This repository contains the source code for the paper A Novel Framework with Information Fusion and Neighborhood Enhancement for User Identity Linkage published in ECAI 2020. The paper can be downloaded from Arxiv or the official site ECAI.
- Ubuntu 16.04
- python3.6
- numpy>=1.17.2
- scipy>=1.1.0
- torch>=1.1.0
- torch-geometric>=1.3.0
The preprocessed data can be downloaded from Google Drive. Due to the privacy concern, the raw data including user profiles and contents are not shared.
All data requires for the information fusion component lie in data/.
adj_s.pkl, adj_t.pkl: adjacency matrices of the source network and the target network, respectively.sims_p.pkl, sims_c.pkl: the ground truth similarity matrices of profile and content, respectively.train_test_{}.pkl: randomly split training and testing anchor user pairs at ratios range from 0.1 to 0.9.
The neighborhood enhancement component requires extra inputs such as the pre-trained node embeddings, candidate users and etc. To ease the evaluation, the ratio of the training set is set to be 0.8, and the pre-trained node embeddings and other data are provided in results/.
emb_0.8.pkl: the node embeddings.candidate_0.8.pkl: candidate user matrix.nei_0.8.pkl:{(u, v): [[], [], [], []]}, potential matched and unmatched neighbors of u and v.adj_list.pkl: the adjacency list of nodes in the source network and the target network.
All arguments can be modified in config.py. Specify the information by modifying the argument --options. The default value is 'structure profile content'.
- Information fusion component
cd runs
python IF.py- Neighborhood enhancement component
cd runs
python NE.pyIf you find this repository useful in your research, please consider citing the following paper:
@inproceedings{chen2020infune,
title={A Novel Framework with Information Fusion and Neighborhood Enhancement for User Identity Linkage},
author={Chen, Siyuan and Wang, Jiahai and Du, Xin and Hu, Yanqing},
booktitle={24th European Conference on Artificial Intelligence (ECAI)},
pages={1754--1761},
year={2020}
}