<![CDATA[Flooding is one of the hydrometeorological disasters that frequently occurs in Indonesia and causes various social and economic losses. This study aims to compare the performance of five machine learning algorithms, namely Random Forest, Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Logistic Regression, as well as one Long Short-Term Memory (LSTM) deep learning model in predicting flood status based on water level data from seven observation posts in the DKI Jakarta area and its surroundings. The research stages include data preprocessing, handling unbalanced data using ADASYN, hyperparameter tuning, and evaluation using accuracy, precision, recall, and F1-score. To avoid data leakage, the data division process is carried out before preprocessing and oversampling. The results show that XGBoost produces the best performance with 96.0% accuracy, 95.5% precision, 96.9% recall, and 96.2% F1-score after hyperparameter tuning. The LSTM model also demonstrated competitive performance with an accuracy of 94.5% and an F1-score of 94.5%. Learning curve analysis showed that all models exhibited normal learning patterns with no indication of data leakage. The results indicate that XGBoost and LSTM have good potential for application in flood early warning systems based on water level data.]]>
Copyrights © 2026
