Density Clustering Algorithm Based On Two-Stage Search

In the era of rapid development of data economy,data mining has a wide range of applications in various fields,and is also the focus of academic circles in recent years.Clustering technology is one of the representative technologies in data mining,and it is also the focus of this paper.Density-based clustering is an important data clustering method,which can be applied to data sets of arbitrary shape,but the current density clustering algorithms have some problems,such as randomness,subjectivity and joint error.To overcome these problems,a new kind of density clustering algorithm based on two-stage search in this paper to further improve the accuracy and stability of clustering.The specific content of this paper is as follows:Firstly,this thesis summarizes the theoretical and application status of clustering algorithms at home and abroad.Then,five classical clustering methods are studied,their clustering principles and representative algorithms are introduced in detail,and the problems existing in DPC algorithm and DBSCAN algorithm are mainly studied.Based on this,the definitions and calculating approaches of density threshold and the nearest neighborhood in clusters are given.Then,the two-stage clustering mechanism is built using the strategies of density sorting,distribution according to the nearest neighborhood in clusters and adaptive search,and the two-stage clustering algorithms of recursive search in neighborhood and the nearest neighbor search in clusters are designed,it makes data points with different densities accurate clustering.Secondly,the experiments on synthetic and real datasets show that the proposed method outperforms DBSCAN,AA-DBSCAN,DPC,RDCA,K-Means,K-Medoids and FCM algorithms,and the robustness of the algorithm parameters is analyzed.The experimental results show that the proposed density clustering algorithm is stable,noiseless,and can automatically determine the number of clusters.The clustering accuracy is better than that of the compared clustering algorithms.