The present paper outlines the extension of the novel
    paradigm of evolutionary support vector
    machines (ESVMs) to regression, in strong correspondence with the similar step that standard
    support vector machines (SVMs) had undertaken in the expansion of the technique from
    classification to regression. In particular, we consider the classical epsilon-support vector
    regression (epsilon-SVR) introduced by Vapnik as the learning component of the hybridization with evolutionary
    algorithms (EAs). Similarly to the application of ESVMs for classification, in epsilon-evolutionary support
    vector regression (epsilon-ESVR) we employ EAs in order to find, in a simple and directly resulting manner,
    the coefficients of the decision function. epsilon-ESVR is validated on the Boston housing benchmark
    regression problem as well as applied to a 2-dimensional set of points. Results complete the arguments
    in sustainment of the new application of ESVMs.