遗传算法优化支持向量机在大坝变形预测中的应用

建立大坝变形预测的支持向量机模型，并用遗传算法对支持向量机模型的核函数参数、惩罚参数和损失参数进行优化。将同一优化方法不同支持向量机核函数、不同优化方法同种支持向量机核函数进行横向对比，将BP神经网络、自回归AR(p)模型、多元回归分析法和周期函数拟合法进行纵向对比。结果表明，该GA-SVM(RBF)模型不仅能较好地预测大坝的变形趋势，而且能大幅提高预测精度。

Dam Deformation Monitoring Prediction on Support Vector Machine Optimized by Genetic Algorithm

A SVM model is established for predicting dam deformation, and optimizing the kernel function parameter, penalty parameter and loss function parameter through the genetic algorithm. We use this model to analyze the long period deformation monitoring data and make predications. In this paper, we compare horizontally different kernel functions of support vector machine using the same optimization method, and the same kernel function of support vector machine using different optimization methods. The results show that GA-SVM(RBF) not only can well predict the dam deformation trend, but also improves the prediction accuracy over contrasting BP neural networks, AR(p), multiple regression analysis and periodic function fitting longitudinally.