基于GEP的最小二乘支持向量机模型参数选择

针对最小二乘支持向量机的多参数寻优问题,提出了一种基于基因表达式编程的最小二乘支持向量机参数优选方法.该算法将最小二乘支持向量机参数(C,σ)样本作为GEP的基因,按其变异算子随着进化代数和染色体所含基因数目动态变化的机制执行,其收敛速度和精确度大大提高.并与基于粒子群算法和遗传算法参数优选方法比较,通过标准测试函数验证了该算法的拟合误差最低.最后用其建立氧化铝生产蒸发过程参数预测模型,应用工业生产数据进行验证,实验结果表明该方法有效且获得了满意的效果.

A parameter selection method of a least squares support vector machine based on gene expression programming

To solve the multi-parameter optimization problem of least squares support vector machines(LSSVM),a parameter optimization method based on gene expression programming(GEP) was proposed.The parameter(C,σ) samples of LSSVM were selected to be genes for GEP according to the mechanism of the dynamic change of the mutation operator with the gene number of the genome and the number of evolutionary generations.As a result,the convergence rate and accuracy were greatly increased.The new method was compared with other parameter optimization methods based on particle swarm optimization(PSO) and a genetic algorithm(GA) by several standard test functions,and the results show that the proposed method obtains the minimum fitting error.Finally,a parameter prediction model of the evaporation process of alumina production was established;the verification results using the industrial production data show that the method is effective and the result is satisfactory.