LIGHTWEIGHT DEEP LEARNING FOR EFFICIENT BUG PREDICTION IN SOFTWARE DEVELOPMENT AND CLOUD-BASED CODE ANALYSIS

Abstract

Bug prediction in software systems is crucial for improving maintenance efficiency and ensuring software reliability. Traditional approaches rely on manual log analysis and patching, which are time-consuming and inefficient, leading to delayed bug resolution and limited scalability. Existing methods lack automation in defect prediction and struggle with handling large-scale bug reports efficiently. This study introduces ConvMixer-T5D, an AI-driven bug prediction model leveraging deep learning for automated defect detection. Unlike conventional techniques, this model integrates cloud-based scalability and bug trend analysis, significantly enhancing predictive accuracy. Experimental evaluations using the Mozilla and Eclipse Defect Tracking Dataset show that ConvMixer-T5D achieves 92% software maintenance efficiency, a 17% improvement over traditional methods, while reducing critical issue response time by 50% and accelerating version upgrades by 25%. The model maintains 98% defect removal efficiency, enhances scalability by 15%, and reduces open-source patch reflection time by 50%, ensuring faster software updates. Compared to manual methods, ConvMixer-T5D automates defect handling, reducing dependency on human intervention while optimizing cloud resource management. These results demonstrate the model’s effectiveness in minimizing software vulnerabilities, reducing maintenance overhead, and improving overall defect management. This approach significantly advances the field of software maintenance by providing a robust, AI-enhanced framework for predictive bug detection, enabling efficient, scalable, and faster software development processes.