I am trying to use select_features
from CatboostRegressor
but I am getting an error that it does not support the sample_weight
parameter. I need this to input exposures/offset for a Poisson loss model. I just want to make sure I am not doing anything incorrectly before I will probably have to submit a feature request or implement it on my own. Here is the code I'm using:
model = CatBoostRegressor(cat_features=feature_names,
task_type=args.task_type,
loss_function='Poisson',
use_best_model=True,
custom_metric=['Poisson','RMSE'],
**hyperparams)
feature_dict = model.select_features(X=train_x, y=train_freq, eval_set=test_pool, features_for_select=feature_names,
num_features_to_select=args.num_features, plot_file='training_plot.html',
sample_weight=train_exposure)
# model.fit(X=train_x, y=train_freq, eval_set=test_pool, plot_file='training_plot.html', sample_weight=train_exposure)