How much training data do you think k-NN will need to make good estimates for the rest? Try out different values for training_fraction and check your intuition.
0 comments