Movie Recommendation System

This project uses a cosine similarity-based retriever and an LSTM-based reranker to make movie recommendations. Input data should be in the form of sequential user ratings on movies.

The model learns to predict the next movie_id based on the user profile — which includes:

• movie_ids that the user has rated,
• the actual rating,
• corresponding tag values of the movie.

It is trained in a supervised manner using the MovieLens 20M dataset.

---

Notes

1. Please refer to the notebook for full training details. main.py only provides a simplified training pipeline.
2. Change configurations in src/config.py.
3. Hyperparameter tuning is not yet set up (currently commented out).

---

How to Run

python main.py