Precision and Recall for Time Series
This work addresses the need for better evaluation metrics in anomaly detection for time series data, particularly for range-based anomalies, but it appears incremental as it builds on existing Precision and Recall concepts.
The paper tackles the problem of evaluating time series classification algorithms for range-based anomalies, which occur over periods of time, by introducing a new mathematical model that extends Precision and Recall metrics to measure ranges and allows customization for domain-specific preferences.
Classical anomaly detection is principally concerned with point-based anomalies, those anomalies that occur at a single point in time. Yet, many real-world anomalies are range-based, meaning they occur over a period of time. Motivated by this observation, we present a new mathematical model to evaluate the accuracy of time series classification algorithms. Our model expands the well-known Precision and Recall metrics to measure ranges, while simultaneously enabling customization support for domain-specific preferences.