<![CDATA[In this research, we have been done developed a prediction model for Sea Surface Temperature (SST) in the sea of Manokwari City, Manokwari Regency using deep learning models, especially the Deep Neural Networks (DNN) model. The SST data used is ERA5 reanalysis data provided by the European Center for Medium-Range Weather Forecast (ECMWF) from 2000-2021 (8233 daily data). The SST data is divided into two parts, namely training and testing data with a proportion of 90% and 10%, respectively. The DNN model developed uses the hyperparameter optimizer adam, the ReLu activation function, the learning rate is 0.01, the Batch size is 30, the number of inputs is 10, the number of epochs is 100 and is equipped with early stops. Meanwhile, the number of hidden layers varied between 1 until 4. Likewise, the number of neurons in each hidden layer varies from 8, 16, 32, 64, or 128 neurons. Based on the test results, the DNN model with 2 hidden layers and 32 neurons per hidden layer gives more accurate results than the other models, with RMSE, MAE, and R2 values respectively 0.121; 0.015; and 0.935. Therefore, this DNN model can be recommended as a model to predict SST in sea of Manokwari City.]]>
Copyrights © 2023
