Improved k-nearest neighbor classification

被引:131
作者
Wu, YQ [1 ]
Ianakiev, K [1 ]
Govindaraju, V [1 ]
机构
[1] SUNY Buffalo, Ctr Excellence Pattern Anal & Recognit, Buffalo, NY 14228 USA
关键词
k-nearest neighbor classification; pattern classification; classifier; template condensing; preprocessing;
D O I
10.1016/S0031-3203(01)00132-7
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
k-nearest neighbor (k-NN) classification is a well-known decision rule that is widely used in pattern classification. However, the traditional implementation of this method is computationally expensive. In this paper we develop two effective techniques, namely, template condensing and preprocessing, to significantly speed up k-NN classification while maintaining the level of accuracy. Our template condensing technique aims at "sparsifying" dense homogeneous clusters of prototypes of any single class. This is implemented by iteratively eliminating patterns which exhibit high attractive capacities. Our preprocessing technique filters a large portion of prototypes which are unlikely to match against the unknown pattern. This again accelerates the classification procedure considerably, especially in cases where the dimensionality of the feature space is high. One of our case studies shows that the incorporation of these two techniques to k-NN rule achieves a seven-fold speed-up without sacrificing accuracy. CD 2002 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.
引用
收藏
页码:2311 / 2318
页数:8
相关论文
共 14 条
[1]   PATTERN-CLASSIFICATION USING AN EFFICIENT KNNR [J].
BELKASIM, SO ;
SHRIDHAR, M ;
AHMADI, M .
PATTERN RECOGNITION, 1992, 25 (10) :1269-1274
[2]   ACCELERATED TEMPLATE MATCHING USING TEMPLATE TREES GROWN BY CONDENSATION [J].
BROWN, RL .
IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS, 1995, 25 (03) :523-528
[3]   FINDING PROTOTYPES FOR NEAREST NEIGHBOR CLASSIFIERS [J].
CHANG, CL .
IEEE TRANSACTIONS ON COMPUTERS, 1974, C 23 (11) :1179-1184
[4]   NEAREST NEIGHBOR PATTERN CLASSIFICATION [J].
COVER, TM ;
HART, PE .
IEEE TRANSACTIONS ON INFORMATION THEORY, 1967, 13 (01) :21-+
[5]  
DUDANI SA, 1991, DISTANCE WEIGHTED K, P92
[6]   K-NEAREST-NEIGHBOR BAYES-RISK ESTIMATION [J].
FUKUNAGA, K ;
HOSTETLER, LD .
IEEE TRANSACTIONS ON INFORMATION THEORY, 1975, 21 (03) :285-293
[7]  
GATES GW, 1972, IEEE T INFORM THEORY, V18, P431, DOI 10.1109/TIT.1972.1054809
[8]  
Hart P.E., 1973, Pattern recognition and scene analysis
[9]   CONDENSED NEAREST NEIGHBOR RULE [J].
HART, PE .
IEEE TRANSACTIONS ON INFORMATION THEORY, 1968, 14 (03) :515-+
[10]   A new nearest-neighbor rule in the pattern classification problem [J].
Hattori, K ;
Takahashi, M .
PATTERN RECOGNITION, 1999, 32 (03) :425-432