一种基于多模态模型的随机子空间分类集成算法-工程技术版.pdf

9 4 () Vo.l 9No. 4 2009 12 JOURNAL OF NAN JING NORMAL UN IVERSITY ( ENG INEERING AND TE HNOLOGY EDIT ION ) Dec, 2009 一种基于多模态模型的随机子空间分类集成算法 叶云龙, 杨 明 (, 210097) [] , . VSM , VSM . , VSM , . , MMRFSEn, ( ), . , . [] , , [ ] TP 301. 6 [ ] A [] 1672-1292(2009) 04-0057-06 A Mult-im odality-based Random Subspace C lassifier Ensemble A lgorithm Ye Y un long, Y ang M ing ( School of omputer Science and Technology, Nanjing NormalUniversity, Nanjing 210097, hina) Abstract: Text lassification is an mi portantmachine learning research, inw hich some progress has been made. M ost of the ex isting classification methods are based on Vector SpaceM odel( VSM ), butVSM does not effectively utilize the structure information hidden in the text samples. A t the same tmi e, VSM vectors are often high-dmi ensiona,l merely u- sing dmi ensionality reduction strategy may lead to the loss of the useful information. To overcome the shortcom ings of the existing algorithm s, w e propose an algorithm calledM ulti-modality-based Random Feature subspace classifier Ensemble (MMRFSEn) , which can effectively use the structure information hidden in the text such as the w ords s average loca- tion and standard deviation, and meanw hile each single classifier is constructed by a random ly selected subspace. The expermi ental results show that the new ly developedmethod is effective and feasible. K ey words: M ulti-modality, random subspace, classifier ensemble , . , . , . [ 1] 20 90, , . , Naive Bayes., , [ 2] .W eiss , em ail. Scha- [ 3] pire ( decision stump) Boostexter, . . , , , (Vector