A new clustering algorithm based on KNN and DENCLUE
Xiao-Gao Yu, Yin Jian · 2005
Clustering in data mining is used for identifying useful patterns and interested distributions in the underlying data. Clustering techniques have been studied extensively in e-commerce, statistics, pattern recognition, and machine learning. This increases the need for efficient and effective analysis methods to make use of this information. Traditional DENCLUE is an important clustering algorithm. But it is difficult to make its two global parameters (/spl sigma/, /spl xi/) be globally effective. A new algorithm based on KNN and DENCLUE is proposed in this paper, which offers DENCLUE the appropriate and globally effective parameters based on KNN and DENCLUE. At the first, the window-width (WW) of each data point is determined and the whole data set is partitioned into some fuzzy cluster (FC) by KNN based on KDE. Then, the local /spl sigma/ of each FC is unsupervised determined according to the entropy theory. At the last, each local /spl sigma/ is mapped to the global /spl sigma/ and each FC is independently clustered, which makes the global /spl sigma/ and /spl xi/ have the global validity. The analysis and experiment prove that our clustering method achieves better performance on the quality of the resulting clustering and the results are not sensitive to the parameter k.