Research On Clustering Algorithm Based On Evolutionary Thought And Cluster Fusion

In order to obtain useful information from the complex data in the real world,many data mining techniques have been proposed continuously.Cluster analysis is an important part of data mining technology,it has been widely used in pattern recognition,image processing,network security and machine learning and other fields.The accuracy and intelligence of the clustering algorithm play a decisive role in the clustering performance.The subject researches and improves two clustering algorithms,K-means and Density Peaks Clustering(DPC),and the details are as follows:Aiming at the shortcomings of K-means clustering algorithm that the number of clusters is not known in advance and cannot handle non-convex distribution data sets,a K-means clustering algorithm based on evolutionary clustering and cluster fusion is proposed.Aiming at the problem that the K-means clustering algorithm does not know the number of clusters in advance,the algorithm embeds the K-means clustering algorithm into the framework of the evolutionary clustering algorithm,and gradually divides the data by adjusting the distance parameter,and automatically determines the number of clusters in the process.The number of clusters is k;in view of the problem that the K-means clustering algorithm cannot deal with non-convex distribution data sets,a fusion algorithm based on the closest distance of intermediate circle density clusters and a fusion algorithm of intermediate circle density clusters based on representative classes are proposed.The clusters are fused so that the k value gradually tends to the true value.Finally,the proposed clustering algorithm is compared with other clustering algorithms,and the experiment proves the validity of the clustering algorithm proposed in this paper.Aiming at the problem that the cut-off distance and the number of density peak points of the density peak clustering algorithm need to be manually selected and the robustness is poor,and the allocation strategy has associated allocation errors,a density peak clustering algorithm based on cluster fusion is proposed.Aiming at the problem that the truncation distance and the number of density peak points of the density peak clustering algorithm need to be selected manually,the truncation distance is substituted into the Theil coefficient formula,and the optimal parameter truncation distance is obtained by minimizing the Theil coefficient;the number of density peak points needs to be In order to solve the problem of artificial determination and associated errors in data point allocation,a fusion decision point algorithm and a density-based cluster fusion algorithm are proposed.First,the DBSCAN clustering algorithm is used to fuse decision points with high similarity into one class,and the number of remaining noise points The purpose is to preselect the number of density peak points p,then distribute the remaining data points,and finally use the cluster fusion algorithm to replace the sample point allocation strategy,and directly fuse the clusters with high similarity to obtain the final clustering result,reducing The number of data sample allocations and the number of density peak points are adaptively selected.Experimental results on artificial and UCI datasets show that the method has good practicability.