Rainfall Prediction in DKI Jakarta Using a Hybrid Model (DWT-SVR-Prophet)

Abstract

Rainfall has a significant influence in the planning and management of various industries such as forestry, agriculture, and water resources. This research aims to develop a Hybrid model that combines Discrete Wavelet Transform (DWT), Support Vector Regression (SVR) and Prophet models to predict rainfall in the DKI Jakarta area more accurately. Using DWT, rainfall data is divided into high and low frequency components, then the prediction results from each model are combined. At Kemayoran Station, the Hybrid model (High Frequency SVR and Low Frequency Prophet) provides the best performance with SMAPE: 36.44%. At Tanjung Priok Station, the non-Hybrid model gave the best results, with Prophet without DWT achieving SMAPE: 29,82%. This study provides a clear understanding of how effective rainfall prediction models are in DKI Jakarta, which helps water resources planning and management.