EADetection: An efficient and accurate sequential behavior anomaly detection approach over data streams

Due to the increasing arriving rate and complex relationship of behavior data streams, how to detect sequential behavior anomaly in an efficient and accurate manner has become an emerging challenge. However, most of the existing literature simply calculates the anomaly score for segmented sequence, and there is limited work going deep to investigate data stream segment and structural relationship. Moreover, existing studies cannot meet...