How to Choose a Bayesian Analyzer
Choosing the right Bayesian-style analyzer to run on your network is difficult because there are many Bayesian implementations to choose from. Some of the questions to ask when choosing a Bayesian implementation include
What level of sophistication are my users?
How many users am I supporting?
Is message scoring (for example, SpamAssassin) “good enough?”
Is retraining the Bayesian analyzer important to my configuration? If so, can I handle the support load? Can my users learn how to manage filter retraining?
How accurate do I want the analyzer to be?
As with most things, choosing which Bayesian-style classifier to use is about managing trade-offs. The more features a solution has, the harder it is to manage. The ...