Conformal k-NN Anomaly Detector for Univariate Data Streams

TL;DR

This paper introduces a simple, model-free anomaly detection method for univariate time-series data that adapts to non-stationarity and provides probabilistic scores, matching the performance of complex models.

Contribution

It proposes a novel conformal k-NN based anomaly detection approach that is simple, adaptive, and effective for univariate data streams.

Findings

Performs on par with complex models on benchmark datasets

Adapts to non-stationary data streams

Provides probabilistic abnormality scores

Abstract

Anomalies in time-series data give essential and often actionable information in many applications. In this paper we consider a model-free anomaly detection method for univariate time-series which adapts to non-stationarity in the data stream and provides probabilistic abnormality scores based on the conformal prediction paradigm. Despite its simplicity the method performs on par with complex prediction-based models on the Numenta Anomaly Detection benchmark and the Yahoo! S5 dataset.