| Abstrak/Abstract |
Heavy rains in Indonesia occur yearly, and one of the impacts is flood disasters. Flooding occurs frequently and unpredictably. Modelling such important occurrences can help to identify vulnerable locations and reduce the effects. Recently, researcher applied machine learning to analyzing data and its correlations in order to predict how the climate will perform. However, most machine learning algorithms cannot automatically detect the dataset’s quality; for example, how long the time interval for the dataset to make good forecasting predictions is. Using ensemble machine learning and Bayesian optimization, we explored for the best interval and model to predict rainfall. The ensemble machine learning algorithm achieved the best result, showing the superiority of ensemble
machine learning over single machine learning in discovering the best interval training set for rainfall prediction. The best interval to predict rainfall is 61-hour, with mean squared error score of 12.97 and mean absolute error score of 2.24. |
