In this research, we present a novel approach for forecasting the Ultraviolet (UV) index by using the capabilities of the transformer-based deep learning models. The UV index is a prominent metric for assessing the risk of overexposure to UV radiation, which can have significant health implications especially for skin as skin cancer and eyes diseases. Traditional forecasting method is rely on statistical models or simple machine learning techniques, which may not capture the complex dynamical dependencies inherent in the UV index dataset. Our approach utilizes the transformer architecture, renowned for its prowess in handling sequential data and capturing long-range dependencies. We trained the model on a comprehensive dataset comprising historical UV Index values, meteorological data, and other relevant environmental factors. The model’s performance was evaluated against several baseline methods, demonstrating superiority in terms of root mean square error (RMSE), mean absolute error (MAE), correlation coefficient (r), and robustness in UV index predictions. The results indicate that the transformer model improves forecast precision and enhances interpret ability, allowing for better understanding of the contributing factors to UV variations. This advancement in forecasting technology provides a valuable tool for public health officials, environmental agencies, and the general public, facilitating informed decision-making and proactive measures to mitigate UV-related health risks.