Spider algorithm for clustering time series
Shohei Kameda, M. Yamamura · International Conference on Artificial Intelligence · 2006
In proportion to the rapid development of information technology, time series are today accumulated in finance, medicine, industry and so forth. Therefore, an analysis of them is an urgent need for these applications. As solving these problems clustering time series has much been paid attention. The similarity for the clustering is commonly measured with Euclidean distance and dynamic time warping. In this paper we propose an innovative and novel algorithm for clustering multivariate time series. The algorithm is called Spider Algorithm. We experimentally show that the similarity from spider algorithm is superior to Euclidean distance or warping path on dynamic time warping, especially when many clusters exist.