Naive Bayesian classification is like a good magic trick, perplexing until explained and
continuing to provide entertainment. The idea is to use observations from the past to classify a new obeservation. The classic Bayesian example made famous by Paul Graham is the Bayesian spam classifier, which is first trained on some messages at explicitly told which messages are "spam" and which are "clean". After training, it is presented with each new email and "guesses" whether it is spam or clean, based on the "features" of the message.

by The author of Reverend, "a general purpose Bayesian classifier" written in Python
[via Keith]

[ by Martin>] [permalink] [similar entries]

similar entries (vs):

• freshmeat.net: Project details for Bayesian Pattern Filtering Library (# 14%)
• Dave Farquhar on the new naive bayesian spam filter in Mozilla (# 12%)
• CRM114 Based Usenet filter (# 11%)
• Automatic Document Classification (# 11%)

similar entries (cg):

• more on Text Summarization (# 8%)
• some more Bayesian Perl (# 7%)
• Due Diligence: Bayesian Nets, Latent Semantics, (# 5%)
• Play with the Machine: Java and classifying stuff (# 5%)
• freshmeat.net: Project details for Bayesian Pattern Filtering Library (# 4%)
• developerWorks: Six approaches to eliminating unwanted e-mail (# 3%)
• Dave Farquhar on the new naive bayesian spam filter in Mozilla (# 2%)

Martin Spernau
© 1994-2003

Big things to come (TM) 30th Dez 2002