Conformal Anomaly Detection on Spatio-Temporal Observations with Missing Data
arXiv:2105.11886
Abstract
We develop a distribution-free, unsupervised anomaly detection method called ECAD, which wraps around any regression algorithm and sequentially detects anomalies. Rooted in conformal prediction, ECAD does not require data exchangeability but approximately controls the Type-I error when data are normal. Computationally, it involves no data-splitting and efficiently trains ensemble predictors to increase statistical power. We demonstrate the superior performance of ECAD on detecting anomalous spatio-temporal traffic flow.
arXiv commentSubmitted to ICML 2021 Workshop--Distribution-free Uncertainty Quantification
References in corpus (10)
- A tutorial on conformal prediction
- Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery
- Testing for Outliers with Conformal p-values
- Real-Time Illegal Parking Detection System Based on Deep Learning
- Conformalized Quantile Regression
- Uncertainty Sets for Image Classifiers using Conformal Prediction
- No Free Lunch But A Cheaper Supper: A General Framework for Streaming Anomaly Detection
- Conformal k-NN Anomaly Detector for Univariate Data Streams
- Classification with Valid and Adaptive Coverage
- Online control of the false discovery rate with decaying memory