I'm trying to look for an idea to create a predictive model having the following data:

Customer_ID - Integer
Catalog_ID - Integer
Country_Code - Integet
Year - Integet
Month - Integer
Day - Integer
Quantity_Purchased - Integer
Product_Purchased - Double

I'm trying to look for a use case that creates a predictive model than can give me the ability to propose a product for a customer the next time that he comes to my website.

Is this a collaborative filtering use case? If yes, I only can use the last two fields of my dataset, right?


  • $\begingroup$ Is there anything in the Product_Purchased field that can be parsed to create product groupings? This can be either an embedded hierarchy or you can cluster similar fields by actual purchases. $\endgroup$ – Paul Dec 13 '16 at 19:54

Recommender systems are usually either collaborative ("other people who bought x also bought y") or content-based ("item y has similar properties as item x"). Given your very few data features, collaborative filtering seems appropriate. In the basic setting it could use only the purchases of your users. Ideally, you would also have some sort of rating of their purchases.

Also, nothing stops you from using a hybrid approach with additional data if available.

  • $\begingroup$ many thanks for your response! :) I was thinking using Collaborative Filtering to predicts something like "Customer ID 10 will buy in Catalog_ID 3 in next week"... like a forecasting time-series. What dou think? $\endgroup$ – Pedro_Rodgers Dec 13 '16 at 20:27
  • $\begingroup$ I thought your goal was to "propose a product". For your time series approach you would need tons of data. If you have a lot of catalogs this seems very difficult. Maybe you can update your question with more context around your business problem. How many of your customers return every week, how many products do you have etc.? Otherwise there is no one-fits-all solution... $\endgroup$ – oW_ Dec 13 '16 at 20:59
  • $\begingroup$ yes and it is my goal ;) I was just asking if with this datset I can think to create a forecasting time series analysis ;) $\endgroup$ – Pedro_Rodgers Dec 13 '16 at 21:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.