Displacement Prediction of the Muyubao Landslide Based on a GPS Time-Series Analysis and Temporal Convolutional Network Model

Landslide displacement prediction is an essential base of landslide hazard prevention, which often needs to establish accurate model. To achieve accuracy displacement, a model based on salp-swarm-algorithm-optimized temporal convolutional network (SSA-TCN) proposed. The TCN model, consisting causal dilation convolution layer residual block, can flexibly increase the receptive fields and capture global information in deeper layer. SSA solve hyperparameter problem well for Muyubao collected from professional GPS monitoring system implemented 2006 used analyze features slope evaluate performance SSA-TCN cumulative time series decomposed into trend (linear part) periodic (nonlinear by variational modal decomposition (VMD) method. Then, polynomial function predict considering response relationship between rainfall, reservoir water. This research also compares proposed approach results with other popular machine learning deep models. demonstrate that hybrid superior more effective than others at predicting displacement.