摘要
代价敏感学习方法常常假设不同类型的代价能够被转换成统一单位的同种代价,显然构建适当的代价敏感属性选择因子是个挑战。设计了一种新的异构代价敏感决策树分类器算法,该算法充分考虑了不同代价在分裂属性选择中的作用,构建了一种基于异构代价的分裂属性选择模型,设计了基于代价敏感的剪枝标准。实验结果表明,该方法处理代价机制和属性信息的异质性比现有方法更有效。
Usually, cost-sensitive learning assumes that different types of cost can be converted into a unified units of the same price. Apparently how to construct appropriate cost-sensitive attribute selection factor is a challenge. In this paper,a kind of heterogeneous cost-sensitive decision tree algorithm was designed, which fully considers the different cost in selecting split attribute, constructs an attribute selection model based on heterogeneous cost-sensitive, designs the price sensitive pruning strategy based on cost-sensitive. The experimental results show that this method is effective and more efficient than the present other methods.
出处
《计算机科学》
CSCD
北大核心
2013年第11A期140-142,146,共4页
Computer Science
基金
国家自然科学基金项目(61170131)
广西创新团队项(GXNSFGA060004)
广西师范大学项目资助
关键词
决策树分类
代价敏感学习
异构代价敏感
Decision-tree classification, Cost-sensitive learning, Heterogeneous cost-sensitive