RED TIDE PREDICTION USING THE MACHINE LEANRINGIN COASTAL AREA

Abstract

 Red tides occurring in the coastal area has a tremendous negative impact on the fishery industry. The occurrence of red tides is strongly affected by the coastal environment. In the Ariake Sea including Isahaya Bay, red tides occur frequently. Although continuous observation of water quality has been conducted for other years in coastal area, the forecasting model of red tides has not been constructed yet. In this study, using big data mesured in Ariake Sea and Isahaya Bay, we tried to make a phytoplankton prediction model from SVR, RF, and GBRT of machine learning. As a result, there was no problem in performance if the data was divided into learning and prediction data at random. However, when the data was divided in time series, the Chl-a fluctuation characteristic in the prediction data was not included in the learning data, so it was confirmed that the reproducibility of the time series model is poor.