In your idea which is the best technique for machine learning. I have used neural networks but NN require a lot of examples to make it work properly. Tell me about some alternatives.

Which method is "best" depends on the context of the problem: the nature of available data (data types, missing values, redundant variables, noisy variables, and so on), the evaluation metric being used, the need to train incrementally, etc.

I will recommend as a start, "Modeling Methodology 3: Algorithm Selection", which is available at:

http://will.dwinnell.com/will/willTechnicalPublications.html

If that doesn't put you to sleep, consider also "Modeling Methodology 4: Localizing Global Models", at the same location, for some more detail on local vs. global models.

-Predictor




[edited by - Predictor on October 7, 2003 8:22:49 PM]