基于加权变异免疫算法的微钙化点特征选择*

给出了一种乳腺X线照片微钙化点的特征选择方法，该方法运用基于加权变异算子的免疫算法进行特征优选。加权变异算子能够动态调整抗体各部位的变异率，在高亲和力抗体的邻近小范围搜索，在低亲和力抗体的周围跳跃式搜索；为了与支持向量机的分类准则保持一致性，该免疫算法在特征空间中通过核函数计算亲和力。实验使用该方法对微钙化点的20种常用特征进行选择，其结果与经验特征集基本相符但更精简，提高了计算效率，是一种可行的特征选择方法。

Micro calcification feature selection based on weighted mutation immune algorithm

This paper proposed a feature selection method of micro-calcification(MC) in mammograms,which used weighted mutation based immune algorithm.The weighted mutation operator could dynamically adjust the mutation rates of different parts of antibodies,making searching nearby the high-affinity antibodies and around the low-affinity ones.Also,according to the classification principal of support vector machine classifier,computed the affinity in the feature space by kernel functions.Experiments show that the result feature set selected by the proposed method among the 20 commonly used features is consis-tent with the empirical one and smaller,which can lead to better performance.It is proved to be a feasible feature selection method.