MUSIC RECOMMENDATION SYSTEM

In this case study, I built recommendation systems using four different algorithms. They are as follows:

• rank-based using averages

• User-user-similarity-based collaborative filtering

• Item-item-similarity-based collaborative filtering

• model-based (matrix factorization) collaborative filtering

We have seen how they are different from each other and what kind of data is needed to build each of these recommendation systems. We can further combine all the recommendation techniques we have seen.

To demonstrate "user-user-similarity-based collaborative filtering", "item-item-similarity-based collaborative filtering", and "model-based (matrix factorization) collaborative filtering", surprise library has been demonstrated. For these algorithms grid search cross-validation is used to find the best working model, and using that the corresponding predictions are made.