0

Normally when I create a claim frequency and severity model at work, I use glm. But now I want to try to create one frequency and one severity model using the xgboost-package in R.

I do the analysis for deductible separately. And the factors I obtain from the analysis, I set as offset in the frequency model. Eg say that the variable for deductible has six classes: 100, 200, 300, 500, 700, 1000. And the factors from the analysis gives med that factors 1, 0.92, 0.85, 0.77, 0.70, 0.64. These are the factors I set as offset in the glm frequency model. Very few policyholders have the higher alternatives for the deductible

My question is, how do I set these factors as an offset in the frequency model using the xgboost-package? If a customer is checking the price for an insurance (all things equal), then the premium for the insurance has to decrease with increasing deductible.

If it is not possible to set a variable as an offset the way I want to, do you have any suggestions how I could work around this problem?

Thanks!

aleket
  • 21
  • 3
  • you can check out https://stats.stackexchange.com/questions/223939/poisson-xgboost-with-exposure, https://stackoverflow.com/questions/34896004/xgboost-offset-exposure – StupidWolf Apr 02 '20 at 10:15
  • you can always check again by using a model with the frequency – StupidWolf Apr 02 '20 at 10:16

0 Answers0