DH-GEM

Source code for the paper, Talent Demand-Supply Joint Prediction with Dynamic Heterogeneous Graph Enhanced Meta-Learning, in SIGKDD 2022.

Requirement

torch==1.10
numpy==1.21.2
pandas==1.3.4
pytorch-lightning==1.5.3
dgl==0.6.1

Usage

You can easily run our code by

python code/main.py -R your-data-path -D dataset-name

More hyper-parameters setting please refer to code/args.py.

Dataset

In this work, we conduct experiments on 3 datasets, i.e. IT, FIN, CONS. They have the same format and in this repository we provide an example of job postings (demand) and work experiences (supply) data. You can collect your own datasets and run our code.

Job Postings (Demand)

Company	Time	Position	Location
Amazon	201903	Information	Boston, MA, US
...	...	...	...

Work Experiences (Supply)

People	Company	StartDate	EndDate	Position
456342	IBM	201603	201603	Information
...	...	...	...

Citation

If you find our work interesting, you can cite the paper as

@inproceedings{guo2022talent,
  title={Talent Demand-Supply Joint Prediction with Dynamic Heterogeneous Graph Enhanced Meta-Learning},
  author={Guo, Zhuoning and Liu, Hao and Zhang, Le and Zhang, Qi and Zhu, Hengshu and Xiong, Hui},
  booktitle={Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  pages={2957--2967},
  year={2022}
}