Research On Blocking Fuzzy Clustering Algorithm Based On Density Of Samples

With the rapid development of modern civilization,mankind has entered the information society.Whether it is life or work,the amount of information from all walks of life showed explosive growth,the data quantity is more and more big.Then how to find useful information from these massive data has become a hot topic.Data mining technology emerges as the times require.As the saying goes: "Like attracts like,Birds of a feather flock together".Clustering analysis is an important means of data mining.And clustering algorithm is a very active research topic in the field of data mining.The purpose of clustering is to divide a large number of samples or abstract data into several subsets according to the similarity between them.Discover the structure of the data to help people understanding some of the information hidden between the data.With the rapid development of modern information technology,the scale of information is growing faster and faster.Many traditional clustering algorithms have been unable to cope with the increasing size of the data.So the researchers turned their attention to some new algorithms that can adapt to the increasing size of data sets.In this research,combined with the traditional fuzzy clustering algorithm,introduces the theory of clustering based on the idea of block,and combines the density property of the data points in the data sets.Several blocking fuzzy clustering algorithms based on density of samples are proposed,which improve the clustering effect of traditional fuzzy clustering algorithm.The main work is as follows:(1)A weighted online fuzzy clustering algorithm based on density(DWOFCM)is proposed.The algorithm divides the data set by using the idea of block,meanwhile,the density weighted method is introduced.According to the density characteristics of samples,all samples of the data set are weighted.Each part of the data were clustered and then the results were unified for processing.Finally,the clustering results are obtained.Compared with the traditional fuzzy clustering algorithm,the proposed algorithm improves the clustering effect to some extent.But the algorithm also has shortcomings and deficiencies.When the size of data continues to increase,the clustering effect of the algorithm will be decreased significantly.(2)A single-pass fuzzy clustering algorithm based on local density of samples(LDSPFCM) is proposed.Change the global density to local density,improved calculation of density,so as to better reflect the actual distribution of samples.To avoid the interference caused by the sample point with a far distance,and the dependence of SPFCM on input sequence is improved.Then the clustering effect has been improved.(3)A weighted single-pass fuzzy clustering algorithm based on density peaks(DPWSPFCM)is proposed.In addition to calculating the local density of the sample points,the algorithm combines the distance between the high density points in the data set.Finding the density peak of data sets,sorting the data set according to the density peak of the data set,and weighted sample points.The result shows that the weight of the sample points in the density peak region has a greater impact on the clustering results.Then the clustering effect has been improved.