内嵌多准则的遗传算法用于尿沉渣红白细胞的特征选择

针对尿沉渣检测中红白细胞的特征选择问题,提出了结合双向选择的内嵌多准则的改进遗传算法(NMGA)(通过多代进化逐步选中优良特征,在每一代进化中基于多种评价准则进行遗传,分别得出各自的最佳特征子集,然后对其进行多票投选,得出一个最佳特征子集,然后继续下一代进化,该方法可结合小生境技术共同搜索最佳特征子集。文末采用多个尿沉渣红细胞和白细胞样本进行验证实验,实验结果表明,算法优选的特征集与未进行特征选择和经过普通遗传算法(SGA)特征选择得到的特征集相比,识别率较高,并且优选后的特征子集的特征数较少,明显降低了分类器的维数复杂度。

Research of Feature Selection of Red Cell and White Cell of Urinary Sediment Images Based on Genetic Algorithm Embedded with Multi-criteria

Based on the feature selection issue of red cells and white cells, one new method (NMGA) for it was proposed. That is: according to the feature selection about red cells and white cells, the modified genetic algorithm was adopted to deal with it. The modified genetic algorithm used gene-fixing technology (gradually fix the features during several generations), and combined the Niche technology together to enhance the performance of the genetic algorithm. Besides, voting mechanism was introduced with multi-criteria during each evolution operation. The effect of these experiments show that this algorithm performs better than simple genetic algorithm (SGA), the features are reduced by this algorithm, and the complexity of the classifier is reduced apparently.