I was taking this course of yours. I am facing some issues with Datasets in Activities Videos of Section “Evaluating Recommender Systems”. The sets you are using has only rating, user id and movie id. If i want to create data set with genre, director, actor ids. How to build that Datasets from Surpise?
Recommender systems typically only work with behavior data, not on content attributes. As such supriselib only deals with ratings.
If you want to build a system that predicts a rating based on features such as director, genre, etc., that lends itself more to traditional machine learning techniques such as deep learning, XGBoost, etc. Recommender systems are build to handle the sparsity of ratings data, but content attributes are not sparse and so don’t need recommender systems.
A little later in the course however, we’ll talk about blending the two – there’s a whole section on “Content-Based Filtering” that I think will clarify things for you.
Hello Sir,
I was taking this course of yours. I am facing some issues with Datasets in Activities Videos of Section “Evaluating Recommender Systems”. The sets you are using has only rating, user id and movie id. If i want to create data set with genre, director, actor ids. How to build that Datasets from Surpise?
Recommender systems typically only work with behavior data, not on content attributes. As such supriselib only deals with ratings.
If you want to build a system that predicts a rating based on features such as director, genre, etc., that lends itself more to traditional machine learning techniques such as deep learning, XGBoost, etc. Recommender systems are build to handle the sparsity of ratings data, but content attributes are not sparse and so don’t need recommender systems.
A little later in the course however, we’ll talk about blending the two – there’s a whole section on “Content-Based Filtering” that I think will clarify things for you.