<![CDATA[Global climate change and the increasing frequency of extreme weather events demand more accurate and adaptive weather prediction systems. This study aims to implement and compare three deep learning models, Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN)+LSTM Hybrid, and Transformer for predicting next-day weather events using daily multivariate meteorological data. The dataset was obtained from the Climatology Station Class IV Lampung and includes air temperature, rainfall, humidity, solar radiation, air pressure, wind direction, and wind speed, collected in CSV format from February 2000 to March 2025. The analysis results indicate that the CNN+LSTM Hybrid model achieved the best performance, with an RMSE of 1.158, MAE of 0.521, R² Score of 0.323, accuracy of 75%, and Macro F1 score of 0.75. The LSTM model demonstrated moderate performance, while the Transformer model yielded the lowest results among the three. These findings suggest that combining CNN's spatial feature extraction with LSTM's sequential processing enhances the prediction quality of short-term weather forecasts based on multivariate data. This study is expected to contribute to the development of AI-based weather forecasting systems in Indonesia, particularly for hydrometeorological disaster mitigation.]]>
Copyrights © 2025
