Recommendation systems with Implicit feedback

Question

I am new to recommender systems and I am trying to build a recommender system based on the articles data. Where we have User, Article ID, Content, action (open, comment, share), timeOfThe Action.

In order to start I have created a baseline scoring system where. {open:1,comment:3,share:4}. Now I would like to add multiple implicit information to my rating system. For example: Time of the day: If action takes place at night its very interesting for the user Conent: User really likes articles about cars (most of the article interactions where about cars) and so on.

How can I 'add' these information to my baseline model. Do I just create the same scorring system and add to the baseline score? I would really appreciate if you could share your thoughts.

Thank you!

Answer 1

I would highly recommend you to use the implicit library.

You can check out the docs here Implicit ALS

For the implicit library, you would only need the base line scoring which you have created and for the certain features that your data has, the ALS will try to find out all the features(factors) on its own. You can easily specify the number factors you wish your model to learn from your dataset.

And for using Implicit library, a rich dataset is required.