Abstract: Aiming at the problems that genetic algorithm can’t make good use of system output information, it’s easily to fall into prematurity, and artificial bee colony algorithm is slow at the beginning of the search, this paper combines improved genetic algorithm and artificial bee colony algorithm to achieve?complementary, and takes the classification accuracy of SVM test set as fitness?function to propose Genetic-Artificial Bee Colony Algorithm（G-ABCA） of support vector machine parameter optimization, and compares G-ABCA with improved genetic algorithm and artificial bee colony algorithm by simulation?experiment. The experiment results show that the performance of G-ABCA in support vector machine parameter optimization is better than other algorithms, G-ABCA greatly improves the classification accuracy of SVM, and it is step-by-step convergent.

Key words: genetic algorithm, artificial bee colony algorithm, combination, support vector machine, parameter optimization

摘要： 针对遗传算法不能充分利用系统中的反馈信息，易陷入“早熟”，以及人工蜂群算法在搜索初期寻优速度慢的问题，将改进的遗传算法与人工蜂群算法融合，实现二者互补，并将由支持向量机训练得到的测试集分类准确率作为算法的适应度函数，提出遗传-人工蜂群算法（G-ABCA），以实现对支持向量机参数的优化选择。通过仿真实验，将其在支持向量机参数寻优中的性能与改进的遗传算法、人工蜂群算法进行比较，实验结果表明：遗传-人工蜂群算法有效地提高了支持向量机的分类准确率，而且算法是逐步收敛的，其表现优于其他算法。

关键词: 遗传算法, 人工蜂群算法, 融合, 支持向量机, 参数优化