Research On Improvement Of Density Peak Clustering Algorithm And Its Application

With the advent of the era of big data,massive amounts of data urge people to efficiently mine data and make effective use of data.Clustering algorithm is a commonly used data mining tool.In the absence of prior knowledge,it explores the internal structure information and similar relationships of data,and effectively processes massive data by dividing multiple objects into different clusters.Therefore,it is also widely used in various fields,such as:information extraction,pattern recognition,image analysis,data compression,and network security.The Density Peak Clustering Algorithm(DPC)is a popular clustering algorithm in recent years.It has the advantages of simple thinking,identification of clusters of different shapes,and good clustering effect,which has attracted the attention of many researchers.Although the density peak clustering algorithm has many advantages,it also has the problem that the clustering result is more sensitive to the value of the cutoff distance dc,and the value of dc and the cluster center point need to be manually selected.Therefore,this paper analyzes the original density peak clustering algorithm and improves it on the basis of predecessors.This paper mainly includes the following aspects:(1)This thesis analyzes the idea and implementation steps of the original density peak clustering algorithm in detail,and summarizes the shortcomings of the original algorithm.Aiming at the problem that the original algorithm requires manual setting of parameters and manual principle of clustering centers,a density peak algorithm(AKNN-DPC)based on K-nearest neighbors for adaptively determining parameters is proposed.The improved algorithm introduces the idea of K-nearest neighbors,which makes the calculation of local density more reasonable,and normalizes the value of y in the decision graph,and selects those K values that make the value of y have obvious mutations to participate in the clustering,(2)This thesis compares the improved AKNN-DPC with the original DPC algorithm and classic clustering algorithms such as K-Means algorithm,DBSCAN algorithm,Mean-Shift algorithm,etc.,and conducts experiments on multiple standard data sets.The comparative analysis of the results and the evaluation of the clustering evaluation indicators show that the clustering effect of AKNN-DPC is better than other algorithms.(3)AKNN-DPC algorithm is a good way to calculate the local density of data points and clustering,so we apply the algorithm to Hubei province non-material cultural heritage of time and space analysis,the calculation of Hubei province intangible point density and clustering,and analysis of Hubei province intangible on the pattern of space distribution and time evolution process,the emergence and development of Hubei province intangible a detailed characterization,for the future protection and inheritance can provide a powerful basis,also reflected the algorithm has a certain practical value.