This document discusses developing a solar flare forecasting model using satellite data and deep learning. It summarizes preprocessing satellite data to extract useful features, analyzing the topology of the solar magnetic field, and training a convolutional neural network called FlareNet. FlareNet uses images of the sun to predict solar flares and x-ray flux levels up to 1 hour in advance. By visualizing what areas of images FlareNet focuses on, researchers discovered it learned the importance of solar active regions in predicting flares. The project achieved dimensionality reduction of satellite data, a topological model of the solar magnetic field, a trained deep learning forecasting tool, and improved solar flare predictions.