Movie Streaming Personalized Recommendation Engine

Project Overview

This project implements a Neural Collaborative Filtering (NCF) based movie recommendation system. It uses a custom NCF model to predict user preferences for movies and generate personalized recommendations. The system is designed to process large-scale user-item interaction data and provide accurate movie suggestions.

Key Features

• Custom Neural Collaborative Filtering (NCF) model implementation
• Efficient data preprocessing and standardization pipeline
• Advanced model training with early stopping and learning rate scheduling
• User-specific movie recommendation generation
• Scalable architecture suitable for large datasets

Project Structure

1. logs and other info: logs of the entire process as well as JSON and YAML files made during the process including screenshots of model usage
2. scripts: scripts of the whole process except for model usage
3. user-recommendations-and-history-script.py: Script for generating user-specific recommendations, get the preprocessed_data.npz file in the same folder as this for it to work
4. preprocessed_data.npz: https://github-1.s3.amazonaws.com/preprocessed_data.npz (not included in the repository)
5. best_ncf_model.pth: Saved best model weights
6. Dataset: https://www.kaggle.com/datasets/parasharmanas/movie-recommendation-system

Documentation

• MODEL.md: https://github.com/DimitriVavoulisPortfolio/movie-streaming-personalized-recommendation-engine/blob/main/MODEL.md
• PROCESS.md: https://github.com/DimitriVavoulisPortfolio/movie-streaming-personalized-recommendation-engine/blob/main/PROCESS.md

Model Performance

• Accuracy: 0.8210
• Precision: 0.8210
• Recall: 1.0000
• F1 Score: 0.9017

Quick Start Guide

1. Clone the repository:

   git clone https://github.com/DimitriVavoulisPortfolio/movie-streaming-personalized-recommendation-engine.git
   cd movie-recommendation-system
   

2. Install dependencies:

   pip install pandas numpy scikit-learn PyYAML torch
   

3. To train the model:

   python model-training-script v1.5.py
   

4. To generate recommendations:

   python user-recommendations-and-history-script.py
   

Future Work

• Implement more advanced recommendation algorithms
• Create a user-friendly web interface for the recommendation system
• Develop an API for real-time recommendation generation
• Optimize model performance for larger datasets

License

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.

Contact

For any questions or feedback, please open an issue in this repository or contact Dimitri Vavoulis.