Rainfall Prediction in North Barito Regency with Recurrent Neural Network and Support Vector Machine Models

Riveroverflows in North Barito Regency often cause flooding, which occurs especially during high rainfall events. Currently, many attempts have been made to utilise technology to predict rainfall. This research uses historical rainfall data of Beringin Meteorological Station, Montallat Rain Post, Kandui Rain Post, Teweh Tengah Rain Post and Beriwit Rain Post from 2010 to 2023. The models used are Recurrent Neural Network (RNN) and Support Vector Machine (SVM). These two models are tested for performance and will be compared using the RMSE, MAE, and MSE values. The evaluation results show that the RNN model at Beringin Meteorological Station has low RMSE, MAE, MSE evaluation metric values and has the smallest difference in metric values between training data and testing data compared to RNN models at other rainfall observation locations and SVM models at all other rainfall observation locations. The RNN evaluation metric value of Beringin Meteorological Station when the data is trained produces an RMSE of 0.0571, MAE of 0.0299, and MSE of 0.0033 and when the data is tested produces an RMSE of 0.0587, MAE of 0.0304, and MSE of 0.0034. The SVM evaluation metric value of Beringin Meteorological Station when the data is trained produces an RMSE of 0.0933, MAE of 0.0393, and MSE of 0.0087 and when the data is tested produces an RMSE of 0.1118, MAE of 0.0594, and MSE of 0.0125.