Rainfall Prediction Using Gate Recurrent Unit (Gru) for The Mataram City Area

Galih Dimas Aryoso (Mataram University)
Bulkis Kanata (Mataram University)
Made Sutha Yadnya (Mataram University)

Jurnal Penelitian Pendidikan IPA (JPPIPA)
Vol 11 No 2 (2025): February

Rainfall prediction is crucial for urban planning, agriculture, and disaster mitigation. This study predicts rainfall intensity in Mataram City using the Gated Recurrent Unit (GRU), a variant of Recurrent Neural Networks (RNN) optimized for sequential data. The dataset consists of hourly rainfall data from NASA's MERRA Power (2010–2021). Data preprocessing includes normalization, feature engineering, and dataset splitting. The GRU model architecture comprises input, GRU, and dense layers. Model performance is evaluated using Root Mean Squared Error (RMSE), yielding 67, 112, 69, and 109 for Ampenan, Cakranegara, Majeluk, and Selaparang, respectively. Results show that the GRU model captures rainfall trends but has limitations in predicting extreme values. This study demonstrates GRU’s potential for improving rainfall forecasting while highlighting the need for further optimization to enhance accuracy.