Research Of Time Series Clustering Algorithm Based On Three-way Decisions Theory

Three-way decision is an important theory to solve the problem of uncertainty.It is the one of the important research directions of the three-way decision to apply the three-way decision to the uncertain problem solving in the field of machine learning.Time series clustering is a hot topic in the field of machine learning.Its main research directions include time series similarity distance computation and clustering optimization based on time series distance.In this paper,we combine three-way decision theories to optimize the different levels of uncertainty in time series clustering,in order to improve the clustering effect of time series.The main work of this paper is as follows.First,the time efficiency of the time series clustering is affected by the problem of high energy and low efficiency in the classical DTW(Dynamic Time Warping)algorithm.In this paper,three-way decision are introduced and a time series similarity algorithm based on three-way decision of warping distance is proposed.The main idea is to set up the DTW three-way decision model based on the three-way decision theory.Based on the error recognition rate optimization method,we give a solution for the decision threshold in the model,and give a concrete simulated annealing algorithm.Finally,through comparative experimental analysis,it is verified that the algorithm propose in this session is more effective than the two decision FTW(Fast Similarity Search under the Time Warping)algorithm,which further proves that the extended research of three-way decision applied to the two decisions is universal and effective.Second,ensemble clustering is a new clustering technique to solve the shortage of single clustering.The quality of the clustering results can be improved effectively.However,the uncertainty of the basic clustering in the process of integration affects the accuracy of the final results.In this study,on the basis of the classical ensemble clustering method,a new ensemble clustering method based on three-way decision theory is proposed by introducing three-way decision techniques.The main idea is to establish three-way decision model based on three-way decision decision and build up the local weighted common joint matrix of three-way decision.The final ensemble clustering results are obtained by the idea of cohesive hierarchical clustering.Finally,through the experiment on public data set,the advantage of 3WD-LWHA algorithm in clustering is verified.Third,based on the above research results,a new time series ensemble clustering method based on three-way decision time series measurement and three-way decision ensemble clustering method is proposed in this paper.Finally,using the public data set design,compared with other algorithms from three angles,the experimental results show that the proposed algorithm can effectively reduce the influence of low quality clustering on the final clustering results in the integration,several data sets have shown great advantages.