A Similarity Model Based on Trend for Time Series

This paper presents a time series similarity matching model based on trend meeting the people´s intuitive sense of trends characterize similarity. At the same time, the concept of similarity value is introduced in order to display the similarity of time series in a more intuitive form. In this model, the original time series are segmented according to the time series segmentation algorithm based on significant points. Each sub-section of the time series are mapped to a two-dimensional vector according to the slope and time span, and then symbolic the two-dimensional vector and calculate the distance between two time series of strings. Finally according to similarity calculation formula proposed, obtain the similarity value between the two time series. Experimental results show that the time series similarity matching model is good. In the aspect of similarity matching, the applicability, high efficiency.