Raising Graphs From Randomness to Reveal Information Networks

This repo is the collection of the codes used in the experiments of Raising Graphs From Randomness to Reveal Information Networks.

Datasets

Datasets can be found here.

• All files have three columns: 1 unix timestamp, 2 user id, 3 user id.
• If the value in the 3rd column is "-1", then instead of an edge a root adopter was observed.
• Each undirected edge occurs once in the file, when it appeared first in the time series.
• Files are indexed by the same codes used in the paper.

Requirements

The following modules are required to run the codes written in Python:

• Numpy
• Scipy
• Matplotlib
• Pandas
• Seaborn
• Networkx

Running our measurements

These scripts generate all the statistics that we investigate in Section 4. Jupyter Notebooks introduced below will use the outputs of these measurements.

• To run all measurements, run the sh/experiment_all.sh
  • For $DATA_FOLDER please set the folder where you downloaded the network files
  • For $STATS_FOLDER please set an output folder where each measurement will be generated
• A short explanation of the executed scripts:
  • experiment_random_shuffle.sh creates randomly shuffled time series that we compare to the original network timelines
  • experiment_degree_distribution_stats.sh computes the degree distribution at different sizes for each network
  • experiment_network_events.sh computes the fraction of R,I,H events as the networks grow
  • experiment_exponent_stats.sh computes the degree distribution parameter estimates as the networks grow

Measurements covered by Jupyter Notebooks

The following notebooks cover our measurements explained in Section 4 of our paper. Please modify the input folder that should be the output of the previously detailed scripts in each Jupyter notebook.

• Average degree
• Evolving power-law degree distributions
• Average degree and the exponent of the power-law degree distribution
• Microscopic view
• No Fit to Uniform Edge Sampling

Exponential Model

The C++ code can be found here. To build the program, run the scons command inside the /cpp/src/ folder.

An example script for running the model is here.