期刊名称：IAENG International Journal of Computer Science
印刷版ISSN：1819-656X
电子版ISSN：1819-9224
出版年度：2019
卷号：46
期号：4
页码：662-669
出版社：IAENG - International Association of Engineers
摘要：The k nearest neighbor (kNN) rule is known as itssimplicity, effectiveness, intuitiveness and competitive classificationperformance. Selecting the parameter k with the highestclassification accuracy is crucial for kNN. There’s no doubt thatthe leave-one-out cross validation (LOO-CV) is the best methodto do this work as its almost unbiased property. However, itis too time consuming to be used in practice especially forlarge data. In this paper, we propose a new algorithm forselecting an optimal neighborhood size k. We found that theclassification accuracy of LOO-CV is approximate concave forthe parameter k. And a search method is proposed to pickout the optimal value of k. An empirical study conducted on8 standard databases from the UCI repository shows that thenew strategy can find the optimal k with significantly less timethan the LOO-CV method.
关键词：k nearest neighbor; leave-one-out cross validation;selecting the parameter k.
