利用一种具有混合编码的二进制差分演化算法进行特征选择,并针对传统评价函数对样本邻域局部信息利用不足的问题,引入邻域粗糙集模型,省去了对数值型数据的离散化过程,同时分析了依赖度度量和邻域识别率度量可能存在的问题,综合两者提出加权邻域识别率,对特征子集适应度进行更好的评价.通过UCI数据进行实验,结果表明该方法能有效进行特征选择. A binary differential evolution algorithm with hybrid coding was used to select features. Aiming at the problem that the traditional evaluation function did not make full use of the local information in the neighborhood of the sample, a neighborhood rough set model was introduced, eliminating the discretization of numerical data. At the same time, the possible problems of the dependence measure and the neighborhood recognition rate measure are analyzed, and the weighted neighborhood recognition rate is proposed to evaluate the fitness of the feature subset better.Experimental results show that the method Effective feature selection.