On CBS MoneyWatch: Dumbest Things You Do With Money
BNET Business Network:
BNET
TechRepublic
ZDNet

FREE Registration is required

Overview:

This paper describes an e-mail spam filter based on local SVM, namely on the SVM classifier trained only on a neighborhood of the message to be classified, and not on the whole training data available. Two problems are stated and solved. First, the selection of the right size of neighborhood is shown to be critical; their solution is based on the estimation of the a-posteriori probability of the correct decision, and the resulting algorithm is called highest probability SVM nearest neighbor (HP-SVM-NN). The second problem is the application of the algorithm in practice, and they propose practical filter architecture based on HP-SVM-NN. Extensive testing is performed on SpamAssassin corpus and TREC 2005 Spam Track corpus, showing that HP-SVM-NN outperforms pure SVM and is applicable in practice.

(Is this item miscategorized? Does it need more tags? Let us know.)

Format:PDFSize:719 KB
Date:Mar 2008
Pages:27
advertisement

White Papers, Webcasts, and Resources

Featured Training Courses

advertisement

SmartPlanet

  • Thought-provoking progressive ideas on diverse topics that intersect with technology, business, and life, and matter to the world at large. Visit SmartPlanet
  • More from IBM
  • Innovate your business' process model, play against the market, compete against others on our scoreboards and WIN! Try INNOV8 2.0: A BPM Simulator
  • Enabling Real-World Business Transformation through IBM Service Management Read the EMA Analyst Report
Click Here

Returning users: Log In Here!

Already registered on BNET, TechRepublic, or ZDNet? Simply log in.

Free Membership: Sign Up Now!

Sign up for a free membership today and get instant and unlimited access to one of the largest databases of white papers, webcasts, and casestudies anywhere. Your FREE membership allows you to:

  • Download an unlimited amount of content, including classic and current white papers, case studies, webcasts and more
  • Track content on your chosen topics of interest
  • Receive targeted email alerts when your favorite content is added
  • Save content for future reading
  • Receive our member newsletter

When you register to access this library, you allow us to share your information with companies that produce products or services featured in the library--so that such companies may contact you with information and offers regarding their products and services. This enables us to keep the library a free service. As a library registrant, you will receive a complimentary subscription to the ZDNet white paper newsletter and e-mail Must-Read News Alerts. You can unsubscribe from these at any time. By clicking the Sign up button, you indicate that you agree to our Terms and Conditions and have read and understand our Privacy Policy (updated).