I recently developed a cross sell application that took product purchase history and flagged the record with a 1 for purchased, and 0 if not, within orders.
I then used Multinomial Logistic Regression to assign new orders to the cluster.
I found that at a certain point, the classification did not work with too many clusters, but was flawless with fewer clusters @ 100% correct classificatio…
You can share this discussion in two ways…
Share this link:
Send it with your computer's email program: Email this