针对支持向量回归模型在实际应用中难以选择建模数据的特征、核函数类型、最优核函数参数以及支持向量的惩罚系数和不敏感损失函数的参数问题，本文提出利用粒子群算法混合模拟退火算法构造协同进化算法，自适应选择建模数据的特征、协同进化选择支持向量回归的核函数类型，匹配最优核函数的参数和模型参数，以此建立协同进化的支持向量回归柳江径流预测模型。本文通过混合模拟退火算法Metropolis接受准则和粒子群算法随机接受准则，结合粒子群算法的并行计算和模拟退火算法的干扰机制，减少在局部搜索点附近的震动偏离构造协同进化算法，自适应筛选支持向量回归模型建模数据的特征、协同进化选择核函数类型，最优匹配核函数参数和支持向量模型参数，建立柳江径流建模，实验结果表明，该方法能够快速有效地选择建模数据的特征、核函数类型、最优匹配核函数参数和支持向量模型参数，其泛化性能明显提升，拟合效果更好。该方法用于柳江径流预测，具有良好的建模效果和更高的预测精度。 Aiming at the difficulty of selecting the appropriate kernel function type, kernel function parame-ters, penalty coefficients of support vectors, and insensitive loss parameters in practical applica-tions of support vector model, in this paper, co-evolution algorithm is employed to simultaneously evolve selection the data features, the kernel type of support vector regression, the parameters of the kernel function and the model parameters, namely HPSOSA-SVR by hybrid the strengths of Particle Swarm Optimization (PSO) and Simulated Annealing (SA). In order to select the optimal support vector regression kernel function type, the optimal matching kernel function parameter and the support vector model parameter for runoff modelling, the co-evolution algorithm focuses on combining the advantages of PSO (fast calculation and easy mechanism) and SA (ability to jump away from local optimum solutions and converge to the global optimum solution) by combined the metropolis processes of SA into the movement mechanism and parallel processing of PSO. The HPSOSA algorithm has the capability of both fast calculation and searching in the direction of the global optimum solution, helping PSO jump out of local optima, avoiding into the local optimal solution early and leads to a good solution. Compared with HPSOSA, PSO and pure SVR, results show that the HPSOSA-SVR model can successfully identify the optimal type of kernel function and all the optimal values of the parameters of SVR with the lowest prediction error values in runoff forecasting, can significantly improve the runoff forecasting accuracy. Experimental results reveal that the predictions using the proposed approach are consistently better than those obtained using the other methods presented in this study in terms of the same measurements.
展开▼
机译：针对支持向量回归模型在实际应用中难以选择建模数据的特征、核函数类型、最优核函数参数以及支持向量的惩罚系数和不敏感损失函数的参数问题，本文提出利用粒子群算法混合模拟退火算法构造协同进化算法，自适应选择建模数据的特征、协同进化选择支持向量回归的核函数类型，匹配最优核函数的参数和模型参数，以此建立协同进化的支持向量回归柳江径流预测模型。本文通过混合模拟退火算法Metropolis接受准则和粒子群算法随机接受准则，结合粒子群算法的并行计算和模拟退火算法的干扰机制，减少在局部搜索点附近的震动偏离构造协同进化算法，自适应筛选支持向量回归模型建模数据的特征、协同进化选择核函数类型，最优匹配核函数参数和支持向量模型参数，建立柳江径流建模，实验结果表明，该方法能够快速有效地选择建模数据的特征、核函数类型、最优匹配核函数参数和支持向量模型参数，其泛化性能明显提升，拟合效果更好。该方法用于柳江径流预测，具有良好的建模效果和更高的预测精度。 Aiming at the difficulty of selecting the appropriate kernel function type, kernel function parame-ters, penalty coefficients of support vectors, and insensitive loss parameters in practical applica-tions of support vector model, in this paper, co-evolution algorithm is employed to simultaneously evolve selection the data features, the kernel type of support vector regression, the parameters of the kernel function and the model parameters, namely HPSOSA-SVR by hybrid the strengths of Particle Swarm Optimization (PSO) and Simulated Annealing (SA). In order to select the optimal support vector regression kernel function type, the optimal matching kernel function parameter and the support vector model parameter for runoff modelling, the co-evolution algorithm focuses on combining the advantages of PSO (fast calculation and easy mechanism) and SA (ability to jump away from local optimum solutions and converge to the global optimum solution) by combined the metropolis processes of SA into the movement mechanism and parallel processing of PSO. The HPSOSA algorithm has the capability of both fast calculation and searching in the direction of the global optimum solution, helping PSO jump out of local optima, avoiding into the local optimal solution early and leads to a good solution. Compared with HPSOSA, PSO and pure SVR, results show that the HPSOSA-SVR model can successfully identify the optimal type of kernel function and all the optimal values of the parameters of SVR with the lowest prediction error values in runoff forecasting, can significantly improve the runoff forecasting accuracy. Experimental results reveal that the predictions using the proposed approach are consistently better than those obtained using the other methods presented in this study in terms of the same measurements.
展开▼
