Chung-Kwei algorithm

[Peter Bishop]

On 12 Aug 2004 at 23:33, Chris Fortune wrote:

> We trained Chung-Kwei on a repository of 87,000
> messages,
> then tested it with a very large collection of 88,000 pieces of SPAM and
> WHITE email: the current prototype achieved a sensitivity of 96.56%
> whereas
> the false positive rate was 0.066%, or one-in-six-thousand. In terms of
> speed,
> we are currently capable of classifying 214 messages/second, on a 2.2
> GHz

Not bad, but no better than my bogofilter setup
~ 10,000 spam+ham corpus 
FP: 0.09%
FN: 0.13% (99.87% sensitivity)

PS 0.066%, is not one-in-six-thousand.
it is about one in 1500 messages
> Intel-Pentium platform. The Chung-Kwei system is part of SpamGuru, a
> collaborative antispam filtering solution that is currently under
> development at
> IBM Research


-- 
Peter Bishop 
Adelard LLP and Centre for Software Reliability, City University
Drysdale Building, 10 Northampton Square, London, EC1V 0HB
Tel: +44-20-7490-9467, Fax: +44-20-7490-9451
pgb at adelard.com, http://www.adelard.com/
pgb at csr.city.ac.uk, http://www.city.ac.uk/