Solar proton events prediction with support vector machine

In this paper the support vector machine (SVM) was applied to model solar proton events. The inputs of the model include area of sunspot group, magnetic class, Macintosh class, solar radio flux and soft x-ray flux, all of which are valued by calculating their solar flare or proton event occurrence rate. Proton occurrence prediction model was created by classifying these inputs as relating to the occurrence or non occurrence. Models were verified by two years data and good prediction rates were demonstrated. A comparison was done between the prediction result of proton occurrence model and that of word warning agency (WWA). It is shown that the model performance can be comparable to that of WWA.