User Ben Reiniger

21 votes

Accepted

L1 & L2 Regularization in Light GBM

answered Aug 12, 2019 at 18:30

15 votes

What would I prefer - an over-fitted model or a less accurate model?

answered Jan 13, 2020 at 2:54

13 votes

Accepted

Using a random forest, would a RandomForest performance be less if I drop the first or the last tree?

answered Oct 5, 2020 at 14:03

11 votes

Does scikit-learn have a forward selection/stepwise regression algorithm?

answered Dec 29, 2020 at 16:07

10 votes

Accepted

Is There a Way to Re-Calibrate Predicted Probabilities After Using Class Weights?

answered Sep 9, 2019 at 11:50

9 votes

Why decision tree needs categorical variable to be encoded?

answered May 17, 2019 at 3:10

9 votes

Accepted

Can Boosted Trees predict below the minimum value of the training label?

answered Jul 7, 2020 at 1:46

9 votes

Why GridSearchCV returns nan?

answered Mar 27, 2021 at 3:53

8 votes

Accepted

Why use fit when already have fit_transform?

answered Apr 12, 2021 at 14:28

8 votes

What is the proper way to use early stopping with cross-validation?

answered May 17, 2020 at 18:30

8 votes

Decision tree with final decision being a linear regression

answered Dec 29, 2019 at 17:04

7 votes

Accepted

ROC AUC score is much less than average cross validation score

answered Feb 10, 2020 at 13:15

7 votes

Accepted

What happens if at leaf node both classes have same number of samples?

answered Jun 18, 2020 at 15:03

7 votes

Understanding predict_proba from MultiOutputClassifier

answered Jul 16, 2019 at 2:50

7 votes

How do we standardize arrays with NaN?

answered Jul 31, 2019 at 16:49

7 votes

Accepted

What makes a Tree-Structured Parzen Estimator "tree-structured?"

answered Jan 17, 2019 at 20:26

7 votes

Accepted

Default value of learning rate in adam optimizer - Keras

answered May 3, 2019 at 3:24

7 votes

Accepted

How to extract trees in XGBoost?

answered Aug 21, 2019 at 18:38

6 votes

Accepted

Can ridge regression be used for feature selection?

answered Aug 16, 2019 at 15:09

6 votes

Accepted

Probability calibration is worsening my model performance

answered Aug 7, 2019 at 18:56

6 votes

Why is there a difference between predicting on Validation set and Test set?

answered Aug 25, 2019 at 1:09

6 votes

Accepted

Why does it not need to set test group when using 'rank:pairwise' in xgboost?

answered Apr 28, 2019 at 16:19

5 votes

Accepted

How to do k-folds in python whilst splitting into 3 sets?

answered Feb 19, 2019 at 4:10

5 votes

xgboost: Is there a way to perform regression on rates/percentages data?

answered Aug 6, 2019 at 14:29

5 votes

Accepted

What is the difference between a regular Linear Regression model and xgboost with objective set to "reg:linear"?

answered Dec 13, 2019 at 22:46

5 votes

handling missing values for LightGBM model

answered Sep 17, 2020 at 15:41

5 votes

Accepted

Logistic regression does cannot converge without poor model performance

answered Jul 16, 2020 at 16:09

5 votes

Accepted

What does the descision boundary of a relu look like?

answered Nov 26, 2020 at 4:04

5 votes

Accepted

Several independent variables based on the same underlying data

answered Jan 30, 2020 at 1:03

5 votes

What's the default Scorer in Sci-kit learn's GridSearchCV?

answered May 10, 2021 at 15:16

Stack Exchange Network

448 Answers

L1 & L2 Regularization in Light GBM

What would I prefer - an over-fitted model or a less accurate model?

Using a random forest, would a RandomForest performance be less if I drop the first or the last tree?

Does scikit-learn have a forward selection/stepwise regression algorithm?

Is There a Way to Re-Calibrate Predicted Probabilities After Using Class Weights?

Why decision tree needs categorical variable to be encoded?

Can Boosted Trees predict below the minimum value of the training label?

Why GridSearchCV returns nan?

Why use fit when already have fit_transform?

What is the proper way to use early stopping with cross-validation?

Decision tree with final decision being a linear regression

ROC AUC score is much less than average cross validation score

What happens if at leaf node both classes have same number of samples?

Understanding predict_proba from MultiOutputClassifier

How do we standardize arrays with NaN?

What makes a Tree-Structured Parzen Estimator "tree-structured?"

Default value of learning rate in adam optimizer - Keras

How to extract trees in XGBoost?

Can ridge regression be used for feature selection?

Probability calibration is worsening my model performance

Why is there a difference between predicting on Validation set and Test set?

Why does it not need to set test group when using 'rank:pairwise' in xgboost?

How to do k-folds in python whilst splitting into 3 sets?

xgboost: Is there a way to perform regression on rates/percentages data?

What is the difference between a regular Linear Regression model and xgboost with objective set to "reg:linear"?

handling missing values for LightGBM model

Logistic regression does cannot converge without poor model performance

What does the descision boundary of a relu look like?

Several independent variables based on the same underlying data

What's the default Scorer in Sci-kit learn's GridSearchCV?