Utilizing target permutation to prevent overfitting.

The author describes a method to assess whether the results of a model are random or statistically significant by permuting the target vector in the training set and evaluating the results on the original target in the test set. The results showed that the non-permuted target vector outperformed most randomly permuted vectors, Show more