CasCIFF

This repo provides a reference implementation of CasCIFF as described in the paper:

CasCIFF: A Cross-Domain Information Fusion Framework Tailored for Cascade Prediction in Social Networks
Hongjun Zhu, Member, IEEE, Shun Yuan, Xin Liu, Kuo Chen, Chaolong Jia, and Ying Qian
The paper will be published in Knowledge-Based Systems

Basic Usage

Requirements

The code was tested with python 3.7, pytorch 1.10, cudatoolkit 11.3.1, and cudnn 6.0. Install the dependencies via Anaconda:

# create virtual environment
conda create --name casciff python=3.7 cudatoolkit=11.3.1 cudnn=6.0

# activate environment
conda activate casciff

# install other requirements
pip install -r requirements.txt

Run the code

cd ./casCIFF

# generate information cascades
python ./preprocess/gene_cas.py 

# generate global graph embeddings
python ./preprocess/gene_global_emb.py 

# preprocess cascades data for training
python ./preprocess/preprocess_graph_signal_time.py

# preprocess global graph embeddings for training
python ./preprocess/preprocess_global_emb.py

# run CasCIFF model
python CasCIFF_train_shuffle.py

More running options are described in the codes.

Evaluate

In addition, we also provide already trained models for you to reproduce the experimental effects in the paper.

# run evaluate CasCIFF model
python CasCIFF_metric.py

Datasets

Our's datasets from CasFlow.

Thanks to Xovee Xu for providing the dataset. Datasets download link: Google Drive or Baidu Drive (password: 1msd).

The datasets we used in the paper are come from:

• Twitter (Weng et al., Virality Prediction and Community Structure in Social Network, Scientific Report, 2013).
• Weibo (Cao et al., DeepHawkes: Bridging the Gap between Prediction and Understanding of Information Cascades, CIKM, 2017). You can also download Weibo dataset here in Google Drive.
• APS (Released by American Physical Society, obtained at Jan 17, 2019).

Contact

For any questions please open an issue or drop an email to: [email protected], [email protected]