Entropy-based dynamic graph embedding for anomaly detection on multiple climate time series

Abstract Abnormal climate event is that some meteorological conditions are extreme in a certain time interval. The existing methods for detecting abnormal events utilize supervised learning models to learn the patterns, but they cannot detect untrained patterns. To overcome this problem, we construct dynamic graph by discovering correlation among series and propose novel embedding model based on entropy called EDynGE discriminate anomalies. measurement quantifies information of graphs constructs space. We conducted experiments synthetic datasets real-world datasets. results showed EdynGE achieved better F1-score than baselines 43.2%, number days has increased 304.5 past 30 years.