<![CDATA[Global warming and climate change have increased the frequency and intensity of extreme weather events, significantly impacting human life and the environment. Urban areas such as Kecamatan Panjang in Bandar Lampung City frequently experience flooding due to extreme rainfall and poor drainage systems. This study compares the effectiveness of three deep learning model architectures- Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Transformers — in predicting rainfall events in Kecamatan Panjang. The data used includes key meteorological variables such as air temperature, dew point, humidity, and air pressure, collected from the Maritime Meteorology Station in Panjang (BMKG) over the past three years. The models were trained using this historical data, with the data divided into training and testing sets. The study results indicate that the Transformer model performs best with the highest accuracy compared to CNN and RNN. The Transformer model efficiently captures long-term dependencies in sequential data, providing more accurate and timely predictions. Model performance evaluation was conducted using accuracy, F1 score, precision, recall, ROC AUC, RMSE, and MAE metrics. The use of deep learning models in rainfall prediction is expected to assist in flood risk mitigation and planning for adaptation to increasingly frequent extreme weather due to climate change. This research significantly advances more accurate and efficient weather prediction systems for urban areas prone to hydrological disasters.]]>
Copyrights © 2025
