Compare model multi-input RNN, LSTM and GRUfor prediction of irrigation canal's water level in Red river delta, North Vietnam

Abstract

Listen

Translate article

Forecasting water level on Red River is an important problem in Vietnam. We need to replace water level predicting models that based on experiences of hydro-meteorologists by machine learning models which provide faster as well as more accurate results. Therefore, we have applied several best machine learning methods with arti?cial neural networks such as ANN, RNN, LSTM, and GRU, compared these models. The results indicated that LSTM is most appropriate to Red River data, with 153.5% better than the worst model ANN (in MSE), and 1.58% better than the second best model GRU (in MSE).