Rezzy Eko Caraka, Rung Ching Chen, Budi Darmawan Supatmanto, Arnita, Muhammad Tahmid, Toni Toharudin
Rainfall is significant in influencing human life. Therefore, it is necessary to predict or forecast rainfall in decision making. Forecasting rainfall can be calculated by the average rainfall of an area and by using the time-series method. Moreover, the government has a climatology station to measure rainfall at specific points or locations in various regions. In Indonesia, they are considered to have potential and represent the surrounding area. However, rainfall outside the climatology station area is not known for sure, while for specific purposes, information about rain is needed at other points. This research work focuses on the application of machine learning methods to the problem of computing prediction on time series as input variables. More specifically, we employ moving average (MA) and long short-term memory (LSTM) method to predict the rainfall in Winangun, North Sulawesi, Indonesia. LSTM is a neural network development that can be used for time-series data modelling. Based on the simulation, the combination of these methods, in-sample data reaches the R2 95.11%, and out-sample data reach R2 90.46% respectively. © 2019 IEEE.
College of Informatics/ins Chaoyang, University of Technology, Taichung City, 41349, Taiwan; Agency for the Assessment and Application of Technology (BPPT), Weather Modification Technology Center; State University of Medan, Department of Mathematics, Medan, Indonesia; Meteorology Climatology and Geophysical Agency (BMKG), Manado, Indonesia; Padjadjaran University, Department of Statistics, West Java, Indonesia
