·156· 智能系统学报 第4卷 [32 ]DOM NGOS P METACOST:a generalmethod for making [42 ]SCHOLKOPF B,PLATT J C,TA YLOR J S,et al Esti- classifiers cost sensitive[C]//Proceedings of the 5th Inter mating the support of a high-dmensonal distribution [J] national Conference on Knowledge Discovery and Data Neural Computation,2001,13(7):1443-1472 Mining San Dieg,CA:ACM Press,1999:155-164 [43 ]BRADLEY A.The use of the area under the ROC curve in [33]CHE H G,BONGER R E,LM CC Dual-nu-support the evaluation ofmachine leaming algorithms[J].Pattem vector machine with eror rate and training size biasing Recognition,1997,30(7):1145-1159 [C]//Proceedings of the 25th IEEE Intemational Confer- [44 ]JOSHIM V.On evaluating perfomance of classifiers r ence on Acoustics,Speech and Signal Processing Salt rare classes [C]//Proceedings of the 2nd IEEE Intema- Lake City,USA:IEEE Press,2001:1269-1272 tional Conference on Data Mining Japan,2002:641-644 [34]FAN W,STOLFO J S,ZHANG J X.et al AdaCost mis- [45 ]PARK KJ,KANEH SA M Prediction of potein subcellu- classification cost-sensitive boosting [C]//Proceedings of lar locations by support vecpormachines using compositions the 16th Itemational Conference on Machine Leaming of am ino acids and am ino acid pairs[J ]Bioinfomatics, San Mateo,USA,1999:97-105. 2003,19(13):1656-1663 (35 ]JOSHIM V,AGARWAL R C,KUMAR V.Predicting [46 ]MALOOF M A.Leaming when data sets are mbalanced rare classes can boosting make any weak leamer strong and when costs are unequal and unknown C]//Intema- [C]//Proceedings of the 8th ACM SIGKDD Intemational tonal Conference on Machine Leaming Washington DC. Conference on Knowledge Discovery and DataM ining Ed- 2003:154-160 monton,Canada:ACM Press,2002:297-306 作者简介： [36 ]CHAWLA N V.C4.5 and mbalanced data sets:investiga- 叶志飞，男，1983年生，硕士，主要 ting the effect of sampling method,probabilistic estmate, 研究方向为统计机器学习和模式分类. and decision tree structure C]//Intemational Conference on Machine Leaming Washingbn DC,2003:125-130. [37 ]ELKAN C The foundation of cost-sensitive leaming[C]// Proceedings of the 17th Intemational Joint Conference on Artificial Intelligence Seattle,Washington,2001:239- 246 文益民，男，1969年生，博士后，副 [38 ]CARD IE C,HOW EN.mprovingm inority class predicting 教授，CCF高级会员，主要研究方向为 using case-specific feature weights[C]//Proceed ings of the 统计学习理论、生物信息学和图像处 14 th Intemational Conference on Machine Leaming San 理.发表学术论文20余篇」 Francisco:Morgan Kaufnann,1997:57-65. [39 ]ZHENG Z H,SR HAR I R Optmally combining positive and negative features for text categorization[C]//Itema- tional Conference on Machine Leaming Washington DC, 吕宝粮，男，1960年生，教授、博士 2003:241-245. 生导师、博士、EEE高级会员，主要研 [40 ]WU G,CHANG E Y.KBA:kemel boundary aligrment 究方向为仿脑计算理论与模型、神经网 considering mbalanced data distribution [J].IEEE Trans 络理论与应用、机器学习、模式识别 on Knowledge and Data Engineering,2005,17(6):786- 脑计算机接口、生物信息学与计算生 795 物学.己在EEE Trans Neural Net [41]HONG X,CHEN S,HARR IS C J.A kemel-based to- works,IEEE Trans Bmedical Engineering,Neural Netorks class classifier for ibalanced data sets[J].IEEE Trans- CCV等国际期刊和会议上发表学术论文80余篇. action on Neural Netorks,2007,18(1):28-41. 1994-2009 China Academic Journal Electronic Publishing House.All rights reserved.hutp://ww.cnki.net© 1994-2009 China Academic Journal Electronic Publishing House. All rights reserved. http://www.cnki.net [ 32 ]DOM INGOS P. METACOST: a generalmethod for making classifiers cost sensitive[C ] / /Proceedings of the 5 th Inter2 national Conference on Knowledge D iscovery and Data M ining. San D iego, CA: ACM Press, 1999: 1552164. [ 33 ] CHE H G, BONGER R E, L IM C C. Dual2nu2support vector machine with error rate and training size biasing [C ] / /Proceedings of the 25 th IEEE International Confer2 ence on Acoustics, Speech and Signal Processing. Salt Lake City, USA: IEEE Press, 2001: 126921272. [ 34 ] FAN W , STOLFO J S, ZHANG J X, et al. AdaCost: m is2 classification cost2sensitive boosting [ C ] / /Proceedings of the 16 th International Conference on Machine Learning. San Mateo, USA, 1999: 972105. [ 35 ] JOSH I M V, AGARWAL R C, KUMAR V. Predicting rare classes: can boosting make any weak learner strong [C ] / /Proceedings of the 8 th ACM SIGKDD International Conference on Knowledge D iscovery and DataM ining. Ed2 monton, Canada: ACM Press, 2002: 2972306. [ 36 ]CHAWLA N V. C4. 5 and imbalanced data sets: investiga2 ting the effect of samp ling method, p robabilistic estimate, and decision tree structure [ C ] / / International Conference on Machine Learning. W ashington DC, 2003: 1252130. [ 37 ]ELKAN C. The foundation of cost2sensitive learning[C ] / / Proceedings of the 17 th International Joint Conference on A rtificial Intelligence. Seattle, W ashington, 2001: 2392 246. [ 38 ]CARD IE C, HOW E N. Imp rovingminority class p redicting using case2specific feature weights[C ] / /Proceedings of the 14 th International Conference on Machine Learning. San Francisco: Morgan Kaufmann, 1997: 57265. [ 39 ] ZHENG Z H, SR IHAR I R. Op timally combining positive and negative features for text categorization [ C ] / / Interna2 tional Conference on Machine Learning. W ashington DC, 2003: 2412245. [ 40 ]WU G, CHANG E Y. KBA: kernel boundary alignment considering imbalanced data distribution [J ]. IEEE Trans on Knowledge and Data Engineering, 2005, 17 ( 6) : 7862 795. [ 41 ]HONG X, CHEN S, HARR IS C J. A kernel2based two2 class classifier for imbalanced data sets[J ]. IEEE Trans2 action on Neural Networks, 2007, 18 (1) : 28241. [ 42 ] SCHO¨LKOPF B, PLATT J C, TAYLOR J S, et al. Esti2 mating the support of a high2dimensional distribution [J ]. Neural Computation, 2001, 13 (7) : 144321472. [ 43 ]BRADLEY A. The use of the area under the ROC curve in the evaluation of machine learning algorithm s[J ]. Pattern Recognition, 1997, 30 (7) : 114521159. [ 44 ]JOSH IM V. On evaluating performance of classifiers for rare classes[ C ] / /Proceedings of the 2nd IEEE Interna2 tional Conference on Data M ining. Japan, 2002: 6412644. [ 45 ] PARK K J, KANEH ISA M. Prediction of p rotein subcellu2 lar locations by support vectormachines using compositions of amino acids and amino acid pairs[J ]. Bioinformatics, 2003, 19 (13) : 165621663. [ 46 ]MALOOF M A. Learning when data sets are imbalanced and when costs are unequal and unknown [ C ] / / Interna2 tional Conference on Machine Learning. W ashington DC, 2003: 1542160. 作者简介 : 叶志飞 ,男 , 1983年生 ,硕士 ,主要 研究方向为统计机器学习和模式分类. 文益民 ,男 , 1969年生 ,博士后 ,副 教授 , CCF高级会员 ,主要研究方向为 统计学习理论、生物信息学和图像处 理. 发表学术论文 20余篇. 吕宝粮 ,男 , 1960年生 ,教授、博士 生导师、博士、IEEE高级会员 ,主要研 究方向为仿脑计算理论与模型、神经网 络理论与应用、机器学习、模式识别、 脑 —计算机接口、生物信息学与计算生 物 学. 已 在 IEEE Trans. Neural Net2 works, IEEE Trans. Bimedical Engineering,Neural Networks和 ICCV等国际期刊和会议上发表学术论文 80余篇. ·156· 智 能 系 统 学 报 第 4卷