An Effective Similarity Measure Algorithm for Time Series Based on Key Points

Time series are ubiquitous in data mining, and how to measure their similarity is a core part in the mining system. At present, there are many researches focused on the problem of time series similarity measurement. However, it will lead to a serious performance degradation if the similarity of original series is measured directly, including high dimensionality and a large computation cost. In this paper, a new algorithm based on the key points is presented. Firstly, we reduce the dimension of the time series based on the key points and retain the basic attributes of the series. On the basis, the length of the key point series are equalized. Finally, we apply the improved morphological similarity distance to measure the similarity between two series. The experimental results show that the algorithm proposed in this paper extends the space of similarity measure for nonlinear series with unequal length and reduces the amount of calculation effectively.