--feature_selection

Switch

--feature_selection NAME (or --feat_selection NAME)

Description

Select feature selection pipeline used in prediction

Argument and Default Value

There is no default value, whatever is uncommented in regressionPredictor.py will be used. To override this you can use this flag with:

Details

Other Switches

Required Switches:

• -d, -g, -t

• -f

• --outcome_table, --outcomes

Optional Switches:

• --nfold_test_regression

• --control_adjust_outcomes_regression

• fwflag_control_adjust_reg

• fwflag_predict_cv_to_feats

• fwflag_predict_combo_to_feats

• fwflag_predict_regression_all_to_feats

• --nfold_test_classifiers

Example Commands

# train a language model for age
dlatkInterface.py -d dla_tutorial -t msgs_xxx -c user_id -f 'feat$cat_met_a30_2000_cp_w$msgs_xxx$user_id$16to16' 'feat$1gram$msgs_xxx$user_id$16to16' --outcome_table masterstats_r500 --group_freq_thresh 500 --outcomes demog_age --train_regression --model ridgefirstpasscv --feat_selection none