Inspiration
When having a few streaming services, I didn't know where to look and what to watch, so I decided to make a website for myself.
What it does
Once you create a user in the system, you can leave ratings of trending movies and also recommended movies by the model.
How we built it
The SVD model was build based on the scikit-surprise library, with a Flask backend that manages users, ratings, the model predictions and re-training as well as API and database communications.
Challenges we ran into
New users have a cold start issue where in order to get any meaningful recommendations, they would need to leave multiple ratings.
Accomplishments that I'm proud of
Being able to create an interactive interface for users, an optimizing all back end services to significantly reduce load times.
What we learned
I learned that large scale projects have to planned fully in advance, and that every component has to be thought of before actually developing.
What's next for What2Watch
I plan on switching the model to something that would handle incremental training, as this would allow the system to handle new ratings much more efficiently.
Log in or sign up for Devpost to join the conversation.