Ensemble Deep Learning for Forecasting 222Rn Radiation Level at Canfranc Underground Laboratory

Ensemble Deep Learning Architectures have demonstrated to improve the performance in comparison with the individual architectures composing the ensemble. In the current work, an ensemble of variants of Convolutional and Recurrent Neural Networks architectures are applied to the prediction of the Rn-222 level at the Canfranc Underground Laboratory (Spain). To predict the low-level periods allows appropriately scheduling the maintenance operations in the experiments hosted in the laboratory. As a consequence of the application of Ensemble Deep Learning, an improvement of the forecasting capacity is stated. Furthermore, the learned lessons from this work can be extrapolated to other underground laboratories around the world.