🎬 Movie Recommender System

A content-based movie recommendation system which was built using :

• First tags were extracted from the dataset such as Director's name , Actor's name , Genre and Overview
• Then I performed data tranformation and remove space in between words
• Next step was to combine all the extracted features in one line called "Tags"
• Then i extracted 5k commong words and capared each movie tags with them to form a 5000x5000 matrix
• I used Text vectorization to convert matrix of each tags into verctor(cordinates) and then we will plot in a 2d space
• Now we have to calculate distance between to vectors in 2d space to calculate the similarity
• I used Cosine distance which is angle between two vectors point
• I did not used Eucladian distance is not a good measure to find distance between two points when we have high dimensionality points
• Again we will plot a 5000x5000 matrix of all the similarity between movies
• After sorting and enumirating values in matrix we get most simmilar movies

Features

• ✅ Content-based movie recommendations using cosine similarity
• ✅ Movie posters from TMDB API
• ✅ YouTube trailer links
• ✅ Interactive Streamlit web interface
• ✅ Dockerized for easy deployment
• ✅ 4800+ movies database

Tech Stack

• Backend: Python, Pandas, NumPy, Scikit-learn, NLTK
• Frontend: Streamlit
• API: TMDB (The Movie Database)
• Containerization: Docker, Docker Compose
• Machine Learning: Bag of Words, Cosine Similarity

Prerequisites

• Docker and Docker Compose installed
• OR Python 3.11+ (for local development)

🐳 Running with Docker (Recommended)

Option 1: Using Docker Compose (Easiest)

# Build and start the container
docker-compose up --build

# Run in detached mode
docker-compose up -d

# View logs
docker-compose logs -f

# Stop the container
docker-compose down

The app will be available at: http://localhost:8501

Option 2: Using Docker directly

# Build the image
docker build -t movie-recommender .

# Run the container
docker run -p 8501:8501 movie-recommender

# Run in detached mode
docker run -d -p 8501:8501 --name movie-app movie-recommender

# Stop the container
docker stop movie-app
docker rm movie-app

💻 Running Locally (Without Docker)

1. Install Dependencies

pip install -r requirements.txt

2. Run the Application

streamlit run app.py

The app will open automatically in your browser at http://localhost:8501

📁 Project Structure

recomender_system/
├── app.py                      # Main Streamlit application
├── movies.pkl                  # Preprocessed movie data
├── similarity.pkl              # Similarity matrix
├── tmdb_5000_movies.csv       # Raw movie dataset
├── tmdb_5000_credits.csv      # Raw credits dataset
├── recomender_system.ipynb    # Model training notebook
├── Dockerfile                  # Docker configuration
├── docker-compose.yml          # Docker Compose configuration
├── requirements.txt            # Python dependencies
├── .dockerignore              # Docker ignore file
└── README.md                   # This file

🎯 How It Works

1. Data Processing: Movies are processed using NLP techniques (stemming, bag of words)
2. Feature Extraction: Extracts genres, keywords, cast, crew, and overview
3. Vectorization: Converts text features into numerical vectors
4. Similarity Calculation: Uses cosine similarity to find similar movies
5. Recommendation: Returns top 5 most similar movies with posters and trailers

🚀 Usage

1. Open the app in your browser
2. Select a movie from the dropdown
3. Click "Show Recommendation"
4. View 5 similar movies with:
   • Movie posters
   • Watch Trailer buttons (opens YouTube)

🔧 Configuration

Environment Variables

You can customize the Streamlit configuration by setting these environment variables:

STREAMLIT_SERVER_PORT=8501
STREAMLIT_SERVER_ADDRESS=0.0.0.0
STREAMLIT_SERVER_HEADLESS=true
STREAMLIT_BROWSER_GATHER_USAGE_STATS=false

TMDB API Key

The app uses TMDB API for fetching posters and trailers. The API key is currently hardcoded in app.py. For production:

1. Get your API key from TMDB
2. Replace the API key in app.py or use environment variables

🐛 Troubleshooting

Docker Issues

Container won't start:

# Check logs
docker-compose logs

# Rebuild without cache
docker-compose build --no-cache
docker-compose up

Port already in use:

# Change port in docker-compose.yml
ports:
  - "8502:8501"  # Use 8502 instead

Missing Posters/Trailers

Some movies might not have posters or trailers due to:

• Movie ID mismatch with TMDB database
• Movie removed from TMDB
• Network issues
• API rate limits

The recommendations will still work correctly.

📊 Dataset

• Source: TMDB 5000 Movie Dataset
• Movies: ~4800 movies
• Features: genres, keywords, cast, crew, overview

🔮 Future Enhancements

• User authentication
• Save favorite movies
• Collaborative filtering
• Deploy to cloud (AWS/Azure/GCP)
• Add movie ratings and reviews
• Real-time search with autocomplete
• Mobile responsive design

👨‍💻 Author

Created with passion by Shrish Mishra

---

⭐ If you find this project useful, please consider giving it a star!