Capturing Delayed Feedback in Conversion Rate Prediction via Elapsed-Time Sampling

Please cite this paper if you used any content of this repo in your work:

@inproceedings{DBLP:conf/aaai/YangLHZZZT21,
  author    = {Jia{-}Qi Yang and
               Xiang Li and
               Shuguang Han and
               Tao Zhuang and
               De{-}Chuan Zhan and
               Xiaoyi Zeng and
               Bin Tong},
  title     = {Capturing Delayed Feedback in Conversion Rate Prediction via Elapsed-Time
               Sampling},
  booktitle = {Thirty-Fifth {AAAI} Conference on Artificial Intelligence, {AAAI}
               2021},
  pages     = {4582--4589},
  publisher = {{AAAI} Press},
  year      = {2021},
  url       = {https://ojs.aaai.org/index.php/AAAI/article/view/16587},
}

This is the code and model checkpoints to reproduce the results on the public dataset.

We also implement Delayed feedback model(DFM, Chapelle 2014), Feedback Shift Importance Weighting (FSIW) (Yasui et al. 2020), Fake Negative Weighted (FNW) (Ktena et al. 2019), Fake Negative calibration(FNC) (Ktena et al. 2019) for comparison.

The criteo dataset is available at https://drive.google.com/file/d/1x4KktfZtls9QjNdFYKCjTpfjM4tG2PcK/view?usp=sharing

For detailed information, please refer to the comments.

Please run

python main.py --help

to see all the arguments.

I uploaded a run.sh file as a reference to run the code, however, the pathes should be modified accordingly.

A preprint version of this paper is available at https://arxiv.org/abs/2012.03245