Exploring a long short-term memory for mountain flood forecasting based on watershed-internal knowledge graph and large language model.

As can be seen from Figs 6 and 7 , the overall stability of the LSTM series, RNN and GRU have a certain vibration, while the LSTM is relatively smooth, with a clear trend and no overfitting but with a certain time lag, indicating that the data collection affecting water level elements has a certain time lag, while KG and LLM can find more relevant influencing factors, improve the problem of time lag and improve the accuracy of prediction.

also in 19,460 other papers