Can we use XGBRegressor?

[hanzigs]

Hi,
The scorecard function param says "model: A LogisticRegression model object."

can we use any regression model. like XGBRegressor with 'gblinear' booster?

[ShichenXie]

Support sklearn lr model object only.

[hanzigs]

Thanks for the reply,

Actually any linear model with coefficient attribute is working including xgb linear, like

model.coef_

I have tried these

from xgboost import XGBRegressor
xgb_fit = XGBRegressor(booster='gblinear')
xgb_model = xgb_fit.fit(X_train, y_train)
xgb_model.coef_

from sklearn.linear_model import LogisticRegression
lr_fit = LogisticRegression()
lr_model = lr_fit.fit(X_train, y_train)
lr_model.coef_

from sklearn.linear_model import SGDRegressor
SGD_fit = SGDRegressor()
SGD_model = SGD_fit.fit(X_train, y_train)
SGD_model.coef_

All these models can generate scorecard using scorecard function

card = scorecard(train_bins, lr_model, training_fields)

Thanks

[autogluonisgod]

'TabularPredictor' object has no attribute 'coef_', I want t ues autogluon to remould this scordcard,but this error is happen. So how should I do to solve this ?

[autogluonisgod]

'TabularPredictor' object has no attribute 'coef_', I want t ues autogluon to remould this scordcard,but this error is happen. So how should I do to solve this ?

[ShichenXie]

A new function scorecard2 is introduced, which requires binning from woebin function and x variable only.