周期性时间序列数据聚类算法的改进研究
Research on Improvement of Clustering Algorithm for Periodic Time-series Data
-
摘要: 聚类分析是时间数据序列分析的一种常用手段, 现有的聚类算法通常从相似性度量方面进行改进.实际的时间序列数据往往具有一定的周期性和连续性, 现有的算法往往忽略时间序列数据周期性和连续性特点对聚类算法的影响.对此问题进行了研究, 尝试采用延拓的方法来解决该问题, 从而改善聚类的效果.初步的实验结果表明了该方法的可行性和有效性.Abstract: Clustering analysis is a common means of time-series data analysis.The existing clustering algorithm for the time-series data is improved from the similarity measure aspects, but the actual time-series data tend to have a certain periodicity and continuity.The existing algorithms often ignore the impact of periodicity and continuity on time-series data clustering algorithms.This issue is researched in this paper and extension method is used to solve this problem.Preliminary results show the feasibility and effectiveness.