標題: Real-time and automatic two-class clustering by analytical formulas
作者: Lin, JC
Lin, WJ
交大名義發表
資訊工程學系
National Chiao Tung University
Department of Computer Science
關鍵字: two-class clustering;high-dimensional space;feature-preserving;analytical fast clustering;general form k-means;hierarchical agglomerative clustering
公開日期: 1-十一月-1996
摘要: Several feature-preserving two-class clustering methods are investigated in this paper. By preserving certain features of the input data, some formulas useful in calculating the two class representatives and population percentages are derived. The derived formulas are expressed in general forms suitable for any dimensionality higher than two. The complexities of the investigated methods are all of order N if the data size is N and hence are much faster than any other clustering method which uses N x N dissimilarity matrix. Additionally, all investigated methods use no initial guesses. Experimental results are included to make a comparison among the four investigated methods so that only two methods are recommended. Further comparisons with the k-means method and hierarchical clustering methods also are included. The proposed feature-preserving approach was found to be fast, automatic and suitable for any held requiring fast high-dimensional two-class clustering. Copyright (C) 1996 Pattern Recognition Society.
URI: http://dx.doi.org/10.1016/0031-3203(96)00033-7
http://hdl.handle.net/11536/967
ISSN: 0031-3203
DOI: 10.1016/0031-3203(96)00033-7
期刊: PATTERN RECOGNITION
Volume: 29
Issue: 11
起始頁: 1919
結束頁: 1930
顯示於類別:期刊論文


文件中的檔案:

  1. A1996VT97300012.pdf