Abstract:
Uniformly distributed constants-based decision tree evolved by Gene Expression Programming(GEP) is a kind of classifier with fairly high accuracy, but its performance on multi-attribute data classification is not satisfactory. This paper presents an algorithm of Differential Evolution (DE)-based decision tree algorithm by GEP. This new algorithm uses differential evolution method to improve the additional threshold, and makes the uniform constant array have both uniformly and diversity. Experiments on benchmark datasets show it performs better on multi-attribute classification problems than basic GEP decision tree.
Key words:
Gene Expression Programming(GEP),
decision tree,
Differential Evolution(DE)
摘要: 基于均匀常数分布的基因表达式编程决策树算法存在对多属性数据分类效果不佳的问题。为此,提出一种基于差分演化的基因表达式编程决策树算法,该算法通过引入差分演化的方法对其附加阈值进行改进,从而使均匀的常数数组在保持均匀分布的同时仍不失多样性。实验结果表明,该方法在多属性数据的分类问题上能够得到良好的效果。
关键词:
基因表达式编程,
决策树,
差分演化
CLC Number:
WANG Wei-Gong, RUAN Wei, LI Qu. Decision Tree Algorithm by Gene Expression Programming Based on Differential Evolution[J]. Computer Engineering, 2011, 37(01): 181-183.
王卫红, 阮薇, 李曲. 基于差分演化的GEP决策树算法[J]. 计算机工程, 2011, 37(01): 181-183.