Multi Time Series WA-LSTM-Adam for Water Level Forecasting in Center Vietnam
Nguyen Duc Khoa
High School for Gifted Students, Hanoi University of Science - VNU, Hanoi, Vietnam.
Nguyen Quang Dat *
High School for Gifted Students, Hanoi University of Science - VNU, Hanoi, Vietnam.
Vo Quang Linh
Hanoi-Amsterdam High School for Gifted Students, Hanoi, Vietnam.
Nguyen Ha Vy
High School for Gifted Students, Hanoi University of Science - VNU, Hanoi, Vietnam.
Vu Hoang Nam Khanh
Hanoi-Amsterdam High School for Gifted Students, Hanoi, Vietnam.
Phan Viet Hoang
Hanoi-Amsterdam High School for Gifted Students, Hanoi, Vietnam.
*Author to whom correspondence should be addressed.
Abstract
The central region of Vietnam suffers from oods almost every year as a result of a combination of frequent storms, heavy rainfall, and short, steep rivers in the region. This is a big problem because they can negatively affect the economy of the region as well as people's lives when not managed properly. Therefore, it is important to have a reliable forecasting method for ooding in order to ensure effective natural disaster management. In this research, we aim at addressing this issue by introducing a multi time series hybrid deep learning model that combines WA (wavelet analysis) and LSTM (long-short-term memory) optimized with the Adam algorithm and uses water level and rainfall data as the input variables. Compared to other traditional methods and some recent models, our WA-LSTM-Adam method shows better results overall.
Keywords: rainfall, water level, time series forecasting, wavelet, LSTM, adam