This is the REPORT code for fake news detection. We are currently using twitter15&16 dataset

Project structures of this repo are listed in project_structure.txt

1.torch_geometric: (also called 'PyG')

torch_geometric does not support conda installation currently, please view your installing versions at here, and install it with pip. you need to install 4 dependencies before installing PyG:

• torch-scatter
• torch-sparse
• torch-cluster
• torch-spline-conv

• Python: 3.8.5
• Pytorch: 1.5.0 (torch_geometric is currently available to up to torch 1.5.0)
• cuda: 10.2

script:

    pip install torch-scatter==2.0.4+cu102 -f https://pytorch-geometric.com/whl/torch-1.5.0.html
    pip install torch-sparse==0.6.3+cu102 -f https://pytorch-geometric.com/whl/torch-1.5.0.html
    pip install torch-cluster==1.5.4+cu102 -f https://pytorch-geometric.com/whl/torch-1.5.0.html
    pip install torch-spline-conv==1.2.0+cu102 -f https://pytorch-geometric.com/whl/torch-1.5.0.html
    pip install torch-geometric
    # change 'cu102' to 'cpu' or 'cu101' or others depending on your PyTorch installation. 

all data are put in datasets/, complete dataset is available at: https://www.dropbox.com/home/FakeNew2020/datasets

1. in each folder that contain task.py, configure the parameter according to your experiments.
   
2. in datasets/twitter15(6)/raw_data/, run raw_data.bat
   
3. in pre_process/, run pre_process.bat
   
4. in load_data15(6)/, run add_label.py and count_tweet.py