Development and Evaluation of an LSTM-Based Rainfall Prediction System at the Class I Sultan Iskandar Muda Meteorological Station, Banda Aceh

Abstract

Rainfall is a key meteorological variable that influences agriculture, water resource management, transportation, and hydrometeorological disaster risk reduction. This study analyses historical rainfall patterns and evaluates the performance of a Long Short-Term Memory (LSTM) model for monthly rainfall prediction at the Class I Sultan Iskandar Muda Meteorological Station in Banda Aceh. Meteorological observation data were processed through several stages, including data cleaning, missing-value treatment, normalisation, model training, and performance evaluation. Model accuracy was assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results show that the LSTM model captures the general temporal pattern of rainfall but produces substantial deviations during periods of exceptionally low or high rainfall. A web-based dashboard was also developed to support meteorological data management, analysis, visualisation, and the presentation of prediction results. The dashboard provides statistical summaries, monthly rainfall trends, rainfall-intensity distributions, and other supporting features that enable efficient monitoring and interpretation of rainfall conditions. These findings contribute to the development of data-driven rainfall prediction systems and provide a basis for supporting meteorological decision-making and hydrometeorological disaster risk reduction in Banda Aceh.