Space weather, i.e. the conditions of the Sun, solar winds an Earth's atmospheric layers, may significantly affect life on Earth, with direct impact on technologies such as spaceships, satellites, planes, energy distribution networks, communication networks and others. Such impact is mainly negative when associated with severe solar flares, i.e. the release of large amounts of energy at the Sun's atmosphere in a short period of time. Therefore, several research groups focus their work on the development of mechanisms capable of forecasting the occurrence of such events, aiming at mitigating their effects on Earth. Among such works, there is no consensus about which features, observable from the Sun and the interplanetary medium, contribute the most to the quality of the forecasts. Thus, this project aims to implement an evolutionary algorithm to automatically identify and select a subset of features, among those adopted in the literature and publicly available, that leads to the smallest errors in the forecast of the occurrence of solar flares. This automatic feature selection mechanism will work together with Multilayer Perceptron (MLP)-based forecasters being developed by the research group to which this project is associated with (High Performance Intelligent Decision Systems - HighPIDS), according to the wrapper method.
News published in Agência FAPESP Newsletter about the scholarship: