Harnessing Feature Clustering For Enhanced Anomaly Detection With Variational Autoencoder And Dynamic Threshold

TL;DR

This paper presents a novel anomaly detection approach for multivariate climate time series data, combining Variational Autoencoders with dynamic thresholding and feature clustering to improve detection accuracy and explainability of climate anomalies.

Contribution

It introduces a new framework that enhances VAE-based anomaly detection with feature clustering and dynamic thresholds, specifically tailored for climate data analysis.

Findings

Higher F1-score on benchmark datasets

Improved detection of localized anomalies

Enhanced interpretability of climate anomalies

Abstract

We introduce an anomaly detection method for multivariate time series data with the aim of identifying critical periods and features influencing extreme climate events like snowmelt in the Arctic. This method leverages the Variational Autoencoder (VAE) integrated with dynamic thresholding and correlation-based feature clustering. This framework enhances the VAE's ability to identify localized dependencies and learn the temporal relationships in climate data, thereby improving the detection of anomalies as demonstrated by its higher F1-score on benchmark datasets. The study's main contributions include the development of a robust anomaly detection method, improving feature representation within VAEs through clustering, and creating a dynamic threshold algorithm for localized anomaly detection. This method offers explainability of climate anomalies across different regions.