面向云原生智能运维系统的因果图恢复与根因定位方法
By constructing natural language scene templates and generating edge-level causal effect estimates using pre-trained large language models, and combining multi-task causal effect signature vectors and path type labels, the technical problem of causal graphs in cloud-native/microservice AIOps scenarios is solved. This addresses the issues of low accuracy in causal graph recovery and misjudgment of root cause localization in existing technologies, achieving accurate recovery of causal graphs and efficient localization of fault roots.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NAT UNIV OF DEFENSE TECH
- Filing Date
- 2026-05-19
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies struggle to accurately recover causal graphs and locate root causes in high-dimensional, small-sample, and knowledge-intensive domains within cloud-native/microservice AIOps scenarios. They also fail to effectively integrate large language models with traditional causal structure learning algorithms, resulting in high misjudgment rates and uninterpretable results in root cause localization.
By constructing natural language scene templates, using pre-trained large language models to generate edge-level causal effect estimates, and combining multi-task causal effect signature vectors and path type labels, a soft skeleton weight matrix is constructed and injected as a soft constraint into a traditional causal structure learning algorithm to optimize the causal graph adjacency matrix to achieve fault propagation analysis and root cause localization.
It achieves accurate recovery of cause-effect graphs and efficient location of root causes of failures, reduces the false positive rate of root cause location, ensures the interpretability and auditability of results, and adapts to the needs of cloud-native/microservice AIOps scenarios.
Smart Images
Figure CN122220138B_ABST