PyTorch implementation for DPLink: User Identity Linkage via Deep Neural Network From Heterogeneous Mobility Data. Jie Feng, Mingyang Zhang, Huandong Wang, Zeyu Yang, Chao Zhang, Yong Li, Depeng Jin. WWW 2019. If you find our code is useful for your research, you can cite our paper by:

@inproceedings{feng2019dplink,
  title={DPLink: User Identity Linkage via Deep Neural Network From Heterogeneous Mobility Data},
  author={Feng, Jie and Zhang, Mingyang and Wang, Huandong and Yang, Zeyu and Zhang, Chao and Li, Yong and Jin, Depeng},
  booktitle={The World Wide Web Conference},
  pages={459--469},
  year={2019},
  organization={ACM}
}

• ISP-Weibo Data (main data used in the paper)
  • This is the private data collected and processed by ourselves and partners. We cannot directly published it due to the privacy issue. If you are interested in it and want to use it in your paper for academic purpose, you can contact with us via the email in this page with your identity information.
• Foursquare-Twitter Data
  • This data is from Transferring heterogeneous links across location-based social networks. Jiawei Zhang, Xiangnan Kong, and Philip S. Yu. WSDM 2014. We have no right to directly publish it. If you are interested in this dataset, you can contact with the original author to access the dataset.

• Python 2.7
• PyTorch 0.4
• tqdm 4.22
• mlflow 0.5
• numpy 1.14.0
• setproctitle 1.1.10
• scikit-learn 0.19.1

• run.py # scripts for run experiments
• match.py # training codes
• preprocessing.py # trajectory data preprocessing
• utils.py # utils for training
• models.py # models
• GlobalAttention.py # attention scripts from opennmt-py

To train a new model (default settings are recorded in the run.py)

python run.py --data=foursquare --model=ERPC --pretrain=1 --pretrain_unit=ERCF

E: embedding, R: rnn, P: pooling, C: co-attention, F: fully connected network. ERPC is the default model in paper, model name can also be ERC(without pooling). ERCF is the default pretrain mode in paper, which means all the components in the model are pretrained. You can choose E, R, C, F for only pretrain selected component and N is for non-pretrain.

Baselines from traditional baselines, TULER and t2vec. Some codes from OpenNMT-py, InferSent and awd-lstm-lm.