Predictive Analytics in Commerce

Chapters 4 › Unit 4: Blow us Away With Your Model! View instructions Hide instructions

Blow us Away With Your Model!

Step 1. Build the best possible model

Nienke has challenged you to build the best possible model, so let’s do that! In this link you can find the two functions, gainLiftChart and glmNextBestVariable, that help you create the best possible model.

Step 2. Visualise your model

Creating a model is important, but communicating the outcome of the model is also important. That’s why we want to challenge you to visualise the outcomes of the model on a poster to inform your colleagues about the profile of the customers with a high probability to buy. In this way they can set up a relevant action to approach these customers.

This poster does not necessarily need to be a piece of white paper, it can also be a powerpoint, a dashboard, whatever you feel comfortable with.

Share your visualisation with the community, let us know where you struggled the most and also what you enjoyed the most about this assignment!

Cross-Selling a product, creating the best possible model


enter image description here

I have based my model on not including more than 5 variables as well as trying to use some "common sense" to avoid variables that that might be included in the model more due to "generalness" rather than causality. HasPhone as an example of a variable that might show as a driver for the model but that doesn't seem to have real life implications on the selection of a travel insurance.

EDIT: I forgot (somehow) to include which variables ended up in my built model: Age, HomeActive, AccidentActive, FamilyStructureMarried, CarActive.

Your Comment

Please login to leave a comment.