This paperback book describes a relatively new, computer based method for deriving a classification rule for assigning objects to groups. As the authors state in their preface:
Binary trees give a n interesting and often illuminating way of looking at data in classification or regression problems. They should not be used to the exclusion of other methods. We do not claim that they are always better. They do add a flexible nonparametric tool to the data analyst's arsenal. The authors use the acronym CART (derived from the book title) in referring to their method and take pains to compare results using CART with those obtained by more conventional types of classification, such as linear discriminant analysis and nearest neighbor classification.
An example of classifying heart attack patients into two groups, those who will survive 30 days (low risk) and those who will not (high risk), is used in the first chapter to introduce the method. After examining 19 variables, including age and blood pressure, the following classification tree was produced:
6 Low admission 14 High Using this simple procedure 89% of low risk and 75% of high risk patients were correctly classified. Linear stepwise discriminant analysis required 12 variables to achieve slightly less accurate results. Logistic regression required 10 variables, including 3 interactions sug-@ 1987 AIan R. Liss. Inc.