The Research And Application Of The Methods To Determine The Clustering Radius

Clustering analysis is a method of unsupervised learning, and it is an important research in the field of data mining, pattern recognition and machine learning. Until now, many clustering algorithms have been proposed by many researchers. And these algorithms are widely used in many different areas, such as data mining, pattern recognition, data analysis, image processing, spatial database technology, biology, and market research, etc.Although cluster analysis has made great achievements in its process of development, but there are still many issues needing people to solve. It is difficult to determine the value of radius and its density threshold in the density based clustering algorithm. Density based clustering algorithm can easily find out any different shape in clusters. However, as same as many clustering algorithms, some values of parameters need to be determined manually. Clustering radius is an important input parameter in the density based clustering algorithm, which affects the quality of clustering. Determing the values of clustering radius and density threshold will increase the burden of users in a certain extent, especially in the high-dimensional data, it is difficult for users to determine the values of clustering radius and density threshold. Therefore, this thesis study how to determine the values of cluster radius and density threshold in the density based clustering algorithm. The main contributions of this dissertation are as follows:Firstly, the thesis proposed a method to automatically determine the value of clustering parameters. It can obtain the suitable value of clustering radius by analyzing the feature of distance between any two points in a data set. Then combine the method with CURD algorithm to achieve a self-adaptive CURD algorithm. Experimental results showed that SA- CURD algorithm can automatically and effectively select the cluster radius and density threshold, and the clustering result had high accuracy by using the two parameters in CURD algorithm.Next, SA-CURD algorithm is used to analysis the three-dimensional clustering phenomenon of housekeeping genes in the Homo sapiens, which is shown with images. This thesis has provided a visualization research tool for three-dimensional genome.