基于分类回归树(CART)方法的统计解析模型的应用与研究

分类回归树是基于统计理论的非参数的识别技术，它具有非常强大的统计解析功能，对输入数据和预测数据的要求可以是不完整的，或者是复杂的浮点数运算。而且，数据处理后的结果所包含的规则明白易懂。因此，分类回归树已成为对特征数据进行建立统计解析模型的一个很好的方法。本文首先介绍了一种构建分类回归树的算法，并对其剪枝策略进行了简单的探讨，最后用统计解析软件S－PLUS对一个应用实例进行了分析，给出结果。

Research and application of the statistical models based on CART

CART (Classification and Regression Tree) is a kind of non-parameter recognizing technology based on the statistical theory. The construction of CART has become a common basic method for building statistical models from simple feature data. CART is powerful because it can deal with incomplete data, multiple types of features floats both in input features and predicted features and the trees it produces often contain rules which are humanly readable. In this article, I first introduce some arithmetic applied to build a kind of CART, and then discuss the prune method of CART At last, I analyze an application example with the software S|PLUS.