|
 Buy Now
 Demo
 Upgrade
More in-depth information about AnswerTree's algorithms
Get unprecedented analytical power for better models
AnswerTree offers four powerful algorithms that enable you to build the best model, for any type of data: CHAID, Exhaustive CHAID, Classification and Regression Trees (C&RT) and QUEST.
CHAID
A CHAID tree is a decision tree that is constructed by splitting subsets of the space into two or more child (nodes) repeatedly, beginning with the entire data set.
Exhaustive CHAID
Exhaustive search CHAID was proposed by Biggs et al. (1991). Biggs suggests finding the best split by merging similar pairs continuously until only a single pair remains. The set of categories with the largest significance is taken to be the best split for that predictor variable.
Classification and Regression Tree (C&RT) Method
The Classification and Regression Trees (C&RT) method of Breiman et al. (1984) generates binary decision trees. The C&RT tree is constructed by splitting subsets of the data set using all predictor variables to create two child nodes repeatedly, beginning with the entire data set. The best predictor is chosen using a variety of impurity or diversity measures. The goal is to produce subsets of the data which are as homogeneous as possible with respect to the target variable.
QUEST
QUEST stands for Quick, Unbiased, Efficient, Statistical Tree. The original method is described in Loh and Shih (1997). It is a tree-structured classification algorithm that yields a binary decision tree like C&RT. A binary tree allows techniques such as pruning, direct stopping rules and surrogate splits to be used. Unlike CHAID and C&RT, which handle variable selection and split point selection simultaneously during the tree growing process, QUEST deals with them separately.
QUEST was demonstrated to be much better than exhaustive search methods in terms of variable selection bias and computational cost. In terms of classification accuracy, variability of split points and tree size, however, there is still no clear winner when univariate splits are used.
Buy Now
Demo
Upgrade
|
