Predicting bugs in software is necessary for enhancing the quality of software project and reducing maintenance-related risks by identifying problematic software components. It is a methodology for creating a model that can be employed by software professionals and academics throughout the commencement of software development life cycle (SDLC) to detect buggy modules or classes. In this study, we introduced a hybrid metaheuristic swarm intelligence based deep learning model for early-stage prediction of software bugs. Combination of whale optimizer algorithm (WOA) as well as grey wolf optimizer (GWO) is used for optimizing the hyperparameters of Convolutional neural network. The effectiveness of proposed model is esitmated against 6 java-based projects. The proposed model’s predictive performance is examined based on several performance measures namely f1 score, MCC, accuracy, precision, AUC, as well as recall. The proposed model’s predictive performance is compared with CNN model without HPT, CNN with GWO as HPT approach, and CNN with WOA as HPT approach. The result of this study indicates that the hybrid model of GWO and WOA has significantly improved the predictive performance of CNN model in respect of accuracy, f1 score, AUC, recall, MCC, as well as precision. It is also found that the rankings and superiority of hyperparameter tuning methods is hybrid of GWO and WOA > GWO > WOA.