🍿 Movie Recommender System

This repository implements an end-to-end movie recommender system that leverages neural collaborative filtering to generate personalized movie recommendations based on historical user data. The core idea is to learn latent representations of users and movies from past interactions, enabling the model to predict the strength of the relationship between a user and a movie. The primary objective is to facilitate intelligent recommendations such that when a user inputs a specific movie or expresses certain preferences, the system accurately predicts similar movies that the user might enjoy.

At the heart of the project is a deep learning model that employs embedding layers for users and movies, which are then combined and processed through a multi-layer perceptron to capture complex, non-linear interaction patterns. This architecture is specifically designed to model the subtleties of user behavior and movie characteristics, ensuring that the recommendations are both personalized and relevant. By training on historical rating data, the model learns to uncover latent features that define user preferences and movie attributes without relying on manually engineered features.

The repository also provides a suite of recommendation functions that cater to various use cases, including similarity-based recommendations, genre-based recommendations, and a combined approach that integrates both collaborative and content signals. These functions allow users to obtain targeted movie suggestions—whether by finding movies similar to a given title or by filtering based on specific genres—making the system versatile and adaptable for different recommendation scenarios.

For more information about the training process, please check the mov-recsys.ipynb file in the training folder.

Check here to see my other ML projects and tasks.

🎈 Demo App

If you encounter message This app has gone to sleep due to inactivity, click Yes, get this app back up! button to wake the app back up.

⚙️ Running Locally

If the demo page is not working, you can fork or clone this repository and run the application locally by following these steps:

1. Clone the repository:

   git clone https://github.com/verneylmavt/st-mov-recsys.git
   cd st-mov-recsys

2. Install the required dependencies:

   pip install -r requirements.txt

3. Run the Streamlit app:

   streamlit run app.py

Alternatively you can run jupyter notebook demo.ipynb for a minimal interface to quickly test the model (implemented w/ ipywidgets).

⚖️ Acknowledgement

I acknowledge the use of the MovieLens 32M dataset provided by GroupLens Research at the University of Minnesota. This dataset has been instrumental in conducting the research and developing this project.

• Dataset Name: MovieLens 32M
• Source: https://grouplens.org/datasets/movielens/32m/
• License: The dataset may be used for research purposes under specific conditions; please refer to the usage license for details.
• Description: This dataset contains over 32 million ratings and 2 million tag applications applied to 87,585 movies by 200,948 users, collected from January 9, 1995, to October 12, 2023.

I deeply appreciate the efforts of GroupLens Research in making this dataset available.