PDA

View Full Version : [newbie question] Expert system?



vertigo
07-05-2007, 07:53 AM
Hi!

I dont't know very much about AI and data mining, sorry if my question is nonsense.

I'm looking for a system to solve this problem:


There are some let's say "reports" to classify in 6 categories: form 1=easy to 6=hard. There are unexperienced people have problems to find out which category to choose. But also, there are experts.

My idea is to create a list of questions with few possible responses for each question, then create some kind of expert system that will help unexperienced people to classify their reports. Such system would be "trained" by experts.

Is this possible with WEKA? or other system?

I would appreciate any help and comment! If you need other information, please ask!!

Thanks,
Stan

xCojotex
07-11-2007, 06:47 AM
Hi vertigo,

your question isn't nonsense, it's just another problem that can be solved by weka *g*.
I don't know how much you have already worked with data mining and what you already know and what you don't know.
In your case Weka can learn models from already classified reports. What you need is a larger amount of data (the answer to your questions and the resulting class value). Weka supports different models for classification like Naive Bayes, 1R, Prism algorithm or Part,.... Which one you should use depends on your data.
Weka learns its model from your training set and tests it on an evaluation set. If you have a great amount of already answered questions you can divide them in 3 Parts. 2 Parts training set 1 part evalution set. If you don't have enough data you should work with crossvalidation. Crossvalidation means that weka automatically divides your data in training an evaluation set and makes multiple runs for different divisions.