This repo is the python implementation of paper "Learning Reliable User Representations from Volatile and Sparse Data to Accurately Predict Customer Lifetime Value" (KDD2021).
Please first install the following python packages:
pytorch
DGL
pandas
numpy
scikit-learn
matplotlib
tqdm
statsmodels
cd ltv-code/src/
python trainer.py -c <config file path> -d <indice of GPU, e.g., 0>
Due to the data privacy restrictions, we have deleted some data preprocessing details with sensitive information in data_prepare.py. Please use your own dataset and customized dataset reader.
If you find our paper or code is useful for your research work, please cite the following BibTex:
@inproceedings{xing2021learning,
title={Learning Reliable User Representations from Volatile and Sparse Data to Accurately Predict Customer Lifetime Value},
author={Xing, Mingzhe and Bian, Shuqing and Zhao, Wayne Xin and Xiao, Zhen and Luo, Xinji and Yin, Cunxiang and Cai, Jing and He, Yancheng},
booktitle={Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining},
pages={3806--3816},
year={2021}
}