Add evaluation methods

Currently only the random subsampling (without training on the entire dataset) is used. That is there is just training and testing datasets, no validation dataset.

Use other non-exaustive methods like random subsampling, k-fold cross evaluation to the model.