T-Net

Code for T-Net for combatting Human Trafficking

Installation

pip install -r requirements.txt

Unzip files

Unzip the data.zip and results/synthetic_asw.zip folder for running rest of the code.

Running T-Net

python3 main.py --data_file data/synthetic_asw/synthetic_labelled_graph.pkl --epochs 100 --save_dir results/synthetic_asw --save_filename tnet_cl_results.pkl

Running baselines

python3 main.py --data_file ht_datasets/synthetic_asw/synthetic_labelled_graph.pkl --epochs 100 --save_dir results/synthetic_asw --save_filename mlp_results.pkl --baseline --baseline_method mlp

Choose a baseline method name from mlp, gcn, nrgnn, pignn. For NRGNN and PIGNN install their code from their official github repository to run them or use the saved model from our results folder

1. NRGNN - (https://github.com/EnyanDai/NRGNN)
2. PIGNN - (https://github.com/TianBian95/pi-gnn)

For Misclassifciation results (Figure 2 in paper)

python3 main.py --data_file ht_datasets/synthetic_asw/synthetic_labelled_graph.pkl --save_dir results/synthetic_asw --get_misclassification

For tabular results (Table 5 in paper)

python3 main.py --save_dir results/synthetic_asw --print_results

For accessing ASW-Synth

If you want to get access to the synthetically generated dataset, send an email with a short description of why you need the data to pratheeksha.nair@mail.mcgill.ca

Labeling Functions

The labeling functions used in the paper are specified in labeling_functions.py and the code for obtaining weak labels are also included. The code for building the graph from the ads is in build_graph.py

Cite

Please consider citing our work if you find it useful,

@inproceedings{nair2024t,
  title={T-NET: Weakly Supervised Graph Learning for Combatting Human Trafficking},
  author={Nair, Pratheeksha and Liu, Javin and Vajiac, Catalina and Olligschlaeger, Andreas and Chau, Duen Horng and Cazzolato, Mirela and Jones, Cara and Faloutsos,    Christos and Rabbany, Reihaneh},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={38},
  number={20},
  pages={22276--22284},
  year={2024}
}