Building the Project

File folders

data: the dataset folder

edge_input: contains the edge files used in preprocessing.

models: contains the main code for the model.

path_data: The result of path collection.

preprocess: contains the preprocessing files.

Requirements

To run the code, you need the following dependencies at least:

networkx==2.3 
numpy==1.21.6 
pandas==1.3.5 
scikit_learn==1.2.1 
scipy==1.4.1
scipy==1.4.1 
torch==1.11.0 
torch_cluster==1.5.9 
torch_scatter==2.0.9 
torch_sparse==0.6.12 
torch_spline_conv==1.2.1 
torch_geometric==2.0.4 
tqdm==4.43.0 
python ==3.8

Data preprocessing

The code executes in the following order: (1) python process_icews.py (2) python path_data.py (3) python init_rw.py (4) g++ gen_merw.cpp -o gen_merw -g -Wall -O2 -mcmodel=medium (5) ./gen_merw [data_name] [path_num] [path_length].

Main model

Files

learner.py: The beginning of the model.

datasets.py: Used to get datasets

models.py: The part of model.

neibs_info_embedding.py: The code that gets path embedding information.

contrastive_learning.py: The code that gets contrastive learning

Run the Experiments

python learner.py --dataset ICEWS14 --rank 800 --valid_freq 2 --max_epoch 20 --learning_rate 0.1 --batch_size 50 --n_hidden 160 --num_walks 1 --walk_len 8 --num_walks_neis 20 --walk_len_neis 6