Finding Time Series Motifs Based On Cloud Model

The research of finding time series motifs has received much attention recently. In an earlier work, we proposed a relatively comprehensive definition of K-motifs to mine more frequent patterns from time series datasets. However, that work has not given a method to select a better K-motif when we encounter the situation that there are several candidate K-motifs. This paper addresses the problem by introducing a novel method inspired by the cloud model theory. Our method can represent qualitative concepts from the quantitative point of view based on the three numerical characteristics of the cloud model and select a better K-motif effectively and accurately. Finally, in order to demonstrate the feasibility of our method, we conduct several experiments. The results show that our method is feasible and effective.