TY - THES ID - 137951196 TI - Forecasting of a Solar Wind Classification using Convolutional Neural Networks AU - Depypere, Gilles AU - Lapenta, Giovanni AU - Amaya, Jorge AU - Pucci, Francesco AU - KU Leuven. Faculteit Wetenschappen. Opleiding Master of Astronomy and Astrophysics (Leuven) PY - 2018 PB - Leuven KU Leuven. Faculteit Wetenschappen DB - UniCat UR - https://www.unicat.be/uniCat?func=search&query=sysid:137951196 AB - Almost everyone has heard about the polar lights, the beautiful events often visible close to the North- and South Pole. Countries where the polar lights are visible are valued holiday destinations, with Norway and Canada as examples. Few people know that polar lights are only a small part of a more complicated story. They are a consequence of what is called "Space Weather", which is also the name of the field that studies the effect of the Sun on the Earth. Next to light and heat, there are other ways in which the Sun influences the Earth. At every moment in time, particles escape the Sun and are set on a voyage in space. This effect is reinforced by very intense eruptions happening now and then on the Sun by which an enormous stream of particles is erupted and accelerated towards the Earth. Luckily the Earth has a shield that protects us from most of those energetic particles. Sometimes however, the events are so big that particles are able to enter the atmosphere. One of the harmless consequences of this are polar lights, a beautiful spectacle on the night sky. Not all the consequences are harmless, though. Energetic particles ejected during these explosive events can cause problems to airplanes and on Earth communications and seriously damage satellites. This can result in vast economic damage. In order to prevent these problems, similarly to what happens for weather forecasting on Earth, scientists are trying to forecast Space Weather. The goal by doing this is to obtain a timely warning about when these space storms will hit the Earth. This would allow us to take countermeasures to reduce the economic damage. Forecasting the normal weather is already hard but Space Weather is even harder, due to the complexity of the physical phenomena involved. The goal of this thesis is to create a part of a model that is able to help such forecasts. It makes use of new techniques from Artificial Intelligence, which make it possible to single out cause-and-effect structures that would be hard to distinguish otherwise. Different information and setups of these techniques are tested in order to obtain the best possible forecast. ER -