Causal structure based root cause analysis of outliers
Dominik Janzing, Kailash Budhathoki, Lenon Minorics, and Patrick Blöbaum Amazon Research Tübingen, Germany {janzind, minorics,
Abstract
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.
中文速览
在已知变量间因果结构（有向无环图，DAG）的前提下，当系统中出现异常观测时，如何找到这个异常的"根本原因"？研究者提出了一套基于信息论的异常评分体系，核心思路是：先用"条件异常分"衡量某个变量的取值在其父节点值已知的情况下是否依然反常，再借助合作博弈论中的夏普利值（Shapley value）把目标变量的高异常分按贡献拆分给各个祖先节点，从而定量地识别出真正"出了问题"的源头节点。在模拟数据和真实数据上的实验验证了该方法的有效性。这项工作将因果推断与异常检测结合起来，为工业监控、气候分析、欺诈检测等领域的异常溯源提供了一个理论严谨且可操作的框架。
原文 arXiv:1912.02724;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1912.02724v1