CineMatch: A Hybrid Movie Recommendation System

CineMatch is a movie recommendation system designed to provide personalized movie suggestions. It leverages data from The Movie Database (TMDb) and MovieLens to deliver relevant recommendations.

The system is built on a hybrid model that combines the strengths of both content-based and collaborative filtering methods. It analyzes movie attributes like genre, cast, and keywords to find similar content, and then re-ranks these suggestions based on predicted user ratings, ensuring the final recommendations are tailored to individual tastes. A popularity-based model also serves as a baseline for general recommendations.

Live Demo

You can try out the deployed application on Hugging Face Spaces:

CineMatch on Hugging Face

How It Works

The recommendation process begins by using a TF-IDF vectorizer to analyze textual features of movies, calculating content similarity with cosine similarity. To personalize these results, a Singular Value Decomposition (SVD) model, trained on user ratings, predicts how a user would rate the similar movies. The final list of recommendations is then sorted based on these predicted scores.

Running the Project Locally

To set up and run this project on your own machine, follow these steps:

1. Place all the necessary .csv data files into a folder named Data/ at the root of the project.

2. Run the exploratory data analysis and feature engineering notebooks (EDA/EDA.ipynb and Model/Prep.ipynb) to process the raw data.

3. After preparing the data, you can train the models. While there are notebooks for content-based and collaborative filtering, the hybrid approach is recommended for the best results.

4. The final, deployable application is a Gradio app located in the cinematch/ folder.