發信人: yaomc (白頭翁&山東大漢), 信區: DataMining
標  題: [合集]有誰用過cart的？
發信站: 南京大學小百合站 (Fri Nov 30 10:44:09 2001), 站內信件

study (.net) 于Mon Oct 15 13:19:58 2001提到：

誰知道cart怎么用啊？
大概介紹一下使用的感受和模塊功能
如何？


roamingo (漫步鷗) 于Mon Oct 15 19:31:42 2001提到：

對, 介紹一下吧! 最好還有輸入輸出的數據格式.



yaomc (白頭翁&山東大漢) 于Mon Oct 15 19:54:06 2001提到：

不知道看看相關的幫助是否可以解決問題？



study (.net) 于Tue Oct 16 00:45:58 2001提到：

這東西好像很難搞定，如果沒有一定的理論知識！
希望有高手先介紹一下！



yaomc (白頭翁&山東大漢) 于Tue Oct 16 08:24:43 2001提到：

確實沒有用過。不過，好像是決策分類的東西，是不是啊？


matrix (天行健，君子以自強不息) 于Tue Oct 16 09:44:17 2001)
提到：

有人做講座嗎？



yaomc (白頭翁&山東大漢) 于Tue Oct 16 09:46:03 2001提到：

現在好像還沒有。
也說不定過一段時間那位專家就給大家上一課呢。
呵呵。


roamingo (漫步鷗) 于Tue Oct 16 09:58:21 2001提到：

A quick Google search:

CART: A Brief Description

CART stands for "Classification and Regression Trees", and, as it's name 
implies, it is a method of categorizing very large sets of data, very 
quickly. You often hear the term "Data-mining" used these days to describe 
the process of analyzing data to find relationships between elements, or 
building a predictive system to explain the data. Most of the data-mining 
done today involves classification, and therefore can benefit greatly 
from the use of CART. 

CART was introduced in 1984 by Leo Breiman, Jerome Friedman, Richard Olshen,
and Charles Stone; all UC Berkeley and Stanford statisticians. Their CART 
methodology has vastly improved various performance bottlenecks, as well as 
problems in accuracy. Some of this was achieved by using only binary 
splitting, and by introducing automatic tree testing. They also came up 
with a completely different, and improved method for dealing with missing 
values.

More information (example, bibliography): 

    http://www.cs.mcgill.ca/~mbatch/CART.html 

CART FAQ: http://www.timberlake.co.uk/software/cart/cartfaq.htm

CART 4.0 是決策樹(decision tree)和回歸樹(regression tree)的商業化軟件, 有很強
的理論背景.