Author : Rahul Jadon and Purandhar. N

The present study aims to develop intelligent bug prediction and fixing systems in software development through deep learning models. The framework brings in attention mechanisms to focus on important parts and improve the prediction accuracy. It utilizes Bidirectional Long Short Memory (BiLSTM) and Transformer networks to find fault-prone code segments based on analyzed historical commit data, version control logs, and past bug reports. Moreover, Discrete Wavelet Transform (DWT) has been applied as a feature extraction method to capture the important pattern embedded within the code. The model aims to predict bugs and provide solutions based on past bug fixes so that debugging can be done more quickly. The performance of the system measured in terms of some important metrics states that it provides accuracy (99.38%), precision (99.01%), recall (99.75%), and F1-score (99.38%), thus proving its role in reducing the debugging time and improving the software quality. This Intelligent Model is fully applicable to the modern software development environment, especially in CI/CD pipelines.

Keywords: Bug Prediction, Deep Learning, Bidirectional Long Short Memory, Feature Extraction, Discrete Wavelet Transform, Software Development, Code Quality, CI/CD Integration.