What is the difference in type of recommendation engines and how to decide which one to use?



I have just started on working on building a recommendation engine for a e-commerce company. I have no experience in building a recommendation engine till now. We have past buying and browsing history of all customers for last one year and want to use that to build a menaingful recommendation system.

From what I read, there are 2 types of recommender systems:

  1. Content based filtering
  2. Collaborative filtering

What are the key differences between the two and how do I decide which algorithm to use in our case?

Any help would be greatly appreciated.



Hi Oliver,
I think first we need to understand the difference b/w the two recommendation engines. So let’s take it one by one -

  1. Content based filtering : In this kind of recommendation system we try to predict user behavior or features based on his ratings or consumption for a product with specific features. Ex: If I have given good ratings to “Gone girl” and “silence of the lamb”, then i like more thriller and psychotic movies and if i have given bad ratings to “titanic”, “if only”, “PS: I love you” than that means I don’t like romantic movies.

So after we have predicted the user behavior we give him recommendations of the products which resemble more to the features he use to like.

  1. Collaborative Filtering : In this kind of recommendation system we try to find users who have given similar ratings to certain kind of products as the customer and based on that we recommend other products which is been rated high by those users.

There is a 3rd type also :
Hybrid Recommendation systems : These recommendation system use both the above techniques simultaneously to cope up with the data gaps.

Here are two links which you can refer for basic understanding of techniques -

Hope this helps.