Modeling Rainfall in Jakarta with Hybrid ARIMAX-ANN Model

Abstract

Abstract. Forecasting Indonesia's rainy and dry seasons is increasingly complex, requiring effective forecasting methods. This research is focused on rainfall forecasting in Jakarta, which includes Kemayoran Station and Tanjung Priok Station, using the Hybrid ARIMAX-ANN model. In ARIMAX modeling, the exogenous variable involved is the humidity variable. The accuracy results of the Hybrid ARIMAX-ANN model will be compared with the ARIMAX model. The results of this study show that the Hybrid ARIMAX-ANN model provides better accuracy. At Kemayoran Station, the Hybrid ARIMAX-ANN (1,0,0) model with 1 hidden layer shows a lower MAPE, 21.145%, than the ARIMAX model. Meanwhile, at Tanjung Priok Station, the Hybrid ARIMAX-ANN (1,0,0) model with 2 hidden layers has a lower MAPE, which is 37.416% compared to the ARIMAX model. The results show that the Hybrid ARIMAX-ANN model provides better accuracy than the ARIMAX model in rainfall modeling in Jakarta. Applying the Hybrid ARIMAX-ANN model at Kemayoran Station produces better accuracy than the application at Tanjung Priok Station.

Keywords: ARIMAX, Hybrid ARIMAX-ANN, Rainfall