中文摘要：

针对简单遗传算法用于特征选择精度不高、过早收敛的问题，提出了链式遗传算法（Link—like Agent Genetic Algorithm），并与多准则（MC）相结合，从而实现了基于多准则竞争策略的链式遗传算法并用于特征选择（LAGA＋MC）研究。LAGA引入了链式个体结构，遗传个体相互进行竞争选择和自适应交叉、自适应变异，从而获得更精确的搜索结果。MC通过对基于单准则进行选择得到的特征子集进行特征位判断，已达到更全面评价选择结果，获得识别率更稳定更高的特征子集。实验结果表明，本文算法获得的特征子集分类准确率比其他几种基于遗传算法的特征选择算法更高、更稳定。

英文摘要：

According to low precision and over early convergence problems, the link-like agent genetic algorithm （LAGA） is presented for combining feature selection with multi-criteria （MC）. LAGA introduces link-like agent structure, competition selection, adaptive crossover and adaptive mutation, so it can obtain more precise search result. MC can judge the feature bits of the feature subset obtained by single criterion, thus final feature subset can obtain a more comprehensive and more stable result. Experimental results show that the feature subset obtained by the algorithm has a better classification rate and a more stable classification result than other algorithms.