Presentation 2007-03-15
An online spam filtering system for changing situations using multiple classifiers
Kenta NARUMI, Kyousuke NISHIDA, Koichiro YAMAUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Spam mail is a serious problem now. To eliminate many spam mails, many spam mail filters referring mail bodies have been proposed. The filters with statistical learners (ex. Naive Bayes) are usually used because of its high accuracy. However, it is difficult for the filters to learn mails that have new tendency after learning of many mails. We have to solve this problem to deal with spam mails that are changing day by day. In contrast, there are filters with instance-based learners (ex. Nearest Neighbor) that are able to respond to the changes quickly. However, they are not used widely because they require large computational complexity and memory resources to store many mails. In this study, we proposed a spam filter that is able to respond to various changes by using an instance-based learner that store recent mails and using statistical learners built from enormous past mails. We showed the proposed spam filter achieved higher accuracy than other spam filters in experiments using real dataset.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) spam mail filter / multiple classifier systems / concept drift / online learning / baysian learning / instance-based learning
Paper # PRMU2006-235
Date of Issue

Conference Information
Committee PRMU
Conference Date 2007/3/8(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) An online spam filtering system for changing situations using multiple classifiers
Sub Title (in English)
Keyword(1) spam mail filter
Keyword(2) multiple classifier systems
Keyword(3) concept drift
Keyword(4) online learning
Keyword(5) baysian learning
Keyword(6) instance-based learning
1st Author's Name Kenta NARUMI
1st Author's Affiliation Division of Synergetic Information Science, Graduate School of Information Science and Technology, Hokkaido University()
2nd Author's Name Kyousuke NISHIDA
2nd Author's Affiliation Division of Synergetic Information Science, Graduate School of Information Science and Technology, Hokkaido University
3rd Author's Name Koichiro YAMAUCHI
3rd Author's Affiliation Division of Synergetic Information Science, Graduate School of Information Science and Technology, Hokkaido University
Date 2007-03-15
Paper # PRMU2006-235
Volume (vol) vol.106
Number (no) 605
Page pp.pp.-
#Pages 6
Date of Issue