Machine learning (ML) has played a major role in enhancing weather forecasting systems in terms of predictive performance and reliability. The traditional numerical weather prediction techniques are limited to incomplete information, nonlinear atmospheric interactions and the challenges of modeling complicated meteorological patterns. These constraints have led to studies of hybrid ML systems that integrate preprocessing, feature engineering and ensemble learning algorithms [1], [4]. This review will discuss weather prediction hybrid models such as Linear Regression, K- Nearest Neighbors (KNN), XGBoost, CatBoost, AdaBoost, and LightGBM. Wavelet packet denoising and ensemble are used to increase the accuracy and robustness of prediction [1]. The existing body of literature shows that there has been significant advancement in ML-based forecasting [18], but there are issues related to scalability, interpretability, and real-time implementation. Hybrid ML models always achieve better accuracy, stability, and flexibility compared to single ones. Recent advancements in machine learning and the increasing availability of meteorological data have enabled the development of intelligent forecasting systems that support accurate, timely, and reliable decision-making across weather-sensitive domains. This work has critically examined the literature available, compared the methods, found deficiencies and future directions on how to come up with scalable and intelligent weather forecasting systems.

Index Terms—Weather Forecasting, Machine Learning, En-semble Learning, CatBoost, Gradient Boosting, Hybrid Models, Weather Prediction.