Reliability is a critical factor for assessing software quality, as it measures the software's ability to perform its intended functions without failure. In recent years, research has focused on developing more robust models for predicting software reliability. This study explores a hybrid approach for software reliability prediction by integrating Artificial Neural Networks (ANN) with the Fireworks Algorithm (FWA) and ensemble learning techniques. By leveraging FWA’s optimization capabilities, the ANN’s weights and biases are fine-tuned, and predictions are further refined using ensemble models consisting of Random Forest and Decision Tree algorithms. Using a real-time dataset of execution time and detected faults, the hybrid model was trained and evaluated, achieving high prediction accuracy, with the ensemble model yielding an R² of 0.972 and FWA optimization achieving an MSE of 0.0369 after 50 generations. The results demonstrate that combining ANN, FWA, and ensemble learning can significantly improve prediction accuracy and model reliability. Future work aims to expand this approach by incorporating additional models, exploring dynamic FWA tuning, and adapting the method across various software environments.

Software reliability Prediction model Gaussian Mutation Real-Time Systems Machine learning model.