remould this card

[autogluonisgod]

I want to use autogluon to remould scordcard,just like this card = sc.scorecard(bins, predictor, X_train.columns)

I use predictor to replace lr,but there is a error

'TabularPredictor' object has no attribute 'coef_'

how should I do?

[Blanket58]

The obvious reason is that TabularPredictor object doesn't have the attribute coef_, any other predictor which does have the attribute will work. But do we really can define or extract the coeffients from that nonlinear model? Is it theoretically possible? You shouldn't use any other model but lr to create a scorecard in most cases, since the result will become unexplainable.

[ShichenXie]

其他模型直接概率转评分呗，不一定需要转成评分卡的形式吧