Improving Weather Forecast Accuracy Using Hybrid Machine Learning Algorithms

Weather prediction, particularly for rainfall, temperature and relative humidity (RH%), is critical for climate-sensitive industries such as agriculture and disaster relief. This paper introduces a predictive modeling framework based on a range of machine learning (ML) methods. In addition, hybrid models such as Catboost-Logistic Regression, XGboost-k-Nearest Neighbors (KNN) and Gaussian ProcessDecision Tree have been investigated to improve prediction accuracy. Using a 10-year Gannavaram dataset, we focus on multi-label classifications (2 to 5 labels) of weather characteristics such as rainfall, temperature, and RH%. Notably, the Gaussian Process consistently predicted rainfall with 100% accuracy, while hybrid models such as CatboostLogistic Regression and XGBoost-KNN performed well across a variety of criteria. The combination of these hybrid models and standalone ML algorithms has considerably increased the resilience of weather forecasting. Our research highlights the effectiveness of combining machine learning models to improve predictive accuracy, offering a valuable contribution to realtime weather prediction systems.