Application of Time Series Anomaly Detection Algorithms in Risk Identification

Traditional risk identification methods often rely on manual inspections and post-event processing, which struggle to cope with the vast amounts of data and the real-time changes in the internet environment. Time series anomaly detection algorithms, as an effective data analysis tool, can identify abnormal changes in time series data. By utilizing spectral residual anomaly detection algorithms and sub-sequence anomaly detection algorithms, models were constructed to conduct an in-depth anomaly detection analysis of price data on the JD platform, providing a new solution for risk identification. The experimental results show that the precision, recall, and F1 scores of these two time series anomaly detection algorithms are significantly better than those of other comparative algorithms, demonstrating significant application value in risk identification.