蚁群算法在KNN文本分类中的应用  

The Application of Ant-Colony-Algorithm to the Knn Text Classification

作　　者：殷宏威[1] 赵伟[1,2] 杨志伟[1] 

机构地区：[1]长春工业大学计算机学院,长春130012 [2]吉林农业大学信息技术学院,长春130118

出　　处：《长春理工大学学报（自然科学版）》2010年第1期159-163,共5页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60573182)

摘　　要：作为一种经典的文本分类算法,KNN简单、实用,在许多实际系统中有广泛的应用,但若待分样本位于易判区域时,KNN却做了许多无用计算。基于此,本文提出一种改进算法,借鉴于蚁群算法,引入了组相似度这个新颖概念,使得当待测样本位于易判区域时,能很快得出判定结果;当待测样本位于难判区域时,该算法退化为KNN的原始算法。As a classic text classification algorithm,KNN is simple,practical and widely used in many practical system .however,if the sample which is being tested is in the easy-to-judge region,original KNN made a lot of useless computation.based on it.Referring to ant-colony-algorithm,this paper proposes an improved algorithm. The group-similarity,as a novel conception,is put forward.When the sample being tested is in the easy-to-judge region,the algorithm can quickly complete the procession;even though the sample is in the difficult-to-judge region,the algorithm complete it in the almost same time as original KNN. Experiments also shows that its precision rate and recall rate are not lower than original KNN.

关 键 词：KNN 文本分类 待测样本 蚁群算法 易判区域 难判区域 组相似度 

分 类 号：TP391.12[自动化与计算机技术—计算机应用技术]