Intelligent Defect Detection and Risk Assessment for Cloud Platforms Using Counterfactual System Modeling

Abstract

This study addresses the challenges of hidden defect behavior, complex causes, and strong sensitivity to operating condition changes in cloud computing environments. It proposes a defect detection method oriented toward understanding system operation mechanisms. The method is based on multi-source monitoring data and models system operation as continuously evolving state sequences. System behavior deviation is characterized through state prediction and consistency analysis. Conditional intervention modeling is introduced within a unified framework to construct alternative operational scenarios and evaluate the impact of key condition changes on system behavior. By jointly modeling observed states and state responses under condition changes, the method highlights core factors that are highly relevant to defect discrimination in complex dynamic environments. It reduces misjudgment risk caused by noise interference and incidental fluctuations. Under a unified data setting, comparative analysis with several representative methods shows that the proposed approach achieves more consistent performance in overall discrimination stability and risk differentiation capability. This work provides a mechanism-aware modeling perspective for defect identification and risk analysis in complex cloud platforms. It contributes to a deeper understanding and more reliable analysis of system anomalies in intelligent operations scenarios.