基于互信息规则剪枝的关联文本分类-工程技术版.pdf

8 4 () Vo.l 8 No. 4 2008 12 JOURNAL OF NAN JING NORMAL UN IVERSITY ( ENG INEERING AND TECHNOLOGY EDIT ION ) Dec, 2008 基于互信息规则剪枝的关联文本分类 商炳章, 白清源 (, 350002) [ ] , , . , , . , ARC- BC , . [ ] , , [ ] TP 18 [] A [] 2008) On Classification of Associative Text Based on Rules Pruning ofM utual Inform ation Shang Bingzhang, BaiQ ing uan ( C ollege ofM ath em atics and C om puter Science, Fuzh ou U n iversit, Fuzhou 350002, Ch ina) Abstract: The traditional associative classif ing algorithm s of associative texts generate a huge mum ber of ru les. If the ru les w ere not pruned, the efficienc of classification would be influenced. H ow ever, if the form er prun ingm ethod were adopted, d ifferent degrees of accurac of classification w ould appear. Therefore, an associative text classification algo rithmbased on ru les prun ing ofmutual inform ation is presen ted to prune the ru les of each class. The ru les w ith h igh clas s if ing capacit are chosen to form classifiers to classif the texts being classified. The stud illum inates that the mutual inform ationbased rules pruning algorithm not on l gets much less rules but ism ore helpfu l for mi proving the accurac of the association categorization. The expermi ental resu lts show the performance of th is m ethod is better than both ARC- BC algorithm and the algorithm wh ich uses all rules. K ey words: mutual in form ation, rules pruning, associative classification 1998, Liu, H suM a CBA[ 1] . CBA , C45. , CBA , CMAR[ 2] ARC - BC[ 3] . , . , , . , , [ 2, 3] . , K , , . 1 基本定义 : , (