Movie Recommendation System

This project implements a movie recommendation system using vector search and cosine similarity. The system takes user input in the form of a movie title or overview and returns the most similar movies based on their embeddings.

Features

• Vectorized search for movie recommendations
• Cosine similarity calculation for finding similar movies
• RESTful API built with Flask
• Docker support for easy deployment

Requirements

• Python 3.10
• Flask
• txtai
• NumPy
• Gunicorn
• Pickle

Installation

1. Clone the repository:

   git clone https://github.com/NexusAurora/MovieApp.git
   cd MovieApp

2. Build the Docker image:

   docker build -t MovieApp .

3. Run the Docker container:

   docker run -p 8000:8000 MovieApp

Usage

Once the application is running, you can make POST requests to the /search endpoint to get movie recommendations.

Request Format

• URL: http://localhost:8000/search
• Method: POST
• Body (JSON):

{
    "query": "Your movie title or overview",
    "results": 5
}

Response Format

The response will be a JSON object containing the recommended movies and their probabilities:

[
    {
        "Output": {
            "title": "Movie Title 1",
            "overview": "Overview of Movie 1"
        },
        "Probability": 0.95
    },
    {
        "Output": {
            "title": "Movie Title 2",
            "overview": "Overview of Movie 2"
        },
        "Probability": 0.92
    }
]

Steps to Create the Recommendation System

1. Import and parse dataset: Load the dataset containing movie titles and overviews.
2. Extract relevant data: Keep only the title and overview, discarding other parameters.
3. Calculate vector embeddings: Vectorize each overview and store the embeddings.
4. Accept user input: Transform user input into an embedding.
5. Find cosine similarities: Calculate the cosine similarities between the user input embedding and the movie embeddings.
6. Return top results: Return the top N movies with the highest similarities.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

• txtai for the embeddings model.
• Flask for the web framework.
• Docker for containerization.

Author

My Name
My GitHub Profile