结合深度核学习与高斯过程的边坡稳定性预测方法OA

Given the complex nonlinear relationships among slope features and between these features and stability evaluation,classical Gaussian process-based slope stability prediction methods are limited in modeling intricate structures and are difficult to scale to large slope datasets.A slope stability prediction method combining deep kernel learning and Gaussian process is therefore proposed.A multilayer feedforward network first performs deep extraction of slope features,and the latent space is then mapped into a Gaussian process with a radial basis function kernel to realize nonparametric uncertainty quantification.By maximizing the marginal log-likelihood,the model jointly optimizes neural network weights and kernel hyperparameters,enabling end-to-end learning of a data-driven optimal kernel.Experiments on a public Kaggle dataset show that the proposed method surpasses classical machine learning algorithms of Random Forest(RF),Support Vector Machine(SVM),and Gaussian Process Regression(GPR),and the deep learning methods of Gated Recurrent Unit(GRU)and Deep Neural Network(DNN),achieving the best performance in root mean square error,mean absolute error,and coefficient of determination,thus providing new technical support for intelligent early warning of slope hazards.