All learning used the same testinf dataset et learning dataset. Better to compare performance of hyper-parameters

K fold

See merge request !3

runKfold return HashMap<AbstractSolution, Double> that contains the solution object and the AUC value