摘要
决策树是数据挖掘领域的经典算法,应用领域非常广泛。在信用评价任务中,数据之间存在序关系,而传统的决策树算法无法解决这类问题。有序决策树的提出有效地解决了此类问题,能够从中发现新的知识,然而很多任务中属性与决策存在单调关系,并且样本之间无法比较,这影响了有序分类器的性能。因此,文章提出一种改进的有序决策树算法(Rank-DT)并应用于信用评价任务中,实验证明提出的算法改进了传统决策树算法的性能,获得了较好的效果。
Decision trees are a kind of classic algorithm in the data-mining field,and the application is very extensive.In the credit evaluation tasks,it exists an ordinal relationship in the data,and the traditional decision tree algorithm cannot solve such problems.The ordinal decision trees are effective to solve these problems,and it can discover new knowledge.However,there are monotonic relationships between attributes and decisions on many tasks.Moreover,it be compared among samples,which affect the performance of ordinal classifiers.Therefore,this paper proposes an improved decision tree algorithm(Rank-DT)and applies it to the credit evaluation tasks.Experiments results show that the proposed algorithm improves the performance of the traditional decision tree algorithm and obtains good effect.
作者
裴生雷
周伟
PEI Sheng-lei;ZHOU Wei(College of Physics&Electronic Information Engineering,Qinghai University for Nationality,Xining 810007,China;Xining big data Service Management Bureau,Xining 810000,China)
出处
《齐齐哈尔大学学报（自然科学版）》
2020年第4期9-13,共5页
Journal of Qiqihar University(Natural Science Edition)
基金
青海省应用基础研究项目(2019-ZJ-7017)
青海民族大学高层次项目(2020XJG13)
青海民族大学多源数据融合及应用科研创新团队。
关键词
序关系
决策树
信用评价
可比较的
ordinal relationship
decision trees
credit evaluation
comparable
作者简介
裴生雷(1980-),男,山东潍坊人,副教授,博士,主要从事机器学习与数据挖掘研究,peishenglei@qhmu.edu.cn。
