Publication: Deep learning neural network for time series water level forecasting
No Thumbnail Available
Date
2021
Authors
Zaini N.
Malek M.A.
Norhisham S.
Mardi N.H.
Journal Title
Journal ISSN
Volume Title
Publisher
Springer Science and Business Media Deutschland GmbH
Abstract
Reliable forecasting of water level is essential for flood prevention, future planning and warning. This study proposed to forecast daily time series water level for Malaysia�s rivers based on deep learning technique namely long short-term memory (LSTM). The deep learning neural network is based on artificial neural network (ANN) and part of broader machine learning. In this study, forecasting models are developed for 1-h ahead of time at multiple lag time which are 1-h, 2-h and 3-h lag time denoted as LSTMt-1, LSTMt-2 and LSTMt-3, respectively. Forecasted water level is significant for determination of effected area, future planning and warning. Root mean square error (RMSE) and coefficient of determination (R2 ) are utilized to evaluate the performance of proposed forecasting models. An analysis of error in term of RMSE and R2 show that the proposed LSTMt-3 model outperformed other models for water level forecasting during training and testing phase. � The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021.
Description
Deep neural networks; Flood control; Forecasting; Learning systems; Long short-term memory; Mean square error; Offshore oil well production; Time series; Water levels; Coefficient of determination; Daily time series; Forecasting models; Learning neural networks; Learning techniques; Root mean square errors; Training and testing; Water level forecasting; Deep learning