Font Size: a A A

Research On Incremental Fuzzy Clustering Algorithm Of Time Series Data

Posted on:2019-09-16Degree:MasterType:Thesis
Country:ChinaCandidate:S WuFull Text:PDF
GTID:2370330599956391Subject:Software engineering
Abstract/Summary:PDF Full Text Request
The time series data which is widely distributed and contains high value information has significant research value.At present,there are many clustering algorithms for time series data,but due to the drift characteristics of time series data,are it is difficult for traditional algorithms to efficiently and accurately cluster.At the same time,with the increasing of data scale,the traditional algorithms cannot cluster and analysis large-scale time series data.So,this paper has researched a large number of relevant literatures,and focuses on fuzzy clustering,dynamic time warping and incremental strategy.Main tasks are as follows:(1)Because the time series data has the time drift characteristic,it is difficult for the clustering algorithm which uses the common similarity measure method to accurately cluster.The existing clustering algorithms based on the dynamic time warping distance cannot efficiently and accurately cluster the time series data.In order to cluster the time series data efficiently and accurately,the third chapter merge fast dynamic time warping and fuzzy clustering,fuzzy c means and fuzzy c Medoid,into a new algorithm(2)The increasing scale of time series data cannot be saved into memory of an ordinary host;high clustering efficiency is needed in applications.Traditional clustering algorithms for time series data have trouble in meeting these requirements.In order to deal with large scale time series data,incremental strategies are introduced into the fuzzy clustering algorithm and four fuzzy clustering algorithms based on incremental strategies(Single-Pass and Online)are proposed in this paper.These algorithms can effectively deal with large scale data sets and ensure high efficiency and accuracy and provide theoretical basis for processing large scale time series data.In order to verify the feasibility of new algorithms,we select several public datasets to do clustering experiments.The experimental results show that new algorithms solve the large-scale data problem to a certain extent,and improve the efficiency and accuracy.
Keywords/Search Tags:large scale time series data, dynamic time warping, Fuzzy clustering
PDF Full Text Request
Related items