融合粒子群的缎蓝园丁鸟优化算法及应用

Satin Bower Bird Optimization Algorithm with Particle Swarm and Application

针对缎蓝园丁鸟优化算法(SBO)收敛精度低、收敛速度慢和全局寻优能力弱的问题,提出了融合粒子群的缎蓝园丁鸟优化算法(PSBO)。该算法在原缎蓝园丁鸟优化算法的基础上,通过引入速度因子和固定惯性权重来提高种群的多样性,进一步提高了原算法的寻优性能。通过8个标准测试函数对PSBO算法、SBO算法、基于自适应t分布变异的缎蓝园丁鸟优化算法、自适应权重缎蓝园丁鸟优化算法和粒子群优化算法这五个算法进行测试比较,实验结果表明,PSBO算法在收敛速度、精度和算法稳定性上都有很大程度的提高。为了进一步说明PSBO算法的有效性,把PSBO算法应用于支持向量机(SVM)内部参数的优化上。SVM是一类按监督学习方式对数据进行二元分类的广义线性分类器,主要影响SVM性能的参数是核参数和惩罚因子。在训练集上,用K折交叉验证的方法算出准确率的均值作为目标函数,通过PSBO算法对核参数和惩罚因子进行寻优,并将参数寻优的结果代入测试集进行样本测试,结果表明PSBO算法在优化参数时拥有更快的收敛速度。

Due to the problem of low optimizationaccuracy,slow convergencerate and weak global Optimization ability of the Satin Bower bird optimization(SBO)algorithm,a Particle Satin Bower Bird Optimization(PSBO)algorithm is proposed.Based on the SBO al⁃gorithm,the speed factor and fixed inertia weight are introduced to improve the diversity of the population and further improve the per⁃formance of the original algorithm.The experimental results in eight test functions show that PSBO algorithm has a higher convergence accuracy,rate,and better optimization ability compared with SBO algorithm,satin bower bird optimization based on adaptive weight(WSBO)algorithm,satin bower bird optimization based on adaptive t-distribution mutation(tSBO)algorithm and Particle Swarm Opti⁃mization(PSO)algorithm.In order to demonstrate the effectiveness of PSBO algorithm,it is applied to the optimization of internal pa⁃rameters in Support Vector Machines(SVM),SVM is a kind of generalized linear classifier which classifies data binary according to su⁃pervised learning,and the main parameters that affect SVM performance are kernel parameter and penalty factor.On the training set,using K-fold cross validation method to calculate the average accuracy as the objective function,the PSBO algorithm is used to opti⁃mize the kernel parameters and penalty factors,and the results of parameter optimization are substituted into the test set for sample test⁃ing.The results show that the PSBO algorithm has a faster convergence rate for the optimized parameters.