针对特征选择问题,提出了基于竞争策略的链式智能体遗传算法(LAGA)。该LAGA算法包含链式智能体网络结构,邻域竞争,自适应交叉,自适应变异,优良个体替换策略,自适应结束等部分,该算法能较好的保持智能体的多样性,在进化中既较佳的继承了优良个体的基因,又有效地搜索了新的空间。多组实验结果表明,通过该算法选择得到的最优特征子集具有较好的稳定性,较高的识别准确率和较低的网络分类器维数复杂度。 Aiming at the problem of feature selection, a chain-agent-based genetic algorithm (LAGA) based on competitive strategy is proposed. The LAGA algorithm includes the network structure of chain agent, neighborhood competition, adaptive crossover, adaptive mutation, good individual replacement strategy, adaptive termination and so on. This algorithm can keep the diversity of the agent well in evolution It is better to inherit the genes of good individuals and effectively search for new space. The experimental results show that the optimal feature subset selected by this algorithm has better stability, higher recognition accuracy and lower complexity of the network classifier dimensionality.