对一种新的决策树建立方法的研究

The Research for a New Method of Building Decision Trees

1 决策树数据挖掘技术中的决策树技术是首先从机器学习领域得来的,它与关联规则技术作为数据挖掘技术的两个大方向,在许多领域都有广泛的应用,如:医学、地质学、天文学、物理学、金融领域等等,并发挥着巨大的作用。决策树技术能建立分类系统及产生预测系统。从训练样本集出发,它的建立是一个速归过程。它重复执行以下过程:根据评价标准选择数据中的某一属性作为分割标准,将当前节点(数据集)分割成子节点(数据子集),如果各个子节点中的数据属于同一类,以类名标注,过程结束,否则过程在包含不同类数据的子节点中进行。

The methods of building decision trees that have been developed all begin from training data sets. And then they select the best attribute to split the data by their different criterion. At last a decision tree is got which can classify and predict. To be different to other methods,a new method of building decision trees is present in this paper. It begins from training data set in which all data belong to one class. At last it gets a composite decision tree.