or mismatched datasets on defect prediction. In the proposed approach, a deep neural network combined with hybrid PSO–SSA hyperparameter optimization is employed to tr ain an optimal dataset and enhance the overall performance of defect prediction.