Classification And Regression Trees
Users: 1 - Average Rating: 4.00
This is one of the pioneering books on classification and regression trees (CART), which constitutes an alternative to the construction of a prediction equation as it is done in linear regression. In CART, data are partitioned along the predictor variables into subsets with homogeneous values of the dependent variable, and then a decision tree emerges to make predictions from new observations in a similar vein as in the Guess Who? board game. But depending on the problem, the basic purpose can be either to produce an accurate classifier or to uncover the predictive structure of the problem, using Bayes rule as an underlying main tool. Along with two introductory and two intermediate illustrative chapters (with medical, ship and mass spectra classification examples) and how to get the right size of the tree and accurate estimates of misclassification probability, some issues as splitting rules (variable misclassification costs by generalizing Gini criterion or by prior alteration), best splitting and Boolean combinations are dealt with to improve flexibility, power, and efficient of classification trees. Then authors proceed to adapt their technique to (claimed simpler) regression trees, so the final tree procedures (Bayesian partitions and optimal pruning using joint distributions, and topics related with resampling the learning sample and its size tending to infinity) can be developed for the general framework.
Type:
Book
Area:
Machine Learning
Target Group:
Basic
DOI:
https://doi.org/10.1201/9781315139470
Cite as:
L. Breiman, J. Friedman, R. Olshen, C. Stone (1984), Classification and Regression Trees. Routledge (Taylor and Francis imprint), New York. ISBN 9781315139470 (ebook).
Author of the review:
Pablo Guerrero-Garcia
University of Malaga
You have to login to leave a comment. If you are not registered click here
Davide Ilardi