Configurations, "HOME_PATH" should be properly set before runing the code.

Code for sampling actions with different strategies.

Preprocessing codes, such as generating the skill graph and itemsets.

The code for the environment, including source code written in C++ and python packages for linux built with swig.

Containing our model and the baseline methods.

Load the training data, train the models, save the models and evaluate the models.

If you use our models, please cite the following paper:

@inproceedings{sun2021cost,
  title={Cost-Effective and Interpretable Job Skill Recommendation with Deep Reinforcement Learning},
  author={Sun, Ying and Zhuang, Fuzhen and Zhu, Hengshu and He, Qing and Xiong, Hui},
  booktitle={Proceedings of the Web Conference 2021},
  pages={3827--3838},
  year={2021}
}