- Unarchived data and put it as
data/directory on the root. - Run
$ automl searchto search best options using Optuna. - Run
$ automl predict --features 32 --category-encoding label --fold 4to make submission file.
See $ automl --help for more details
$ automl predict --help
Usage: automl predict [OPTIONS]
Train and predict using LightGBM.
Options:
--features INTEGER The number of features.
--featuretools / --no-featuretools
Synthesis features using feature tools if
True
--category-encoding [label|ohe]
--fold INTEGER The number of LightGBM model learned from N
folded dataset.
--save-model / --no-save-model Save models in output directory, if true.
--save-feature-importance / --no-save-feature-importance
Save image of feature importance, if true.
--help Show this message and exit.