一种混合编码的差分演化算法应用于特征选择问题

A differential evolution algorithm with hybrid encoding in feature selection

针对特征选择问题,引入一种具有混合编码的二进制差分演化算法,融合随机和优选两种方法自适应对特征子集个数进行控制,并在评价函数中引入权重和相关性抑制.利用淋巴瘤细胞核形态特征进行实验,结果表明该方法能有效进行特征选择.

This paper applys a binary differential evolution algorithm with hybrid encoding（HBDE）in feature selection.Aims to adjust the size of feature subset adaptively,a method which combines randomization with optimization is proposed.the combination of random and the principle of superior is priority is advanced.And apply the weight and correlation analysis in the evaluation criteria.The instance of the malignant lymphoma morphology feature selection shows the method is available and feasible.