A recommendation engine involves connecting users to products/ items.
Lets take an example. Let’s assume we have ratings for products . In an e-commerces store these might be implicit signals such as view, add to cart, buy , remove from cart, time spent on a product and much more.
For simplicity we will take ratings on a scale of (1-5). C1,C2.. are customers and P1, P2 .. are products. C1 has rated P1 4, C2 has rated P2 5 etc . The recommendation engine’s goal is to predict what would customer 3’s rating would be for P2 using the given ratings(or any other blank cell in the matrix).
- Recommendations are classified according to how these blanks in this matrix are filled.
- Content based- using properties of items.
- Collaborative filtering (CF)- using similarity of items and/or similarity of users. The items recommended to the user.
- Neighborhood based approach
- Model based approach.
- Hybrid models.
- Hybrid – CF + content based approach.
Hope this gives a simple introduction into what recommender systems are all about, I will discuss in detail more of these methods in my future posts.