JIITA, Vol.10 no.3 pp.1370-1387 (2026), DOI: 10.22664/ISITA.10.3.1370
Soumya Yattinahalli , Vijaya Ramineni 2, Ramya M S and Dr Prakash Kuppuswamy
Abstract. The forecasting of rainfall is critical to the management of water resources, reducing disaster risks, and resiliency to climate change. The application of machine learning techniques to modeling complex meteorological relationships has shown promise; however, many existing studies use random data splits, accuracy-centered evaluations, and opaque black-box models that limit their usefulness in real-world
applications. Using Australian Weather AUS data, this study proposes a probabilistic, explainable, and temporal machine learning framework for rainfall prediction. The primary predictive model was an XGBoost classifier, with class imbalances handled through cost-sensitive learning rather than resampling.
Keywords; Rainfall Prediction; Temporal Evaluation; Explainable AI; Probability Calibration; Extreme Rainfall; XGBoost
Fullpaper: