This research introduces a hybrid data-driven weather prediction system that enhances the accuracy and reliability of forecasts by integrating atmospheric numerical weather prediction (NWP) models with near-surface observational datasets. Traditional numerical models lack localized meteorological data, imprecise parameterization, and low spatial resolution. Utilizing state-of the-art machine learning techniques, the system incorporates numerical atmospheric model outputs with real-time near-surface meteorological data, including humidity, temperature, wind speed, and pressure. In order to more accurately represent intricate atmospheric patterns and local weather fluctuations, the hybrid approach employs physical simulations and data driven learning. This system is able to better adapt to new environmental conditions, reduce prediction errors, and enhance short-term forecasting by integrating historical data with model outputs. For weather predictions that are precise, practicable, and broadly applicable, we suggest the hybrid approach. Numerical atmospheric models and near-surface data enhance prediction capacity, as demonstrated by experiments.