I have been working at a new company for around 4 months now as a statistical modeling analyst, and the existing data science manager and modeling analysts are very familiar with regression techniques, but not newer techniques such as Random Forest and XGBoost. I have built some models for them using these techniques which have shown good results, but they are concerned with the interpret-ability of the models for business use. They want to be able to explain what is happening within the models so that we can explain to our business units what the model is doing.

I am familiar with Importance in Random Forest, and have messed a little bit with importance in XGBoost, and the model plotting in XGBoost, but I was wondering if any of you had ways of getting more information out of these models in order to better explain what is happening and how many a certain variable might be impacting the model.

I know that these models are somewhat blackboxish as to exactly what is happening, but if I could extract as much information as possible from them then I can make a good case for us as a team to start transitioning to using these models more as a team.

I appreciate any advice that anyone can give.



Browse other questions tagged or ask your own question.