Novel clustering algorithm based on central symmetry

Autor: Qi-Lun Zheng, Jia-Meng Xie, Jia-Yi Lin, Hong Peng
Rok vydání: 2005
Předmět:
Zdroj: Proceedings of 2004 International Conference on Machine Learning and Cybernetics (IEEE Cat. No.04EX826).
DOI: 10.1109/icmlc.2004.1381979
Popis: Cluster analysis is an important research field in data mining. One key of the clustering algorithms is the distance measure. A novel distance measure based on central symmetry is proposed in this paper. This kind of distance measure can be used to detect symmetrical patterns in data set. Then a modified version of K-means algorithm employing the central symmetry distance is presented. The proposed algorithm can be used for data clustering in data mining. It divides a given data set into several clusters of different geometrical structures. While detecting hyperspherical-shaped patterns, the clustering algorithm with the central symmetry distance measure performs much better than the preview algorithms with the ordinary measures. The novel clustering algorithm can also be used for human face detection. Finally, some experimental studies and results demonstrate the feasibility and effectiveness of the proposed algorithm.
Databáze: OpenAIRE