递减步长果蝇优化算法及应用

Research on and application of diminishing step fruit fly optimization algorithm

提出一种递减步长果蝇优化算法(diminishing step fruit fly optimization algorithm,DS-FOA).该算法的搜索步长随果蝇觅食进程逐步减小,从而使果蝇群体在觅食初期具有较强的全局搜索能力,在觅食后期具有较强的局部寻优能力,从而实现全局搜索能力和局部寻优能力的平衡.将该算法用于支持向量机(support vector machine,SVM)回归模型的惩罚因子和核函数参数优化中,结果表明,DS-FOA收敛速度快,全局搜索与局部寻优能力强.与其他算法相比,由DS-FOA优化参数的SVM回归模型均方误差最低,回归效果好.

A diminishing step fruit fly optimization algorithm （DS-FOA） is proposed. The step length is decreased progressively along with the process of foraging. DS-FOA demostrates preferable global optimization capability in early stage and local optimization capability in later period. Dynamic balance is achieved between global and local optimizing capability. Also DS-FOA is applied in the field of support vector machine （SVM） regression model parameter optimization. Experimental results show that the DS-FOA has fast convergence speed and powerful global and local optimization capability. The SVM model using DS-FOA has the lowest error of mean square and the best optimization result.