Settlement prediction of Nanjing Metro Line 10 with HOA-VMD-LSTM

• The HOA-VMD-LSTM model is proposed to estimate settlement accurately. • The key hyperparameters of VMD and LSTM were optimized using HOA. • HOA-VMD-LSTM is able to accurately predict different types of settlement. • The HOA-VMD-LSTM-based application was developed. The development of the subway will cause settlement, posing a risk to the safety of surrounding structures. The empirical formula method, physical experiment method and numerical simulation method have challenges in predicting settlement. To accurately, promptly, and cost-effectively predict settlement, this study introduces the hiking optimization algorithm-variational mode decomposition-long short-term memory model (HOA-VMD-LSTM). HOA is employed to optimize the parameters of VMD and LSTM. The optimized VMD algorithm is applied to decompose the settlement data, and the resulting components of the decomposition and the original data are used as inputs and outputs to form a new dataset, respectively. The resulting reconstructed data is subsequently utilized for optimized LSTM prediction. The viability of the proposed model was exhibited through the anticipation of the settlement of four monitoring points on Nanjing Metro Line 10. The programmed model is user-friendly and is expected to be further promoted in other settlement predictions.