Abstract: The paper proposes a classification method of Gene Expression Programming（GEP） based on the principle of maximum degree of membership, which is named MDM-GEP. Describing fuzziness of classification by membership degree of fuzzy set, the GEP classifier approximating membership function is obtained on training data set. For the instance to be classified, it computes the membership degree of in fuzzy sets, and determines the final class based on the principle of maximum degree of membership. The experiments carried on three datasets from the UCI machine learning repository show that MDM-GEP not only is effective for classification, but also resolves the unclassifiable region problems in the conventional simple GEP classification strategy.

Key words: classification, principle of maximum degree of membership, gene expression programming

摘要： 提出了一种基于最大隶属度原则的基因表达式编程（Gene Expression Programming，GEP）分类方法MDM-GEP。引入模糊集合中的隶属度描述分类的模糊性，在训练集上得到逼近各类别隶属函数的GEP分类器。对于待分类实例，计算其在各模糊集中的隶属度，基于最大隶属度的模糊模式识别原则确定最终归属类，并在三个UCI数据集上对该算法进行了实验。实验结果表明，MDM-GEP不仅具有较好的分类性能，而且有效解决了传统的简单GEP分类方法中存在的拒分区域问题。

关键词: 分类, 最大隶属度原则, 基因表达式编程