Causal structure based root cause analysis of outliers
About
We describe a formal approach to identify 'root causes' of outliers observed in $n$ variables $X_1,\dots,X_n$ in a scenario where the causal relation between the variables is a known directed acyclic graph (DAG). To this end, we first introduce a systematic way to define outlier scores. Further, we introduce the concept of 'conditional outlier score' which measures whether a value of some variable is unexpected *given the value of its parents* in the DAG, if one were to assume that the causal structure and the corresponding conditional distributions are also valid for the anomaly. Finally, we quantify to what extent the high outlier score of some target variable can be attributed to outliers of its ancestors. This quantification is defined via Shapley values from cooperative game theory.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Root Cause Analysis | CausalMan | Top-1 Accuracy11 | 11 | |
| Root Cause Analysis | CausalChamber | Top-1 Accuracy26 | 11 | |
| Root Cause Analysis | ER graph 50-node GT | Top-1 Accuracy30 | 11 | |
| Root Cause Analysis | ER graph 50-node XGES-estimated | Top-1 Accuracy27 | 11 | |
| Root Cause Analysis | ProRCA | Top-1 Accuracy50 | 11 | |
| Root Cause Analysis | 50-node ER graph (30% corruption) | Top-1 Accuracy18 | 11 | |
| Root Cause Analysis | 50-node ER graph 70% corruption | Top-1 Accuracy17 | 11 | |
| Root Cause Analysis | 50-node ER graph (50% corruption) | Top-1 Accuracy10 | 11 | |
| Root Cause Analysis | Synthetic Graph n=40 | Top-5 Precision0.00e+0 | 11 | |
| Root Cause Analysis | Synthetic Graph n=60 | Top-5 Precision0.00e+0 | 11 |