antico

antico copied to clipboard

Introduction

Implementation of our KDD paper "Detecting Cash-out Users via Dense Subgraphs". [Paper] [Code] [Video] [Link]

ANTICO is developped for spotting cash-out users based on bipartite graph and subgraph detection. It is designed for credit card services and real-world banking data.

Main Contact:

• Yingsheng Ji ([email protected])
• Zheng Zhang ([email protected])

Running ANTICO on a test example

Support python 3.6

pip install -r requirements.txt

The direct way to use ANTICO is to run in command line:

python toy.py

where default demo is running on a synthetic dataset.

Input Data Fromat:

• cards: account id, personal credit limit, label
• merchants: account id, prior, label
• transactions: card account id, merchant account id, amount, timestamp, label

The main configuration items involve start_date and ending_date, and time_spans in config.py.

Cite

@inproceedings{ji2022detecting,
  title={Detecting cash-out users via dense subgraphs},
  author={Ji, Yingsheng and Zhang, Zheng and Tang, Xinlei and Shen, Jiachen and Zhang, Xi and Yang, Guangwen},
  booktitle={Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  pages={687--697},
  year={2022}
}