An Optimized Chameleon Algorithm based on Local Features
Xiaoxiao Cao, Tianyun Su, Pengyu Wang, Guoyu Wang, Zhihan Lv, Xinfang Li · 2018
Clustering algorithm plays important roles in many fields, such as data mining, data visualization, and so on. In a variety of clustering algorithms, the CHAMELEON algorithm has become a commonly used algorithm because of its ability to discover clusters with arbitrary shapes. This paper makes an intensive study of clustering algorithm, especially CHAMELEON algorithm. After the study of the structure and features of the proposed algorithm, this paper introduces an optimized CHAMELEON algorithm based on local features and grid structure aiming at the deficiency of CHAMELEON. This algorithm generates neighbor graph adaptively, partitions the graph and merges sub-clusters based on local features, so it can produce high quality clustering results. The proposed clustering algorithm is experimented with four data sets, and its performance is compared with CHAMELEON algorithm, DBSCAN algorithm and K-means algorithm. The experimental results show that the clustering algorithm proposed can obtain satisfied clustering effect of data sets with complex distribution.