Current Proceedings on Technology
Yazarlar: Damodar Reddy Edla, Prasanta K. Jana
Konular:-
Anahtar Kelimeler:Clustering,K-means,Kd-tree,Outliers,Biological data,Intra-inter ratio validity index
Özet: K-means is one of the popular partitional clustering techniques that has been researched over the decades. But it suffers from the random selection of initial cluster centres. If an outlier is assigned as an initial seed then, K-means usually merges two or more clusters into a single one which is undesirable. In this paper, we propose a new scheme that avoids the outliers during initialization using kd-tree. We have performed extensive experiments of the proposed algorithm on various artificial and biological data. The results show that the proposed method produces encouraging results over the existing algorithms.