🌐 REPOmmend

Network Analysis Project: Building a Recommendation System from a Bipartite Network

This repository is used to save the progress of the Fall 2021 Network Analysis Project. The goal of this project was to build a recommendation system for github repositories, based on the 2009 GitHub bipartite network, representing User-Repository-Watches (unweighted) relationships. A similar pipeline as used within this project can be applied generically to any bipartite graph, to build a recommendation for either of the two classes.

While the focus of this project was on understanding, applying and argue in favor or against theoretical concepts of network analysis using real-life data, the final product - a recommendation graph (that was statically saved for a max of 5 recommendations) - was translated into a simple web-app. Below, you can find all the important resources:

▶️ Web application

▶️ Project presentation

▶️ Course notes

Contributors

Ludek Cizinsky

Jonas-Mika Senghaas

Lukas Rasocha

Get started

Virtual Environment (Dependencies)

The project was (excluding the web-app) entirely written in Python and makes use of several external libraries. If you want to run the scripts and notebooks yourself, it is recommened to create a virtual environment (using the python environment manager) of you choice and then install all dependencies of it from the requirements.txt. Follow the following steps for create a stadard python venv:

1. Create a virtual environemnt: python3 -m venv [name of venv]

2. Activate the venv using: source [name of env]/bin/activate. You can decativate using the command deactivate

3. Update pip: pip install --upgrade pip

4. Install all dependencies: pip install -r <path/to/requirement.txt>

Github API

We use the public GitHub API to dynamically query additional metadata to be displayed in our web-app and to gain additional metadata for our recommendation system. While the API has free access, using an API token increases the number of API requests to 5000/hour. If you want to run the scripts yourself, you must create your own free, GitHub API token. Follow the following steps:

1. Visit your personal GitHub Token Settings

2. Create a new personal token and set the scope to public_repo.

3. Copy the token and save it into a python file in the following location csripts/envvars.py. Format your file as follows:

   GITHUB=TOKEN=[your_token]
   

4. The file is automatically ignored by .gitignore.