Hybrid Gradient Ensemble for Urban Flood Alerting Using XGBoost, LightGBM, and CatBoost Fusion

Flooding in the urban areas has become a great issue ever since urbanization and climate change came into force and therefore a need for a better flood modeling system. In this research, a new algorithm named as Hybrid Gradient Ensemble Model (HGEM) which is a combining model of XGBoost, LightGBM, and CatBoost is introduced to enhance the accuracy of the flood alerting system. The model under consideration aims to intercept the advantages of each gradient boosting algorithms, making the ensemble learning strategy to boost predictive accuracy and lower types of errors. The application of HGEM it was proved accurate and efficient with the accuracy of 94.6 % compared to that of the single models and Mean Absolute Error of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{0. 0 7 4}$</tex> and Root Mean Square Error of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{0. 1 5 1}$</tex>. It also equally remarkable for various levels of rainfall and various geographical regions to show that it is suitable for realtime flood notification. Nevertheless, predictions of the current model cost slightly more than estimates obtained from standalone models, which is reasonable for a more accurate predictor. The present study therefore implies the future possibilities of existence and applicability of ensemble learning techniques in urban flood prediction and management of disaster. For future work, deep learning integration and further optimization for real-time purposes may be done.