RESEARCH ON AN ALGORITHM FOR TIME SERIES PATTERNS DISCOVERY

In this paper, a new way for logic expression of time series patterns is presented for time\|series data. By a piecewise linear approximation, the data is modeled as sequence of straight lines and the sequence of arctg. （slope） is used as logic expression of time series patterns. In this way, a new algorithm for time series patterns discovery is presented. It can find all of the patterns or sub patterns automatically. The experiment results prove that the algorithm can be put into practice.