Abstract: Based on a new automatic programming method——gene expression programming(GEP), this paper presents an improved GEP mining algorithm by designing new fitness function, optimizing the processing of initialization, introducing new genetic operators and adopting(λ+μ) selection strategy which is used in evolution strategy. Experiments are done to test the new algorithm and the data sets are from UCI machine learning repository. The capability of the new algorithm to mine accurate classification rules is compared with C4.5 algorithm and traditional GEP algorithm.

Key words: GEP, Classification rules, Data mining

摘要： 基于一种新的自动程序设计方法基因表达式程序设计（GEP），通过设计适应函数、初始化群体的优化、增加新的遗传算子以及采用演化策略中的（λ+μ）淘汰策略等对原始GEP算法进行有效的改进，设计出一种新的数据挖掘算法。采用UCI机器学习知识库中的数据集对该算法进行了实验，并通过与C4.5及文献[3]的比较，检验了该算法的准确性。

关键词: GEP, 分类规则, 数据挖掘

CLC Number: 

• TP301.6