Customer Reviews

Classification and Regression Trees

5 star:		(5)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

In 1984 Brieman, Olshen, Friedman and Stone published this book and produced a software product called CART that made tree classification popular. These algorithms were very useful in medical applications and the book illustrated some simple success stories particularly ones from Richard Olshen's experience working in the Medical School at UC San Diego.

Olshen and Gordon did some of the work on the asymptotic theory of recursive partitioning that made the methodology credible to the statistical research community. The methods began to be applied to pattern recognition problems and also to the development of expert systems. Today data miners use these tools.

These ideas goes back a lot further than these authors. However, previous attempts at recursive partitioning algorithms tended to grow trees with too many terminal nodes. These authors introduced two important ideas. One was to grow the trees overly long and then prune them back. The second was to continually use cross-validation to evaluate the trees.

This book is still very valuable 24 years after it was first published. It is also readible by general audiences for the most part. It now stands as a classic text on the subject of classification and regression trees. There are also books that followed in its footsteps and other places where tree structure comes into play.

This book is a must-have for all serious decision trees researchers. It explains the underlying algorithms of classification and regression trees methods in details. It's not for beginners though. It's a bit outdated by now as trees methodology has advanced much with the invention of boosting, bagging, and arcing.

A good introduction to classification and regression trees with a variety of examples. You need never regress again! Many will find some of the technical topics difficult but then I found the statistical grounding to be rewarding in the end. My only complaint is that the book is near worthless to practitioners like myself without software which is a little hard to find and then pricey.

The book was a very good introduction to Classification andRegression Trees, and coupled with the software ( ) make for apowerful approach to solving traditional classification problems.

For me, this 1984-issued monograph is more valuable than many contemporary books on CART (I actually read the 1983 edition with gold & brown hardcover, but the contents is the same). The reason is that it shows the tortuous, painstaking trial and error process of creating and adjusting CART methodology.

Therefore, if your goal is to learn yet another statistical recipe quickly, this book is not for you. If, on the other hand, you want to pick up some of the research methodology of the great quantitative modelers, I can highly recommend it. It will positively affect your thinking and, in the end, you will be able to tackle quantitative problems a lot better, be it with or without CART.

This is the original textbook written by the pioneers of the Classification And Regression Trees algorithm, which has now been cited in over 2200 academic journals. While some of the material can be fairly complex, the authors take great pains to make the material accessible. Many examples are given, and the algorithm and process are broken down into discrete pieces. It is the single best resource I have encountered on the topic.

NOTE: One odd aspect of the text is that there has been absolutely ZERO effort has been expended on either graphic design or typesetting. Instead, it looks like a high school student typed it out late one night. If such details are important to you, you should skip this one.