DM: Decision Trees

[canarelli]

hi all,
Some times ago, there has been a discussion in this forum regarding 
pros and
cons of various decision trees algorithms (in particular CARTŪ  vs 
CHAID).
Basically the conclusion was the following:

1.There are 3 families of decision tree algorithms:  
  1) The CARTŪ  family (CARTŪ , IND CARTŪ , Splus CARTŪ , etc.) 
  2) The ML family (ID3, C4.5, C5 and other derivatives, etc.)
  3) The AID family (THAID, CHAID, XAID, TREEDISC, etc.)

2. The differences between the 3 families are small. They concern:
  o motivation behind the algorithm
  o splitting criteria 
  o stopping criteria
  o scale type of the dependent/criterion variable
  o scale type of the independent/input variables.

3. All lead to similar results and none outperforms the others on a 
large
number of datasets.

I am more interested in knowing precisely what are the fundamental 
technical
differences  between (or assumptions behind) these 3 families of 
decision trees.
Does anyone can help me in understanding this (pointers to literature,
comments, related web site, ...) ?

Thanks in advance.
Patrick.

_______________________________________________________________

Patrick Canarelli
Managing Director
COMPLEX SYSTEMS
18 rue d'Abbeville
F-75009 PARIS

Tel/Fax: (33) 01 40 82 93 12
Email: patrick.canarelli@filnet.fr
_______________________________________________________________