ABSTRACT
Detecting anomalies from high-dimensional multivariate temporal data is challenging, because of the non-linear, complex relationships between signals. Recently, deep learning methods based on autoencoders have been shown to capture these relationships and accurately discern between normal and abnormal patterns of behavior, even in fully unsupervised scenarios. However, validating the anomalies detected is difficult without additional explanations. In this paper, we extend SHAP -- a unified framework for providing additive explanations, previously applied for supervised models -- with influence weighting, in order to explain anomalies detected from multivariate time series with a GRU-based autoencoder. Namely, we extract the signals that contribute most to an anomaly and those that counteract it. We evaluate our approach on two use cases and show that we can generate insightful explanations for both single and multiple anomalies.
- A. Del Giorno et al. 2016a. Informative Features for Anomaly Detection. In ICML Anomaly Detection Workshop.Google Scholar
- A. Shrikumar et al. 2017a. Learning Important Features Through Propagating Activation Differences. In ICML.Google Scholar
- C. Zhang et al. 2018a. A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data. In arXiv:1811.08055.Google Scholar
- E. Gutflaish et al. 2019 a. Temporal Anomaly Detection: Calibrating the Surprise. In AAAI.Google Scholar
- M. Sundararajan et al. 2017b. Axiomatic Attribution for Deep Networks. In ICML.Google Scholar
- M. T. Ribeiro et al. 2016b. Why Should I Trust You?: Explaining the Predictions of Any Classifier. In KDD.Google Scholar
- R. Andrzejak et al. 2001. Indications of Nonlinear Deterministic and Finite Dimensional Structures in Time Series of Brain Electrical Activity: Dependence on Recording Region and Brain State. In Phys. Rev. E.Google Scholar
- U. Bhatt et al. 2019 b. Towards Aggregating Weighted Feature Attributions. AAAI Workshop on Network Interpretability for Deep Learning.Google Scholar
- Y. Guo et al. 2018b. Multidimensional Time Series Anomaly Detection: A GRU-based Gaussian Mixture Variational Autoencoder Approach. In ACML.Google Scholar
- Y. Ikeda et al. 2019 c. Estimations for Dimensions Contributing to Detected Anomalies with Variational Autoencoders. In AAAI.Google Scholar
- P. W. Koh and P. Liang. 2017. Understanding Black-box Predictions via Influence Functions. In ICML.Google Scholar
- S. M. Lundberg and S. Lee. 2017. A Unified Approach to Interpreting Model Predictions. In NIPS.Google Scholar
Index Terms
- Additive Explanations for Anomalies Detected from Multivariate Temporal Data
Recommendations
Explaining anomalies detected by autoencoders using Shapley Additive Explanations▪
AbstractDeep learning algorithms for anomaly detection, such as autoencoders, point out the outliers, saving experts the time-consuming task of examining normal cases in order to find anomalies. Most outlier detection algorithms output a score for each ...
Highlights- Explaining anomalies identified by autoencoder using shapley values.
- Explain features with high reconstruction error.
- Evaluated correctness and robustness of explanations.
- Explanations can assist in reducing anomaly score.
- ...
Reliable Counterfactual Explanations for Autoencoder based Anomalies
CODS-COMAD '21: Proceedings of the 3rd ACM India Joint International Conference on Data Science & Management of Data (8th ACM IKDD CODS & 26th COMAD)Autoencoders have been used successfully for tackling the problem of anomaly detection in an unsupervised setting, and are often known to give better results than traditional approaches such as clustering and subspace-based (linear) methods. A data ...
Deep learning for anomaly detection in multivariate time series: Approaches, applications, and challenges
AbstractAnomaly detection has recently been applied to various areas, and several techniques based on deep learning have been proposed for the analysis of multivariate time series. In this study, we classify the anomalies into three types, ...
Highlights- The methods for anomaly detection on multivariate time series are reviewed.
- The ...
Comments