An AutoML Approach Integrated with Live Weather Data in Rain Forecasting System (RFS)

The increasing climate unpredictability highlights the critical need for accurate rain prediction methods. Traditional approaches often struggle to adapt and accurately forecast rain across varied climatic conditions. This research work investigates the fusion of automated machine learning (AutoML) with real-time weather data to tackle these challenges. Previous methodologies have relied on manual algorithm selection and optimization, leading to inefficiencies and suboptimal outcomes. In contrast, the proposed AutoML-based approach automates algorithm selection and optimization, resulting in heightened accuracy and efficiency. Integration of diverse meteorological data streams and advanced preprocessing techniques enhances data quality for model training. Evaluation against traditional methods showcases significant enhancements in rain prediction accuracy, with scores ranging from 80.60 to 86.09% across distinct algorithms. Specifically, the Random Forest Classifier achieved an accuracy score of 85.43% and an AUC score of 0.88, Gaussian NB achieved 80.60% accuracy and 0.82 AUC, K Neighbors Classifier achieved 82.89% accuracy and 0.82 AUC, and XGBoost achieved 86.09% accuracy and 0.89 AUC. The developed approach not only advances rain forecasting but also offers potential applications in agricultural planning, urban water management, and disaster preparedness. Future research avenues may explore further enhancements to the AutoML framework and its extension to other environmental prediction domains.