River tide level prediction: A data mining approach for hydrographie time series data analysis

摘要

Prediction is one of the most complicated and challenging tasks; if it comes to tidal prediction then it becomes more complicated because of the chaotic nature of gradual increment in water level. Since Bangladesh is a low lying land, performing tidal prediction in seaport area is really a key factor since the economy of the country largely depends on the smooth operations of exports and imports in the 3 major seaports. Predicting hundred percent accurate tide levels is very difficult. The main objective of this study was to propose a good machine learning model with the aid of data mining techniques for the projection of water level with higher accuracy for Karnaphuli River, which is one of the mainstream rivers in Bangladesh. The river comprises of 4 major boatyards and shipyards for anchoring lighter ship vassals and boats; those are Sadarghat, Kalurghat, Canal no-10, Canal no-18. For undertaking the experiments, we collect 5years (2008-2012) historical tide level dataset of Karnaphuli River from Chittagong Sea Port Authority (CPA). Finally, we got our proposed model that can predict tide level with more than 96% accuracy in overall scenarios, which outperform the previous models that are used in literature review and reference sections.