Research on similarity comparison method of engineering time series data based on improved dynamic pattern matching algorithm

Traditional time series similarity comparison method is difficult to calculate the similarity of curve shape quantitatively. A new method is proposed in this paper in order to solve this problem. This method consists of two steps: extracting the pattern of time series and measure the similarity of the pattern. Experiments are carried out based on the workflow data of a certain device. The proposed method and traditional method are used to find the time series which has the largest deviation from others. In these experiments, the proposed method can achieve 91.7% accuracy while the accuracy of traditional method is 50% or less. This result shows that the proposed method has better discrimination ability on the shape of the time series curve.