Achieving high accuracy for Global Navigation Satellite Systems (GNSS) positioning at low cost, is the main reason for the wide acceptance that has received the technique of Precise Point Positioning (PPP) from the geodetic community. However, the long convergence time required to achieve centimeter-level accuracy in positioning, remains of PPP’s main disadvantage. Our proposed method is aiming to provide accurate prediction of ionosphere variations at regional level, in a form suitable to be applied in PPP processing as external ionospheric information, so that to successfully deal with the ionospheric error effects. Machine Learning regression-based approaches for sequence modeling are suitable for predicting the ionospheric variability. Gaussian Process Regression (GPR) and Support Vector Regression (SVR) are introduced for ionospheric variations modeling at different locations of the International GNSS Service (IGS) network. The proposed algorithms predict the future electronic content per satellite from a specific station, fitting a non-linear model. The evaluation of our proposed methods for vertical total electron content (VTEC) values prediction is compared against the traditional Autoregressive (AR) and Autoregressive Moving Average (ARMA) methods, per satellite. Additional, mean TEC timeseries are compared against classical Global Ionospheric Maps (GIM), NeQuick and International Reference Ionosphere (IRI) models.