Hi,
I'm trying to fit a logistic regression model regularized by the L1 norm. I am using the glmnet package in R. I have a dataset of a 100k records with roughly 60 regressors (binary response). Here is my method: 1. Cut the original dataset into a test dataset and a train dataset (stratified). 2. Run glmnet with 10-fold cross validation on the train dataset. 3. Fit on the test dataset using the predicted model from step 2 and calculate the mean square error of prediction.
Is this the correct approach? Any help would be much appreciated.
[link] [10 comments]
