Multiscale Time Series Clustering : A Consensus Clustering Approach(Artificial Intelligence III)

概要

論文の詳細を見る
Multiscale analysis has played an important role in signal processing and statistics. Multiscale analysis analyzes the time series data in multiple scales, and each scale catches specific features of the data, thus it gives us the ability of observing time series data in various views. When applying multiscale analysis to time series clustering, it is not trivial to determine the appropriate scale. This problem is circumvented in this paper by getting a consensus of the clustering solutions over each scale. We first perform clustering with the data of each scale, then merge the cluster solutions together. Two contributions of this paper are: we propose a multiscale time series clustering scheme based on the consensus of multiple clustering and a novel consensus clustering algorithm is presented in this paper. The benefits of the proposed approach are experimental evaluated with several real time series data.
一般社団法人情報処理学会の論文
2004-12-04