Closed jackvfb closed 5 days ago
I think rather than make a branch for different arbitrary choices of K, I should so I more formal train/test split and validation to optimize K.
I can holdout some years, say the last three years and do kfold cross validation on the training data, then get metrics from the test set.
I think rather than make a branch for different arbitrary choices of K, I should so I more formal train/test split and validation to optimize K.