RFECV for classification giving KeyError: 'weight'

Viewed 928

Recursive feature elimination with cross-validation (RFECV) is not working, getting KeyError: 'weight' . As I see it is not able to calculate coefficients, hence the weights are missing. I have estimator as XGBClassifier.

1 Answers

RFECV is not supported with new version of XGBoost - 1.0.0, hence try running the same with older versions of XGBoost package.

Uninstall package: !pip uninstall xgboost --y

Install old package: !pip install xgboost==0.90 or 0.80

Also, RFECV will not work (take long time) with large datasets viz: column size >10000

Related