A movie recommender by year: didn't go a lot to the movies in 2012? Check here what you could have liked!
Some years are busier than others and sometimes at the end of the year, I notice I didn't see a lot of movies. My goal here is to create a movie recommender that recommends movie based on what you liked the other years.
- Find dataset
- Create a general purpose movie recommender on small dataset
- Test movie recommender
- Add a filter to recommend only movies of a given year
- Visualize latent matrix
- Select evaluation metrics (MAP@K, RMSE, MAE...)
- Adapt to new user: from notebook, ask if the user liked a bunch of movies and return recommendation
- Implement best sampling strategy to ask which movies the user saw and how they rate it
- Fine tune recommender using adapted metrics
- [Optional] Develop Dash app to use the movie recommender
- [Optional] Use PySpark to run the SVD on the large dataset
- [Optional] Finish interpreting SVD with proper visualizations