A Clustering Algorithm Based On Density Partition

With the development of information science,the research of data mining technology has received more and more attention from scholars.As an important branch of data mining,the research of cluster analysis has made great progress in recent years.This paper mainly proposes a clustering algorithm based on density partition(CABODP).For the density-based clustering algorithm,the density parameter is global and it does not consider the relationship between adjacent clusters.The CABODP algorithm has the following improvements: First,the CABODP algorithm makes use of its own characteristics of the data,the density parameters of each density partition are automatically generated,the steps of the user inputting the density parameter are omitted,and the influence of the global density parameter on the clustering quality is eliminated.Secondly,membership degree coefficient is introduced in the algorithm.This parameter sets a judgment rule for the separation and combination between clusters and clusters in the dataset.In the numerical experiment part,seven sets of two-dimensional data and three sets of high-dimensional data are used to test the clustering accuracy,parameter sensitivity,and clustering time performance of the algorithm.Compared with the DBSCAN algorithm and K-means algorithm,the experimental results show that the CABODP algorithm has good clustering effect for multi-density,adjacent,complex-shaped data,it can identify the noise points and isolated points,and it is suitable for clustering high-dimensional data.