Movie-Recommendation-System

Content-based movie recommender: a training pipeline (TF-IDF, SVD, top-K cosine neighbours) and a Django app with search, autocomplete and a JSON API. Small mod

What is Movie-Recommendation-System?

Content-based movie recommender: a training pipeline (TF-IDF, SVD, top-K cosine neighbours) and a Django app with search, autocomplete and a JSON API. Small models, about 1 ms lookups, ships with a demo model.

Overview of Movie-Recommendation-System

What it is

This is a content-based movie recommendation system. It includes a training pipeline that uses TF-IDF, truncated SVD, and top-K cosine neighbors to build a model from a TMDB-shaped CSV. It also includes a Django web application that serves the model, providing search with autocomplete, ranked results, and a JSON API. A demo model covering 6,248 movies is included in the repository, so the application works immediately after cloning with no additional setup. The system is designed to be small and fast, with neighbor arrays stored as memory-mapped files for efficient lookups. The training pipeline stores only the 50 nearest neighbors per film, avoiding large N x N matrices. The application is tested and includes safety features such as refusing to start with the development secret key when DEBUG is False, and reporting a starting status during model loading for platform health checks.

Key features

  • Content-based recommendations using TF-IDF, SVD, and cosine similarity, requiring no user history.
  • A training pipeline that converts any TMDB-shaped CSV into a small, fast model directory.
  • A Django web application with search, autocomplete, ranked results, and a JSON API.
  • Small model sizes: stores only the top 50 nearest neighbors per film, not a full similarity matrix.
  • Low memory usage: neighbor arrays are memory-mapped, so each request reads only one row.
  • Safe deployment: refuses to start with the development secret key when DEBUG=False, and reports a starting status during model loading.
  • A demo model of 6,248 movies is committed and works immediately.

Use cases

  • Running a movie recommendation website or service that works from the first query without collecting user history.
  • Using the training pipeline to build a custom recommendation model on a different movie dataset or catalogue.
  • Integrating the recommendation engine into other Python code using the provided inference class, which supports year, genre, and rating filters.
  • Using the JSON API endpoints for autocomplete and health checks in a larger application or microservice architecture.

Getting started

bashgit clone https://github.com/inboxpraveen/Movie-Recommendation-System.gitcd Movie-Recommendation-Systempython -m venv venvsource venv/bin/activate # Windows: venv\Scripts\activatepip install -r requirements.txtpython manage.py migratepython manage.py runserverOpen http://localhost:8000. Search works immediately because the demo model is found automatically.

License and pricing

The license is MIT. The project is open source.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/inboxpraveenmovie-recommendation-system.svg)](https://yunodb.com/tools/inboxpraveenmovie-recommendation-system?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/inboxpraveenmovie-recommendation-system?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/inboxpraveenmovie-recommendation-system.svg" alt="Featured on YunoDB" height="20"></a>