A time series anomaly detection method based on improved clustering algorithm
Time series anomaly detection is widely used in the field of civil aviation.Anomaly detection of time series data collected by aircraft quick access recorder provides a powerful means to identify events that reduce safety margin.In order to im-prove the accuracy of time series anomaly detection,an improved clustering algorithm based time series anomaly detection method is proposed.The Euclidean distance measurement method of K-Medoids clustering algorithm is replaced by the dynamic time regu-lar distance measurement method,and the anomaly detection method is determined according to the distance between the sample point and the center point,and the validity of the time series anomaly detection method is studied by aircraft flight parameter over-run detection.Experimental results show that the proposed method has higher anomaly detection accuracy and F1 score than the traditional clustering algorithm.The clustering algorithm optimizes the accuracy of time series similarity measurement by using the dynamic time regularity measurement distance,which can cluster the time series data with similar morphological characteristics better and improve the accuracy of the clustering algorithm.
time seriesflight dataclusterdynamic time warpinganomaly detection
万方数据
维普
