ICI Reproducibility

Diffusion models benchmarking

Requirements

• python2.7 or python3.7
• sklearn, pandas, networkx, numpy

Datasets

• path: ICI/datasets/
• sources for raw data: Digg, Twitter.
• data preprocessing scripts: clean_diggs.py;clean_twitter.py
• preprocessed datasets: [filename].edgelist;[filename].seed;[filename].spread

Model implementations

• path: ICI/utils/

Run

• command:

python benchmark.py [--model model_name] [--dataset file_name] [--output 0/1] [--repeat simulation_times] [--step spread_in_each_step] [--beta ICI_param] [--gamma ICI_param];

• example:

python benchmark.py --model ici --dataset digg --output 1 --repeat 1000 --step 10 --beta 0.9 --gamma 0.6;

• Input arguments

  • datasets: --dataset={"digg","twitter"}
  • support models: --model={"ic", "icm", "icn", "ici", "lt", "ftm", "ltc"}
  • output mode: print all results by --output 1

Citation

Please kindly cite our work if you find our paper or codes helpful.

@inproceedings{zhang2024information,
  title={Information Diffusion Meets Invitation Mechanism},
  author={Zhang, Shiqi and Sun, Jiachen and Lin, Wenqing and Xiao, Xiaokui and Huang, Yiqian and Tang, Bo},
  booktitle={Companion Proceedings of the ACM on Web Conference 2024},
  pages={383--392},
  year={2024}
}