i don't under stand the technique of using CART to select the most important features and how when a CART is growing up, the input attribute appeared during splitting process plays an important role in the training data specially after 25 splits in the CART. Thank you in advance
Hello,
CART is a binary tree. During its construction best variable is identified that potentially leads to the largest decrease of impurity observed in two new child nodes. The impurity is effectively measured by a specific index, for instance, entropy index, Gini's index.
In the course of tree construction all variables and all possible splits are evaluated. The tree is grown untill a certain number of objects are found in terminal nodes. This parameter is usually specified by user.
Another issue is related to the selection of the optimal size of a tree, i.e. finding the optimal number of terminal nodes. This is usually guided using the V-fold cross validation and the cost-complexity pruning procedure.
More explanations can be found in original Breiman's et al. book the publication indicated below.
Kind regards,
Michal
Article Classification and Regression Trees-Studies of HIV Reverse T...
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification