It's movie night and the question comes up again: "Which movie should we watch?". Worry no more, this app will make your life easier! Just pick two movies and the app will return the best recommendations for both.
Live demo here
- The recommendation system works as a hybrid model.
- It uses content-based based on similarity of movie attributes: genre, actors, and other descriptive metadata associated with a movie. And collaborative filtering based on user ratings from the movieLens dataset .
- Both attributes and ratings were transformed into vectors in order to create a latent factor model with the best features. Those that explain at least 80% of the variance.
- Then used cosine similarity as a similarity measure to find the most similar movies for each of the user choices.
- Lastly, a global measure average for both movies was calculated for both content and collaborative.
- Python
- Streamlit
- Scikit-learn