Unless I'm missing something, the usual suspects don't have this....
Asked
Active
Viewed 1.8k times
15
-
9Consider the catch-22... we could give you a decision tree for choosing one but you'd have no way of evaluating it. :-) – cletus Jun 27 '10 at 16:12
-
Drop Python and use R (or Rpy). – mbq Jun 27 '10 at 16:24
-
1@mbq: prepare to defend such a bold claim. Why drop Python and use R? – Eli Bendersky Jun 27 '10 at 16:26
-
@Eli that is just a suggestion; it may be a simpler solution, so I mentioned it, but of course only Ash R can judge that. – mbq Jun 27 '10 at 16:57
-
Just to keep you from going insane implementing your own logic in Python, why not wrap C4.5 in a subprocess or use that through an Orange module or use the tree-building modules in Orange? http://www.ailab.si/orange/doc/reference/C45Learner.htm – ddotsenko Jun 27 '10 at 20:27
-
@mbq: Ash's last name is R, so he's already using it. :) – tzot Jun 27 '10 at 20:35
-
@ΤΖΩΤΖΙΟΥ: I don't think so; in R it is so simple that there is no need to ask. – mbq Jun 27 '10 at 20:40
-
@mbq: …so he's already using it (as a last name, in the Ash R persona)… I just reply because your answer seems to be serious. Anyway, it's not your fault, just my unfunny sense of humour. – tzot Jun 27 '10 at 21:48
-
I know; probably I should have used a ';-)'. Nevertheless its true. – mbq Jun 27 '10 at 23:32
-
Thanks folks - in the end I found using weka and jython was the best solution. – Ash Jun 29 '10 at 14:24
-
There is a related question on Stats Stackexchange site: http://stats.stackexchange.com/questions/2419/boosted-decision-trees-in-python – jb. Dec 07 '11 at 13:56
3 Answers
4
There is also Scikit Learn: http://scikit-learn.org/stable/modules/tree.html#classification Haven't tried it though (but I'm about to).
EDWH
- 491
- 2
- 5
- 11
4
There is a DecisionTreeLearner class as part of the Python library for Russell & Norvig's "Artificial Intelligence: A Modern Approach" textbook.
Brandon
- 3,684
- 1
- 18
- 25
0
I was finding python decision tree library, too. there are many open source decision tree libraries on the internate, and I found out DecisionTree from Kak, who is a professor in Purdue, is the most useful one.
just want to update the information, so people who are looking for decision tree library can save some effort.
Unfortunately, the library does not implement numeric feature values and treats each number as a different class.
