Using Data Mining to Preprocess Data for the Neural Network Model to Predict Water Level Applied for Northern Vietnam’s Agriculture

Water resource in agricultural canal is a necessary requirement in water resource system management. Nowadays, in Bac Hung Hai, a small area in Red River Delta (Northern Vietnam), it is essential to have a system for water level forecasting. In this research, we suggest several methods which are SARIMA, long short-term memory (LSTM) and hybrid seasonal-wavelet-LSTM to predict water level for water administration. We employed the data from Cau Cat station located on Red River (the water input of Bac Hung Hai canal system) to predict water level in 7 days. The results will be used to prepare for the accumulating or discharging of water for the canal. The study demonstrates that the seasonal-wavelet-LSTM model performs better than the single SARIMA model and the classic single LSTM model (in MSE and MAPE), respectively. The investigation provides a promising option for operating the water level of Bac Hung Hai canal, where the water flow is often unstable. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.