zhoudinglive/ahead

KDD21 Attentive Heterogeneous Graph Embedding for Job Mobility Prediction

Ahead

This is the source code of KDD21 paper "Attentive Heterogeneous Graph Embedding for Job Mobility Prediction", where the file structure is shown below. Although we cannot publish our dataset for the security problem, you can find the input&output format of our model in source_data/dataset.py. Feel free to contact us if you have any issues.

• file structure

.
├── LICENSE
├── README.md
├── main.py
├── model
│   ├── __pycache__
│   │   ├── ahead.cpython-38.pyc
│   │   ├── config.cpython-38.pyc
│   │   ├── dgru.cpython-38.pyc
│   │   ├── hgcn.cpython-38.pyc
│   │   └── utils.cpython-38.pyc
│   ├── ahead.py
│   ├── config.py
│   ├── dgru.py
│   ├── hgcn.py
│   └── utils.py
├── model_save
│   └── ahead_duration
├── requirements.txt
└── source_data
    ├── dataset.py
    ├── mock_company_ppr.npy
    ├── mock_dur_of_company.npy
    ├── mock_dur_of_title.npy
    ├── mock_external_graph.npy
    ├── mock_internal_graph.npy
    ├── mock_pid_str_dict.json
    ├── mock_final_records.npy
    ├── mock_skill_embedding.npy
    └── mock_title_ppr.npy

• code running

pip install requirements.txt
python main.py

If you find Ahead useful in your research, we ask that you cite the following paper

@inproceedings{zhang2021attentive,
  title={Attentive heterogeneous graph embedding for job mobility prediction},
  author={Zhang, Le and Zhou, Ding and Zhu, Hengshu and Xu, Tong and Zha, Rui and Chen, Enhong and Xiong, Hui},
  booktitle={Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery \& Data Mining},
  pages={2192--2201},
  year={2021}
}

ahead's Issues

Questions of the meaning of input data

Hi. This is not an issue that I am reporting, but I want to understand the following :

1. Can you please clarify the meaning of the acronym "ppr"? From the research paper, it appears to refer to the total number of job transition relationship. Is this correct?

2. Regarding the duration file, I am wondering if the value in the matrix would be the average of the blank periods (1+2)/2 or the average of the employment periods. For example, if person A worked at company A for 3 years, took a one-year break, and then worked at company B for 4 years, and person B worked at company A for 5 years, took a two-year break, and then worked at company B for 6 years.

3. I am curious why the internal graph values are either 0 or 1 instead of the number of job transitions. Can you please explain the reasoning behind this?

I will be grateful for any information you can provide.