Solar radiation directly affects human health and the surrounding environment. Therefore, scientists are paying much attention to this aspect to control the level of radiation. This paper introduces a new model to predict solar radiation using the collected dataset. Our approach focuses on predicting solar radiation frequency with a deep-learning network model. Instead of ideas directly indicating the outcome with one regression model (deep learning or machine learning), we take inspiration from the saying "divide and conquer" to propose a layered learning model. We implement classification models before building local regression models for classes. Our proposal obtains the expected results with 99% accuracy for the classification and an MAE of 17.8556 for the regression model. In this paper, we also compare our approach with existing models. Two highlights are: (1) our model is better than several approaches, and (2) it forecasts the ability of solar radiation in the next fifteen minutes based on the current information/data.