Indonesia’s tropical climate, marked by rainy and dry seasons, is increasingly affected by extreme weather events driven by climate change. Rising temperatures, shifting rainfall patterns, and sea-level rise have intensified health risks such as malaria, dengue hemorrhagic fever (DHF), and gastrointestinal infections. Accurate weather forecasting is essential for mitigating these challenges and informing risk management strategies. This study develops and evaluates a hybrid SARIMA-LSTM model for weather forecasting in Bandar Lampung, integrating time series analysis with deep learning to enhance predictive accuracy. SARIMA captures seasonal variations, while LSTM models nonlinear relationships, offering a robust approach to forecasting complex weather patterns. The SARIMA (6,1,0)(3,1,0)26 model was selected for its effective seasonal representation and combined with LSTM to leverage its capability in modelling nonlinear dependencies. Hyperparameter optimization using grid search further improved model performance. Two data partitioning approaches were tested: 70%-30% and 80%-20% splits for training and testing, respectively. The SARIMA-LSTM hybrid model demonstrated superior performance with the 80%-20% split, achieving MSE, RMSE, and MAPE values of 0.1174, 0.3426, and 0.0104%, respectively. The model accurately forecasted weather conditions over 21 weeks, aligning closely with observed trends and effectively capturing seasonal patterns. These findings underscore the model’s potential to support public health strategies, including disease outbreak mitigation for malaria and DHF, and enhance disaster preparedness in flood-prone areas.
Kurniasari et al. (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: