Detail Abstract

Application of Machine Learning Model in predicting Hanoi Water Level

Abstract

The lower portions of the Red River and Thai Binh River in the North are where Hanoi, the capital of Vietnam, which is important to the socioeconomic development of the nation, is situated (Red River Delta or Northern Delta). Therefore, accurate flood forecasting, specifically the water level, in Hanoi will be crucial to preventing flooding, protecting public safety, and promoting social and economic improvement. For predicting water levels in Hanoi, authors have developed the Long Short-Term Memory Neural Networks (LSTM) model, a special type of Regression Neural Network (RNN). The average daily water level at three stations in Son Tay, Hanoi, and Ba Lat from 1962 to 2019 and the average daily traffic at two stations in Son Tay and Hanoi from 1962 to 2019 are the input data for the forecast model. Forecasting value accuracy evaluation is carried out by using the Coefficient Nash Sutcliffe Efficiency (NSE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). The findings reveal a similar trend between the anticipated and actual values. The best forecast values are for one day; the Nash coefficient was 0.93, MSE at 0.13, and RMSE was 0.37; this indicates that the forecast\'s quality is satisfactory. As a result, this model has the ability to estimate Hanoi\'s water levels, providing a foundation for future research.