Movie Recommendation System Demo App 🎬

An interactive web application demonstrating three recommendation algorithms on the MovieLens dataset.

Features

• 3 Recommendation Algorithms:

  • Popularity-Based (non-personalized baseline)
  • Item-KNN using SAR (Smart Adaptive Recommendations)
  • ALS available in notebook version (requires PySpark)

• Interactive UI:

  • 🏠 Home: Dataset statistics and exploration
  • 🎯 Get Recommendations: Personalized movie recommendations for any user
  • 📊 Model Evaluation: Compare algorithms with MAP@K and NDCG@K metrics
  • ℹ️ About: Learn about algorithms and metrics

• Evaluation Metrics:

  • MAP@K (Mean Average Precision)
  • NDCG@K (Normalized Discounted Cumulative Gain)

Quick Start

Installation

1. Install dependencies:

pip install -r requirements.txt

Run the App

streamlit run app.py

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

Usage

1. Explore the Dataset (Home page):

   • View dataset statistics
   • See rating distributions
   • Discover top-rated movies

2. Get Recommendations:

   • Select a user ID
   • Choose an algorithm
   • Get personalized movie recommendations
   • View user's rating history

3. Evaluate Models:

   • Run evaluation on all algorithms
   • Compare MAP@K and NDCG@K scores
   • View performance benchmarks
   • See visualization charts

Dataset

The app uses the MovieLens dataset from GroupLens Research:

• Default: 100k ratings (for fast demo)
• Can be configured for 1M, 10M, or 20M ratings
• Contains user ratings, movie IDs, and movie titles

To change dataset size, edit the MOVIELENS_DATA_SIZE variable in app.py:

MOVIELENS_DATA_SIZE = '100k'  # Options: '100k', '1m', '10m', '20m'

Algorithms Explained

1. Popularity-Based

• Recommends globally popular movies
• Score based on: 70% rating count + 30% average rating
• Fast but not personalized
• Good for cold-start scenarios

2. Item-KNN (SAR)

• Item-based collaborative filtering
• Uses Jaccard similarity to find similar items
• Personalized based on user history
• Good balance of speed and quality

3. ALS (Notebook Only)

• Matrix factorization approach
• Learns latent factors for users and items
• Best personalization but requires PySpark
• See movie_recommendation_demo.ipynb for implementation

Project Structure

.
├── app.py                          # Main Streamlit application
├── requirements.txt                # Python dependencies
├── movie_recommendation_demo.ipynb # Jupyter notebook with ALS
├── als_movielens.ipynb            # Original ALS example
└── README.md                      # This file

Configuration

Edit these constants in app.py to customize:

TOP_K = 10                          # Number of recommendations
MOVIELENS_DATA_SIZE = '100k'       # Dataset size
TRAIN_RATIO = 0.75                 # Train/test split ratio
RANDOM_SEED = 42                   # Random seed for reproducibility

Performance Notes

• 100k dataset: Runs smoothly on any machine
• 1M dataset: Still fast, recommended for demos
• 10M/20M datasets: Slower, recommended for servers with more RAM
• Models are cached using @st.cache_resource for faster subsequent runs

Requirements

• Python 3.8+
• 4GB+ RAM (for 100k dataset)
• 16GB+ RAM recommended (for 20M dataset)

Technologies

• Streamlit: Web framework
• Microsoft Recommenders: ML library
• Plotly: Interactive visualizations
• Pandas: Data manipulation
• NumPy: Numerical operations

References

• Microsoft Recommenders
• MovieLens Dataset
• Streamlit Documentation

License

This project is for educational purposes. MovieLens data is provided by GroupLens Research.

Troubleshooting

Issue: App is slow

• Solution: Use smaller dataset (100k or 1m)
• Clear Streamlit cache: Press 'C' in the running app

Issue: Memory error

• Solution: Reduce dataset size or increase available RAM

Issue: Module not found

• Solution: pip install -r requirements.txt

Next Steps

• Try different dataset sizes
• Experiment with algorithm parameters
• Compare results with the notebook version (includes ALS)
• Deploy to Streamlit Cloud for public access

---

Enjoy exploring movie recommendations! 🍿