Anomaly Attribution of Multivariate Time Series using Counterfactual Reasoning

Abstract--There are numerous methods for detecting anomalies in time series, but that is only the first step to understanding them. We strive to exceed this by explaining those anomalies. Thus we develop a novel attribution scheme for multivariate time series relying on counterfactual reasoning. We aim to answer the counterfactual question of would the anomalous event have occurred if the subset of the involved variables had been more similarly distributed to the data outside of the anomalous interval. By determining which variables yield the lowest anomaly score Finding causes of extreme weather events, power outages after the replacement, we can conclude that the subset of and abnormal fluctuations in financial data can be of crucial variables in question was the reason why the anomaly had importance for their understanding and taking precautionary occurred. We propose a novel anomaly attribution scheme Our attribution method can be applied to any multivariate to analyze anomalous intervals of multivariate temporal and time series data regardless of potential outliers and missing spatio-temporal data and attribute those anomalies to a set of values.