AUTOMATIC QTS ANOMALY DETECTION FRAMEWORK BASED ON ATTENTION MECHANISM LSTM-CNN

In order to improve the generality and accuracy of the detection method, an automatic quasi periodic time series anomaly detection framework based on attention mechanism LSTM-CNNis proposed. The purpose of the QTS segmentation algorithm was to automatically and accurately segment QTS into continuous high-quality quasi periods and improve the ability of anti-noise. The purpose of the LSTM-CNN model was to accurately simulate the quasi periodic fluctuation pattern by using the overall trend and local characteristics of the quasi periodic at the same time. Experimental results on four common datasets show that the proposed method can effectively improve the detection versatility and accuracy.