Landslide Displacement Prediction Based on a Two-Stage Combined Deep Learning Model under Small Sample Condition

The widely distributed “Step-type” landslides in the Three Gorges Reservoir (TGR) area have caused serious casualties and heavy economic losses. prediction research of landslide displacement will be beneficial to establishment local geological hazard early warning systems for realization scientific disaster prevention mitigation. However, number observed data like displacement, rainfall, reservoir water level this is very small, which results difficulties training advanced deep learning model obtain more accurate results. To solve above problems, a Two-stage Combined Deep Learning Dynamic Prediction Model (TC-DLDPM) predicting typical TGR under condition small samples proposed. process method as follows: (1) Time warping (DTW) used enhance cumulative obtained by Global Positioning System (GPS); (2) A Difference Decomposition Method (DDM) based on sequence difference proposed, decomposes into trend periodic then cubic polynomial fitting predict displacement; (3) component predicted proposed TC-DLDPM combined with external environmental factors such rainfall level. combines advantages Convolutional Neural Network (CNN), Attention mechanism, Long Short-term Memory network (LSTM) carry out two-stage parameter transfer, can effectively realize construction high-precision samples. variety models are compared model, it verified that accurately especially case drastic changes factors. capture spatio-temporal characteristics dynamic evolution reduce complexity calculations. Therefore, provides better solution exploration idea