Improved Chameloen Clustering Algorithm Based On K-medoids

Big data has become the focus of social development in recent years,because of the rapid development of modern technology.Clustering problem is an important research topic in the field of data mining.Clustering analysis can discover the characteristics of data sets and help data mining algorithm to preprocess data.Therefore,improving the clustering performance of clustering algorithm has become a research hotspot.Hierarchical clustering algorithm is a common way to solve the problem of large data aggregation class.Chameleon algorithm is a relatively common multi-stage hierarchical clustering algorithm with simple structure,which can process large data sets and dynamically build models according to the similarity between clusters.This paper improves on the traditional Chameleon algorithm,and conducts empirical analysis to verify the practicability of the improved algorithm.Firstly,the paper studies the basic principle and algorithm structure of Chameleon clustering algorithm,discovers the problem that the abnormal data in Chameleon algorithm can not be correctly processed,and gives the improved method of Chameleon algorithm.The Knearest neighbor graph in the first stage of Chameleon algorithm is subdivided by K-medoids algorithm,and the abnormal points in the data set are classified correctly.It ensures the high cohesion of the sub-clusters and reduces the impact of outliers.It provides more accurate sub-clusters for the construction of the local dynamic model in the second stage,and then gets more accurate clustering analysis results.Secondly,the paper studies the Chameleon algorithm based on K-medoids algorithm to improve the effect of practical application.The improved Chameleon algorithm is empirically analyzed by using six eigenvalues extracted from the transaction information data of 649 securities customers of a securities company in 2018.The improved Chameleon algorithm is compared with several improved Chameleon algorithms.The practicability and effectiveness of the improved Chameleon algorithm based on K-medoids algorithm are further verified.