论文部分内容阅读
鉴于支持向量机(SVM)存在结构稀疏化不足、缺乏概率信息等缺陷,将性能更具优势的相关向量机(RVM)理论引入到大坝变形预测的应用中。选择高斯径向基函数作为RVM模型的核函数,核参数用基于模拟退火的混合粒子群算法(SAPSO)进行寻优,进而建立SAPSO-RVM回归预测模型。实例应用结果表明,RVM模型的向量数量远小于SVM模型,在保持良好泛化能力的前提下计算结构得到简化,混合粒子群算法相较于一般粒子群算法其全局寻优能力也有所提高,SAPSO-RVM模型回归预测精度较高。
Due to the lack of structural sparsity and the lack of probability information, support vector machine (SVM) is introduced into the application of dam deformation forecasting based on the more advanced RVM theory. Gaussian radial basis function is selected as the kernel function of RVM model. The kernel parameters are optimized by SAPSO (Simulated Annealing) algorithm, then the SAPSO-RVM regression prediction model is established. The results show that the number of vectors in the RVM model is much smaller than that in the SVM model, and the computational structure is simplified under the premise of maintaining good generalization ability. Compared with the general PSO algorithm, the hybrid particle swarm optimization algorithm also improves the global optimization ability. The SAPSO -RVM model regression prediction accuracy is higher.