This project aims at using an SVM (in this case with rbf kernel) to predict if the email described by the data given (more infos in spambase.names and spambase.DOCUMENTATION) are spam mails or not.
The process involves performing a GridSearch on gamma and C parameters of the SVM classifier, using a 3Fold cross-validation.
After that the learning curve of the algorithm with optimal parameters is plotted (RBF_learning_curve.png).
At last the model is trained over the full training set and testing it on the test set. The final accuracy is 0.93.