决策树的自动生成模板

用于知识发现的大部分数据挖掘工具均采用规则发现和决策树分类技术来发现数据模式和规则。该文通过采用基于仿真属性的离散化方法，基于概率统计的未知属性与噪声数据处理方法以及基于误差的剪枝算法，实现了用于自动生成决策树的通用算法模板。利用该模板，决策树算法的设计者可以快速验证为解决特定决策问题而设计的新算法。构造决策树的基本机制是算法的设计者利用其自己定义的公式来初始化通用算法模板。然后利用该系统提供的交互式图形环境，针对不同的决策问题测试该算法，从而找出适合特定问题的算法。

A Template for Automatic Generation of Decision Trees

Most data mining tools for knowledge discovery generally use rule discovery and decision tree technology to extract data patterns and rules. A general template for automatic generation of decision trees is provided by attribute based descrization methor, by statistic based unknown attributes and noisy data processing method, and by error based pruning method. The designers of automatic tree building algorithms can quickly evaluate the new algorithms for solving specific decision problems with this template. The basic mechanism for creating new tree building is a generic algorithm template, which is initialized by the algorithm designers with his own formula. Using the interactive graphic environment provided by this system, the new algorithms can be tested on different decision problems.