Machine Learning to Predict Climate Change in Coastal Areas of Indonesia

Abstract

Indonesia's coastal regions face significant threats from climate change, including rainfall uncertainty, rising temperatures, and sea level rise. This study aims to explore the potential of machine learning algorithms in predicting climate parameter changes in the coastal areas of Minangkabau, Pesawaran, and Maritim Panjang. Daily climatological data obtained from the Meteorology, Climatology, and Geophysics Agency (BMKG) were used as the basis for model training. Three primary algorithms were tested Random Forest, XGBoost, and Long Short-Term Memory (LSTM) selected for their capability to handle complex and temporal data. The research methodology included data preprocessing, model training, cross-validation, and predictive performance evaluation using metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). Preliminary results show that LSTM excels in time series prediction, while XGBoost offers a good balance between speed and accuracy. These findings indicate that machine learning-based approaches have strong potential as decision-support tools for climate change mitigation and adaptation planning in Indonesia’s coastal regions.