Diploma

A movie recommender platform implemented as the link weight prediction task with Graph Neural Networks and Graph Databases

System Architecture

The platform consists of the following main components:

• Graph Database
  • A Neo4j instance is utilized to store the dataset
• Movies API
  • A REST API responsible for serving the movies and their metadata to the front-end component
  • Developed with NodeJS in Typescript
• Users API
  • A REST API responsible for performing the CRUD operations on users accounts and their ratings on the movies
  • Developed with NodeJS in Typescript
• Model API
  • A REST API responsible for the integration of the machine learning model that performs the predictions into the platform
  • Developed with Flask in Python
• Model
  • A machine learning model that utilizes Graph Neural Networks to perform predictions on the movies ratings for each user
  • Developed with Pytorch Geometric in Python
• Front-end
  • A friendly and easy-to-use web app that allows users to:
    • explore multiple movies and their metadata
    • gain insights on the underlying graph structure of the data via numerous graph visualizations
    • submit their ratings on the platform's movies
    • access personalized predictions on the movies' ratings
  • Developed with ReactJS in Javascript and Typescript