--folds

Switch

--folds

Description

Number of folds for functions that run n-fold cross-validation.

Argument and Default Value

Argument is integer representing number of folds. Default number of folds is 5.

Details

The original sample is randomly partitioned into n subsamples, n-1 of which are used for training. This is repeated n times. The n trained models are then combined into a single model.

Other Switches

Required Switches:

• -d, -g, -t

• -f

• --outcome_table, --outcomes

Optional Switches:

• --nfold_test_regression

• --control_adjust_outcomes_regression

• --control_adjust_reg

• --predict_cv_to_feats

• --predict_combo_to_feats

• --predict_regression_all_to_feats

• --nfold_test_classifiers

Example Commands

# Runs 10:doc:`fwflag_fold` cross validation on predicting the users' genders from 1grams.
# This essentially will tell you how well your model & features do at predicting gender.
# Splits the data in 10 chunks, for each chunk training a model on the remaining 9 chunks.
dlatkInterface.py -d fb20 -t messages_en -c user_id -f 'feat$1gram$messages_en$user_id$16to16$0_01' --outcome_table masterstats_andy_r10k --outcomes gender --combo_test_classifiers --model linear-svc --folds 10