Deep Learning Driven Bug Detection and Severity Classification in Multi-Tenant Cloud Systems

Bug tracking in cloud-based environments—especially within multi-tenant SaaS systems—faces challenges such as incomplete user reports, delayed detection, and inaccurate defect classification. Traditional approaches often fail to handle the dynamic complexity of cloud infrastructures and fail to ensure timely resolution of issues. In this study, we propose an improved bug tracking system that integrates deep learning with interactive real-time feedback to overcome these limitations. A Multilayer Perceptron (MLP) model is used for early defect detection and severity classification, while an interactive reporting interface prompts users to submit more comprehensive and structured bug reports. Furthermore, the system incorporates the Cloud-Odc framework to classify defects based on severity, affected components, and cloud service layers (SaaS, PaaS, IaaS). Experimental evaluations on a dataset of more than 8,000 annotated bug reports demonstrate a 40% improvement in detection accuracy and a significant reduction in resolution time. The proposed system improves report completeness, reduces developer-user interaction overhead, and improves the scalability and efficiency of defect management in modern cloud-based workflows.