Ensemble Learning and Active Learning Based Personal Spam Email Filtering

Abstract

Abstract:

This paper proposes a personal spam email filtering method, which can learn a user’s interests and update it automatically according to the user’s feedback. The proposed method extracts the linguistic features and behavior ones to build some rulebased individual filters, and uses the SVM ensemble learning method to combine the multifilter’s results. Applying an active learning method to choose those knowledgeable emails with the user’s labels, the method can minimize the number of labeled emails and reach steadystate performance more quickly. The experimental results show the personal filtering method based on ensemble learning and active learning can capture personality, and achieve high performance with the considerations on accuracy, efficiency and learning ability.

Key words: spam email filtering;personal;ensemble learning;active learning;SVM

LIU Wuying,WANG Ting. Ensemble Learning and Active Learning Based Personal Spam Email Filtering[J]. J4, 2011, 33(9): 34-41.

References

［1］Denning P J. Electronic Junk ［J］. ACM Communications, 1981, 25(3):163165.
［2］Lindberg G. AntiSpam Recommendations for SMTP MTAs ［M］. Chalmers University of Technology,1999.
［3］Leiba B, Fenton J. DomainKeys Identified Mail (DKIM): Using Digital Signatures for Domain Verification［C］∥Proc of the Fourth Conf on Email and AntiSpam (CEAS 2007),2007.
［4］Fleizach C, Voelker G M, Savage S. Slicing Spam with Occam’s Razor［C］∥Proc of the Fourth Conf on Email and AntiSpam (CEAS 2007), 2007.
［5］Cohen W W. Fast Effective Rule Induction［C］∥Proc of the Twelfth Int’l Conf, 1995:115123.
［6］Carreras X,Marquez L. Boosting Trees for AntiSpam Email Filtering［C］∥Proc of Euro Conf Recent Advances in NLP (RANLP2001), 2001:5864.
［7］Nicholas T. Using AdaBoost and Decision Stumps to Identify Spam Email［R］. Stanford University Course Project Report, 2003.
［8］刘洋, 杜孝平, 罗平, 等. “垃圾邮件”的智能分析、过滤及Rough集讨论［C］∥第12届全国计算机网络与数据通信大会, 2002.
［9］Sahami M, Dumais S, Heckerman D,et al. A Bayesian Approach to Filtering Junk EMail［C］∥Proc of AAAI Workshop on Learning for Text Categorization,1998:5562.
［10］Androutsopoulos I,Paliouras G,Karkaletsis V,et al.Learning to Filter Spam EMail: A Comparison of a Naive Bayesian and a MemoryBased Approach［C］∥Proc of the 4th European Conf on Principles and Practice of Knowledge Discovery in Databases (PKDD 2000), 2000:113.
［11］Drucker H, Wu D, Vapnik V N. Support Vector Machines for Spam Categorization［J］. IEEE Transactions on Neural Networks, 1999, 20(5):10481054.
［12］Littlestone N. Learning Quickly When Irrelevant Attributes Abound: A New LinearThreshold Algorithm［J］. Machine Learning, 1988, 2(4):285318.
［13］刘伍颖, 王挺. 一种多过滤器集成学习垃圾邮件过滤方法［C］∥第三届全国信息检索与内容安全学术会议(NCIRCS2007),2007.
［14］Cristianini N,ShaweTaylor J. An Introduction to Support Vector Machines and Other KernelBased Learning Methods［M］. Cambridge University Press, 2000.
［15］Engelbrecht A P. Incremental Learning Using Sensitivity Analysis［C］∥Proc of the Int’l Joint Conf on Neural Networks, 1999:13501355.
［16］Lewis D D, Catlett J. Heterogeneous Uncertainty Sampling for Supervised Learning［C］∥Proc of the Eleventh Int’l Conf on Machine Learning, 1994:48156.
［17］Seung H S, Opper M, Somepolinsky H. Query by Committee［C］∥Proc of the 5th Annual Workshop on Computation Learning Theory, 1992:287294.
［18］Tong S, Koller D. Support Vector Machine Active Learning with Applications to Text Classification［J］. Journal Machine Learning Research, 2001(2):9991006.
［19］Cohn D A, Ghahramani Z, Jordan M I. Active Learning with Statistical Models［J］. Artificial Intelligence Research, 1996,4(1):129145.
［20］Wuying Liu, Ting Wang. Active Learning for Online Spam Filtering［C］∥Proc of the Asia Information Retrieval Symp(AIRS2008), 2008:555560.
［21］Cormack G,Lynam T. TREC 2005 Spam Track Overview［Z］. University of Waterloo, Waterloo, Ontario, Canada.
［22］Chang ChihChung, Lin ChihJen. LIBSVM: A Library for Support Vector Machines［EB/OL］.［20090427］.http://www.csie.ntu.edu.tw/~cjlin/libsvm.