随机决策树[1](Random Decision Trees),是由Wei Fan等人提出的一种决策树构建方法,它的一大特点就是不需要任何特征选择。因其生成过程具有随机性,对集成学习准确性与差异性的研究提供了新的思路[2][3]。1随机决策树算法描述在整个构建过程中,任何一个结点的生成都是从特征集中随机选取的。对于离散特征来说,每个子树的生成都对应一个离散特征值,且不可以被重复选取。对于连续特征而言,可将其离散 Random Decision Trees [1] is a decision tree construction method proposed by Wei Fan et al. One of its major characteristics is that it does not require any feature selection. Because of its randomness, the research provides a new way to study the accuracy and diversity of integrated learning [2] [3]. A Random Decision Tree Algorithm Description Throughout the construction process, the generation of any one node is randomly selected from the feature set. For discrete features, the generation of each subtree corresponds to a discrete eigenvalue, and can not be repeatedly selected. For continuous features, it can be discrete