Pick a Movie for Me 🎬

A multi-user movie recommendation web application with reinforcement learning capabilities. The system learns from each user's preferences over time to provide personalized movie suggestions.

📖 For complete documentation, see DOCUMENTATION.md

Features

• 🎲 Random Movie Recommendations - Get personalized movie suggestions
• 🤖 Reinforcement Learning - The system learns from your feedback to improve recommendations
• 🎨 Beautiful UI - Modern, responsive design with TailwindCSS
• 📊 Genre-Based Learning - Uses multi-armed bandit algorithm to learn your genre preferences

Tech Stack

• Backend: Flask (Python)
• Frontend: HTML + TailwindCSS (CDN) + JavaScript
• Movie Data: TMDB (The Movie Database) API
• RL Model: Multi-armed bandit with epsilon-greedy strategy
• Storage: JSON file for RL feedback data

Setup Instructions

1. Activate Virtual Environment

# On Windows
venv\Scripts\activate

# On macOS/Linux
source venv/bin/activate

2. Install Dependencies

pip install -r requirements.txt

3. Get TMDB API Key

1. Go to TMDB
2. Create a free account
3. Go to API Settings
4. Request an API key (free)
5. Copy your API key

4. Set Environment Variable

Windows (PowerShell):

$env:TMDB_API_KEY="your_api_key_here"

Windows (CMD):

set TMDB_API_KEY=your_api_key_here

macOS/Linux:

export TMDB_API_KEY="your_api_key_here"

Or create a .env file (not included in git):

TMDB_API_KEY=your_api_key_here

5. Run the Application

python app.py

The application will be available at http://localhost:5000

How It Works

Reinforcement Learning Model

The system uses a multi-armed bandit approach where:

• Each genre is an "arm" that can be pulled
• The system tracks:
  • Total recommendations per genre
  • Positive feedback (likes) per genre
• Expected reward = positive_count / total_count

Recommendation Strategy

1. Epsilon-Greedy Algorithm:

   • 15% of the time: Explore (random genre)
   • 85% of the time: Exploit (genre with highest expected reward)

2. Feedback Loop:

   • When you click "I Like This" → reward = 1
   • When you click "I Don't Like This" → reward = 0
   • The system updates genre statistics and saves to feedback_store.json

3. Over Time:

   • Genres you like get recommended more often
   • Genres you dislike get recommended less often
   • The system adapts to your preferences

Project Structure

recom sys/
├── app.py                 # Flask application and routes
├── rl_model.py           # Reinforcement learning model
├── tmdb_service.py       # TMDB API integration
├── requirements.txt      # Python dependencies
├── feedback_store.json   # RL data (created automatically)
├── templates/
│   ├── index.html        # Landing page
│   ├── movie.html        # Movie recommendation page
│   └── error.html        # Error page
└── README.md             # This file

API Routes

• GET / - Landing page
• GET /recommend - Get a movie recommendation
• POST /feedback - Submit feedback (JSON: {"genre": "Action", "liked": true})
• GET /stats - View RL model statistics (debug endpoint)

Example Usage

1. Visit http://localhost:5000
2. Click "Pick a Movie for Me"
3. View the recommended movie
4. Click "I Like This" or "I Don't Like This"
5. Click "Pick Another Movie" to get a new recommendation
6. The system learns from your feedback!

Troubleshooting

"TMDB_API_KEY environment variable not set"

• Make sure you've set the environment variable before running the app
• Check that your API key is valid

"No genres available from TMDB"

• Check your internet connection
• Verify your TMDB API key is correct

"Could not fetch a movie"

• TMDB API might be temporarily unavailable
• Try again in a few moments

Future Enhancements

• User sessions to track individual preferences
• More sophisticated RL algorithms (UCB1, Thompson Sampling)
• Movie filtering by rating, year, etc.
• Watchlist functionality
• Multiple user support

License

This is a personal project for educational purposes.

Credits

• Movie data provided by The Movie Database (TMDB)
• UI inspired by pickamovieforme.com