Deep Learning Models for Predicting High and Low Tides With Gravitational Analysis

Abstract

In this chapter, many forecasting models were established, and each model was produced in three forms, each with a distinct machine learning algorithm: regression, random forest, and artificial neural network. All of the presented models performed well in predicting the tide level at Visakhapatnam port. All models based on supervised learning had an accuracy of 85 to 9%, with a relative absolute error of 5 to 7.5. Finally, with a forecast horizon of several hours, satisfactory results were obtained, and a further specific comparison revealed that the models based on the considered machine learning algorithms outperform the auto learning integrated evaluation models with exogenous input variables in forecasting high/low tides.