NetCause: Counterfactual Learning for Root Cause Analysis in Large-Scale Networks
Researchers have developed NetCause, a self-supervised learning framework designed to identify the root causes of network incidents. This system models network failures as graph-temporal processes and employs counterfactual simulations to rank potential root causes, offering an interpretable output for operators. Trained on over 1,500 incidents from a major cloud provider, NetCause demonstrated a 16.1% improvement in root cause ranking accuracy compared to a rule-based baseline, with efficient inference times. AI