|Table of Contents|

A Multi-modality-based Random Subspace Classifier Ensemble Algorithm(PDF)

南京师范大学学报（工程技术版）[ISSN:1006-6977/CN:61-1281/TN]

Issue:
2009年04期
Page:
57-62
Research Field:
Publishing date:

Info

Title:
A Multi-modality-based Random Subspace Classifier Ensemble Algorithm
Author(s):
Ye Yunlong; Yang Ming
School of Computer Science and Technology,Nanjing Normal University,Nanjing 210097,China
Keywords:
M ult-im odality;  random subspace;  classifier ensemb le
PACS:
TP301.6
DOI:
-
Abstract:
Tex t C lassifica tion is an im portant m ach ine learn ing research, in w hich som e progress has been made. M ost o f the ex isting class ification me thods are based on Vecto r SpaceM ode l( VSM ), but VSM does not e ffective ly u tilize the structure in fo rm ation h idden in the text sam ples. A t the same tim e, VSM vectors are o ften h igh-d im ensiona,l m ere ly us ing d im ensiona lity reduction stra tegy m ay lead to the lo ss of the use fu l in fo rm ation. To overcom e the shortcom ings o f the ex isting a lgo rithm s, w e propose an algorithm ca lledM ult-i modality-based Random Feature subspace classifier Ensem ble (MMRFSEn) , wh ich can e ffective ly use the structure in fo rm ation h idden in the text such as the w ords’ s average location and standa rd dev ia tion, and m eanw hile each sing le class ifier is constructed by a random ly se lected subspace. The experim ental resu lts show tha t the new ly deve loped m e thod is e ffective and feasib le.

References:

[ 1] Sebastiani F. M ach ine learn ing in au tom ated tex t ca tego rization[ J]. ACM Computing Survey, 2002, 34( 1): 1- 47.
[ 2] W e iss SM, Apte C, Dam erau F J. M ax im izing tex t-m in ing perfo rmance[ J] . IEEE Inte lligent System s, 1999, 14( 4) : 63-69.
[ 3] SchapireR E, S ingerY. Boostex ter: a boosting-based system fo r tex t ca tego rization [ J]. M achine Lea rn ing, 2000, 39( 223):
135-168.
[ 4] Lu Yuchang, LuM ingyu, L i Fan. Ann lys is and construction of w ord we igh ing function in VSM [ J] . Journa l o f Computer Research
& Deve lopm en t, 2002, 39( 10): 1 205-1 210.
[ 5] Babaguchi N, K aw a iY, K itahash i T. Event based index ing o f broadcast spo rts v ideo by interm oda l co llaboration[ J]. IEEE
T rans onMu ltimed ia, 2002, 4( 1): 68-75.
[ 6] Snoek C G M, Wo rringM. M ultim ed ia event-based v ideo index ing using tim e in terva ls[ J]. IEEE Trans onM u ltimedia, 2005,
7( 4): 638- 647.
[ 7] H u N, W ang Y W, Lv N. Study on mu ltimodel retr ieva lm ethod o f content-based v ideo [ J]. Journa l o f Jilin Un iversity: Inform
ation Sc ience Edition, 2006, 24( 3): 265-270. ( in Ch inesew ith Eng lish abstract).
[ 8] 吴飞, 刘亚楠, 庄越挺. 基于张量表示的直推式多模态视频语义概念检测[ J]. 软件学报, 2008, 19( 11): 2 583-2 868.
W u Fe,i Liu Yanan, Zhuang Yue ting. Transductive m ult-im oda lity v ideo concept detec tion w ith Tenso r representa tion[ J].
Jou rnal of Softw are, 2008, 19( 11): 2 583-2 868. ( in Ch inese)
[ 9] Xue X iaob ing, Zhou Zh ihua. Distributional fea tures for tex t categor ization[ C ] / / Pro ceedings o f the 17 th European Conference
onM ach ine Learn ing ( ECML’ 06). Berlin, Ge rmany, LNAI 4212, 2006: 497-508.
[ 10] 孙春红, 杨明. 一种嵌入分布信息的W eb文档相似性度量[ J]. 南京师范大学学报: 工程技术版, 2008, 8( 3): 67-68.
Sun Chunhong, YangM ing. A novel sim ilar ity measurem ent forw eb pages by incorporating d istribution in fo rm ation[ J]. Journa
l of Nanjing No rm alUn iv ers ity: Eng inee ring and Techno logy Ed ition, 2008, 8( 3): 67-68. ( in Ch inese)

Memo

Memo:
-
Last Update: 2013-04-23