Forecast Rainfall Data Time Series Using Multi-Attribute Long Short-Term Memory

The high rainfall in Indonesia cannot be easily predicted based on the moon due to climate change. The method that is often used to obtain accurate prediction results is based solely on rainfall data on one rain gauge without considering the effects of the location of other rain gauge stations. In this study a new model was developed by presenting in a web form that can predict rainfall. Daily rainfall data obtained online from 3 rain gauge stations belonging to the Meteorology, Climatology and Geophysics Agency (BMKG) in Surabaya is processed as input in the form of training data and testing data In this modeling, after the rainfall data is obtained, the data is divided as many as n attribute features over several periods and labeled. Then the prediction process is carried out using Multi-Attribute LSTM. The results of the predicted values obtained are also labeled for comparison with labeling in the actual data and error values are generated in the form of RMSE and MSE values. This process is carried out by producing a prediction of daily rainfall for 1 year for each BMKG climatology station in the city of Surabaya. The results of the analysis are in the form of error ratios obtained through the testing process by comparing the results of predictive data with actual rainfall data from BMKG.