为了克服排序学习算法不能处理包括名词性特征的复杂数据类型的局限性,设计一种新的排序学习算法.在决策树学习算法中,采用新的等级不纯度定义,修改决策树的分裂规则,得到具有直观解释的排序算法,并给出了相关理论基础.实验结果表明:排序树的平均等级损失明显优于感知机类算法和序回归类算法,且具有较快的收敛速度.基于决策树的排序学习算法,可以处理名词性数据和选择相关的特征. In order to overcome the limitation that the ranking learning algorithm can not handle complex data types including nominal features, a new ranking learning algorithm is designed.In the decision tree learning algorithm, a new definition of grade impurity is introduced to modify the split rules of decision tree, We get the order algorithm with intuitionistic interpretation and give the relevant theoretical foundation.The experimental results show that the average rank loss of the ranking tree is obviously better than the perceptual machine type algorithm and the ordinal regression algorithm and has a faster convergence rate.Based on the decision tree The sort learning algorithm can handle nominal data and select related features.