Application of Deep Neural Network for Predicting River Tide Level

Abstract

In riverine countries like Bangladesh, residents have to depend on rivers for many things for their daily life. If the river has river port or intercostals waterways or estuaries than it becomes more significant to country. So, knowing the upcoming water height or tide level at the river becomes an important fact to people for planning their daily activities, like: fishing, water way communication, port activities controlling etc. Moreover, people living at the bank of the river can anticipate sudden flooding by having updated information about tide height at the river. The aim of this study is to propose a machine learning model which will be able to percept short term future tide level at a river. For undertaking the experiments, we have collect 10 years (2007–2017) of historical dataset of the Karnaphuli river from Chittagong Port Authority's (CPA) hydrographic department. It is the mainstream river of Bangladesh having four tide gauge stations named Kalurghat, Canal-10), Canal-18 and Sadarghat. We have designed our study using Support Vector Machine (SVM), Artificial Neural Network with back propagation (BP-ANN) and Deep Neural Network (DNN). After careful and meticulous analysis we have found DNN model outperformed the others with almost 99% accuracy in future water level prediction.