AI-based feature selection techniques: can we predict geoeffective solar events?

In machine learning enhancing sparsity is a crucial issue. When working in a supervised setting, this aspect is frequently investigated by means of methods that are independent of the classifier used to perform the prediction with that reduced number of features. Some of the state-of-the-art methods include the Least Absolute Shrinkage and Selection Operator (LASSO), Recursive Feature Elimination (RFE) or Support Vector Machines (SVMs). Here we explore a new feature selection approach, which allows to perform an iterative selection of the most important feature according to selected classifier and performances score. The benefits of such scheme are theoretically investigated in terms of model capacity indicators that prove that this iterative algorithm is able to construct classifiers whose complexity capacity grows at each step.