I'm setting up a new advertisement recommendation engine which will show some textual message to the customer. How do I select the message which is most suitable to the customer?
What we have done till now is come up with a classifier which gave us the segment of customers who have a high score for clicking on the Advertisement. The way we did it was by setting up a logistic classifier and training it to score customers based on their propensity of clicking on a given Advertisement. The model is using all the features that we could collect about the customers in general and is not specific to any particular Advertisement. We can do a feature selection her but the current prediction accuracy (AUC) is good for us.
Let's say we have a set of 10 messages that we want to show to our customers. Now, is there any way that we could identify the message that we should be showing to the customer based on the feature values that the given customer has? If there is a way, it'd be great if I can get some pointers on the pros & cons of using one over the other.