Research On Clustering Ensemble With Base Clustering Alignment

In today's digital age,a huge amount of data will be generated every day.In order to find more valuable information,data mining technology has also entered a period of rapid development.Clustering ensemble is an important branch of clustering analysis in data mining.The process of clustering ensemble includes generating base clustering results and final consistent clustering result.The second process needs to use consistency function to aggregate multiple base clustering results obtained in the first process to obtain final consistent clustering results.At present,there are still some difficulties in the design of consensus function in clustering ensemble.In this paper,by reference to domestic and foreign literature in this regard and achieved the following results:(1)A clustering ensemble algorithm based on clustering hard alignment is proposed.Firstly,a base clustering result is taken as the benchmark,and other base clustering results are used as the permutation matrix to transform with it.Then,each transformation result is weighted.Finally,the final consistency representation is obtained by summation.The experimental results show that the algorithm needs to be further improved.Based on this method,a clustering ensemble algorithm based on maximization alignment criteria is proposed.(2)A clustering ensemble algorithm based on maximization alignment criteria is proposed.The final consistent clustering result is found by maximizing the alignment of permutation basis,and all clusters are mapped from the base clustering result to the new permutation cluster.On the one hand,this method finds the optimum weak correspondence between clusters from the base cluster and the final consistent cluster;on the other hand,it obtains the final clustering result by weighted aggregation of aligned base clusters.It is worth mentioning that this method does not use any hyper-parameters and has linear time complexity.(3)A clustering algorithm analysis system based on clustering ensemble is designed and implemented.In view of the problem that many researchers with new contact clustering ensemble analysis and weak code ability are faced with the need to constantly manually adjust the source code the source code and cannot focus on the algorithm itself in the experiment,this system encapsulates some existing clustering ensemble algorithms and the algorithms proposed in this paper into the system,helping new researchers to solve this problem.To sum up,aiming at the problem of consistency function design for clustering ensemble to generate final consistent clustering results,this paper proposes two new methods.For the new researchers who are new to cluster analysis and have not strong code ability,this paper designs and implements a cluster analysis system based on clustering ensemble.