Unlocking Enhanced Rainfall Prediction: Leveraging Stacking Classifier Ensembles for Accurate Forecasting and Real-World Applications

Accurate prediction of rainfall is crucial for ensuring flight safety and various other applications. This study explores the use of Stacking Classifier ensemble models to predict rainfall, focusing on the challenging 1- and 24-h scenario. The intricate nature of rainfall data demands the integration of diverse machine learning models to enhance prediction accuracy. Leveraging surface observation data from Soekarno-Hatta International Airport, approach utilizes various single models including Random Forest, Extra Trees, Adaptive Boost, and Gradient Boost as the base learners within the Stacking Classifier. The results show that the Stacking Classifier exhibits a compelling performance, outperforming in terms of accuracy. It achieves an accuracy of 0.85 and 0.95 in 1- and 24-h scenario on the test data, showcasing its potential for real-world application. This research also presents an in-depth analysis of the Stacking Classifier’s performance through various evaluation metrics including confusion matrices, ROC curves, precision–recall curves, classification reports, and learning curves. The Stacking Classifier demonstrates strong discriminative capabilities across classes and exhibits consistent learning progress with increasing training data. Furthermore, the successful implementation of the Stacking Classifier model is extended to a user-friendly web application and demonstrates the potential of ensemble learning in addressing complex meteorological prediction tasks. The results underscore the significance of utilizing advanced machine learning techniques for enhancing forecast accuracy and promoting their accessibility to a wider audience.