A movie recommendation engine suggests what to watch next. Instead of hand-written rules, it learns from data: either from the movies themselves (genres, keywords, cast, descriptions) or from the ratings and viewing habits of many users. This is the same idea behind the suggestions you see on Netflix, YouTube and Spotify.
The lessons start with the intuition behind collaborative filtering. Next you build a content-based recommender on the TMDB 5000 movie dataset using TF-IDF and cosine similarity, then a collaborative filtering recommender from user ratings. The last lesson goes one step further with matrix factorization in PyTorch, the technique made famous by the Netflix Prize.
Watch the lessons in order and code along in Google Colab or a Jupyter notebook using Python, pandas and scikit-learn. Once it works, try wrapping your recommender in a small Streamlit app so you can pick a movie and see the top suggestions.
The core intuition and the maths behind collaborative filtering, one of the most widely used techniques for building recommendation systems.
Video 4minBuild a content-based recommender on the TMDB 5000 movie dataset: learn TF-IDF and cosine similarity, then recommend similar movies.
Video 15minA full walkthrough of building a movie recommender from user ratings with collaborative filtering.
Video 57minImplement Matrix Factorization (MF) and Generalized Matrix Factorization (GMF) in Python with PyTorch, then train the GMF model.
Video 23min