How does DataRobot do regularization for its GA2M model?
It has been discussed a few times that regularization is the best practice for insurance pricing models. And understandably, GA2M is the recommended model if a rating table is needed. Curious how DataRobot handle regularization in a GA2M model?
Interesting question. I wasn't able to find any advanced tuning parameters for regularization, but looks like GAM blueprints have an ElasticNet model on residuals after the GAM model. The ElasticNet model has l2 regularization in it at least.
Let me know if you find a GAM BP without another model after the GAM one in it.