Emrecan Ulu
Research View on GitHub
Research · 2025

Rec-Dating Project

This project studies the rec-dating dataset as a role-based bipartite network.

The notebooks are arranged in four steps and can be run in order from data preparation to the final figures.

Core Idea

  • rater nodes send ratings
  • profile nodes receive ratings
  • edge weight is the observed score from 1 to 10

This framing lets us separate outgoing activity from received attention and apply network measures such as HITS in a role-consistent way.

Main Questions

  1. How strongly do popularity and prestige align on the profile side?
  2. How concentrated is received attention?
  3. Do the profiles dominating overall interaction also dominate high-rating buckets?
  4. Which profile-side features are most aligned with elite interaction and elite high-rating status?

Project Structure

  • data/: raw dataset
  • src/rec_dating_project/: reusable project code
  • scripts/: analysis scripts used by the notebooks
  • notebooks/: the ordered notebook workflow
  • paper/: LaTeX paper draft and bibliography
  • outputs/data/: generated tables reused across notebooks
  • outputs/figures/: generated figures reused in notebooks and paper

Environment Setup

The project was last checked with Python 3.11.9.

From the project root:

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

The raw dataset is expected at:

data/rec-dating/rec-dating.edges

Download it manually from the Network Repository page and place it at the path above:

https://networkrepository.com/rec-dating.php

Recommended Notebook Workflow

Run the notebooks in this order:

  1. notebooks/01_data_preparation.ipynb
  2. notebooks/02_rec_dating_exploration.ipynb
  3. notebooks/03_applications.ipynb
  4. notebooks/04_final_plots_for_paper.ipynb

What each notebook does:

  • 01: inspects the raw file, explains the role-based modeling choice, and builds the cached dataset summary
  • 02: explores popularity, prestige, inequality, and descriptive plots
  • 03: applies the framework to bucket concentration and feature alignment
  • 04: gathers the final figures and reference values used to check the project outputs

How To Run The Notebooks

Option A: Interactive

Launch JupyterLab from the project root:

jupyter lab

Then open the notebooks and run them in the numbered order above.

Option B: Fully Reproducible Terminal Execution

If you want to execute everything from the terminal:

MPLCONFIGDIR=/tmp/matplotlib-codex python3 -m nbconvert --to notebook --execute notebooks/01_data_preparation.ipynb --inplace --ExecutePreprocessor.timeout=1200
MPLCONFIGDIR=/tmp/matplotlib-codex python3 -m nbconvert --to notebook --execute notebooks/02_rec_dating_exploration.ipynb --inplace --ExecutePreprocessor.timeout=1200
MPLCONFIGDIR=/tmp/matplotlib-codex python3 -m nbconvert --to notebook --execute notebooks/03_applications.ipynb --inplace --ExecutePreprocessor.timeout=1200
MPLCONFIGDIR=/tmp/matplotlib-codex python3 -m nbconvert --to notebook --execute notebooks/04_final_plots_for_paper.ipynb --inplace --ExecutePreprocessor.timeout=1200

The MPLCONFIGDIR prefix helps on headless or restricted environments where Matplotlib cannot write to its default cache directory.

Generated Outputs And Rebuild Behavior

  • The notebooks reuse cached files in outputs/data/ and outputs/figures/ when they already exist.
  • Missing artifacts are rebuilt automatically by the relevant scripts.
  • If you want a full refresh, set FORCE_REBUILD = True in the setup cell of the notebook you are running.

Scripts Used By The Notebooks

The notebooks rely on these scripts:

  • scripts/01_dataset_overview.py
  • scripts/02_full_project_analysis.py
  • scripts/03_profile_rating_extremes.py
  • scripts/04_profile_feature_alignment.py
  • scripts/05_degree_distribution_fit.py

If the notebook templates ever need to be regenerated, run:

python3 scripts/06_rebuild_notebooks.py

Paper Reproduction

After 04_final_plots_for_paper.ipynb has been executed, the paper figures should be available under outputs/figures/.

To compile the paper from the project root:

latexmk -pdf -cd paper/main.tex

Notes

  • The dataset is large, so cached artifacts are used intentionally to keep the notebook workflow responsive.
  • The last notebook is a final check that the generated figures and summary values still line up with the rest of the project.