Multi-dimensional Time Series Modeling of Ionospheric foF2

A multi-dimensional time series method for modeling ionospheric foF2 is developed using the scikit-learn technique. Knowledge about the character of the foF2 is used to reduce the complexity of machine learning methods. Then, Linear regression, Bayesian ridge regression, Elastic net and SVM are chosen to predict foF2. Compared with measurements data and the conventional models, the proposed approach shows its ability to model foF2 and can greatly improve the prediction accuracy.