12th December 2005, 09:52 PM
|
|
Member
|
|
Join Date: Jan 1970
Posts: 4,426
|
|
The reason I didn't reply to your post, is because you asked the questions and then went on to basically answer them yourself.
Personally, I would not dissect the systems too much, as you risk trimming down the data which will distort the results.
You are right, there are dangers in using too many rules, the best systems are straightforward and have the least number of rules.
Bear this in mind, the more rules you have, the more data you need to try it over, as the probability of backfitting increases with each rule.
|