IMPACT: Influence Modeling for Open-Set Time Series Anomaly Detection
Researchers have developed a new framework called IMPACT for open-set time series anomaly detection. This method uses influence modeling to estimate the impact of individual training samples, enabling the generation of realistic unseen anomalies and the repurposing of high-influence samples for anomaly decontamination. Experiments demonstrate that IMPACT significantly outperforms existing state-of-the-art methods across various settings and contamination rates. AI
IMPACT Enhances anomaly detection capabilities for time series data, potentially improving applications in fraud detection and system monitoring.