Artists Recommendation System

Combining Neural Collaborative Recommendations and Content Based Recommendations

An artists recommendation system that combines:

1. Neural Collaborative filtering that extends the collaborative filtering matrix factorization technique to a deep Neural Net.
2. Content Based recommendations that help understand how users are making decisions.

There are four main steps I took to build the final model:

1. Database: Using EMI Music Kaggle competition \cite{kaggle}, I pre-processed a ratings matrix for the recommendation task.
2. Model Design: Combined Neural Pytorch model and content based model to generate top 5 unseen artist to any user.
3. Evaluation: loss and accuracy metrics on validation and test sets.
4. Interpretation: Understanding what the model is learning by comparing content based recommendations on given artists features versus Neural model artist embedding matrix.