<![CDATA[This study compares the performance of Long Short-Term Memory (LSTM) and Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) models for daily rainfall forecasting in South Tangerang City using meteorological data from January 2005 to July 2025. Data from official meteorological stations was processed with mean imputation for missing values and MinMaxScaler normalization. Models were evaluated based on Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Squared Error (MSE), and coefficient of determination R². Results show CNN-LSTM outperforms with RMSE 0.79, MAE 0.63, MSE 0.62, and R² 0.61, compared to LSTM (RMSE 0.83, MAE 0.60, MSE 0.68, R² 0.58). Prediction visualizations confirm CNN-LSTM's accuracy in capturing extreme patterns, with statistically significant differences via t-test. The novelty lies in using a long-term (20-year) dataset for tropical Indonesia, demonstrating the hybrid model's efficacy for complex spatio-temporal predictions. Findings support flood early warning systems and water resource management, recommending additional climate variable integration for further development.]]>
Copyrights © 2025
