Customizing Collaborative Filtering for Product Affinity

Question

I'm trying to build a recommendation system and I am trying to use Collaborative Filtering (please let me know if other models fit better for my use case).

My Data:

My data is for an e-commerce site where I have some data on Transactions. I have another table which gives me some product features.

Transaction Table includes: CustomerID, ProductID, OrderID, Units Bought, Revenue, Purchase Country

Issue

Most Collaborative Filtering articles/examples I've seen use Product Ratings and Time (Purchase/Transaction Timestamp) and I don't have these 2 features available.

Question & Goal

Can I use a normal CF model when I don't have product ratings? How would I implement it?

Also, does the model just utilize customer preferences individually or does it also account for items that were purchased together (can be determined with orderID). If not, how do I make a model that uses that information as well?

Answer 1

• About product rating

  There are two types of feedback (rating) in CF, one is Explicit, and another is Implicit

  • Explicit: users specify how much they liked
  • Implicit: if a user by a product, the system infers that the user is interested.

• About detailed feature for model input

  Using e-commerce as an example, CF model use transaction record as input data, and generate a feedback matrix (user x item), then calculate similarity. ➡️ Individually buy or purchased together have no difference for the feedback matrix (There is no time series concept in CF model)