文本分类中特征权重算法的改进-南京师范大学学报.PDFVIP

8 4 () Vo.l 8No. 4 2008 12 JOURNAL OF NAN JING NORMAL UN IVERSITY ( ENG INEERING AND TECHNOLOGY EDIT ION ) Dec, 2008 文本分类中特征权重算法的改进 沈志斌, 白清源 ( , 350002) [ ] TFIDF. , , . , BORTF IDF, , , . TFIDF, . [ ] , , TF IDF, , BORTFIDF [ ] TP 18 [] A [] 2008) Improvem ent of FeatureW eighting A lgorithm in Text Classification Shen Zh i in, BaiQ ingyuan ( College ofM athem atics and Com puter Science, FuzhouU niversity, Fuzhou 350002, China) Abstract: TFIDF is a k ind of comm on m ethods used to m easure the term s in a docum en t. The method is easy ut ig nores the distri ution of the feature in each class. So, it can not really reflect each feature s contri ution to each class. A mi ing at th is shortage, w e put forward the BORTFIDF and use it to readjust each feature s d ifferentiation to each class, i. e. , mod ifies each feature sw eight. Then the classifier is used to check its validaty. The m ethod is etter than trad itional TFIDF and proves that the BORTFIDF m ethod is feasi le. K ey words: text classification, feature weight, TFIDF, class difference, BORTFIDF [ 1] : , . , . 1, 1 . , , [ 2] [ 3] . , Bayes , [ 4] [ 5] [ 6] [ 7 ] [ 8] KNN , Centroid , LLSF , Boosting SVM . TFIDF: . . , TFIDF. 1 传统的TFIDF及相关工作 TFIDF Gerald SaltonM cG ill . , (Term), , , , . TFIDF : TF IDF. TF(Term Frequency), ; IDF( Inverse Document Frequency), 1 1 . IDF= log(N / n) . , N , n t. : : ( SYSKF0701)( 2005XQ13) ( JB06023) . : ,, : . Em ail: a iqy@ fzu. edu. cn