Data Augmentation for Rainfall Classification in Bogor and Palu Using SMOTE

Weather predictions provide detailed and accurate information about weather conditions, including cloud cover, rainfall, temperature, humidity, wind conditions, and sun exposure. Indonesia, a tropical country with only rainy and dry seasons, experiences relative variations in these seasons. Due to global warming, rainfall in Indonesia has become increasingly unpredictable. This study addresses the class imbalance in rainfall data by using data from Bogor, which has the highest rainfall, and Palu, which has the lowest rainfall in Indonesia. The aim is to develop a robust classification method for imbalanced data using the Synthetic Minority Over-sampling Technique (SMOTE). We combined datasets from Bogor and Palu to create a balanced baseline dataset. We then evaluated four classification methods—Naïve Bayes, Logistic Regression, Support Vector Machine (SVM), and Kernel Support Vector Machine (KSVM)—using precision, recall, and F1-score as evaluation metrics. Our results indicate that KSVM outperformed the other methods in terms of robustness to class imbalance with F1-score of 48.6%, 67.1%, 67.4%, 71.3% for the dataset of Bogor, Bogor with SMOTE, Palu, and Palu with SMOTE respectively. The application of SMOTE improved the F1-score by 38% for the Bogor dataset and 6% for the Palu dataset.