Link predictability classes in large node-attributed networks

This work is an addition to the paper "Link predictability classes in large node-attributed networkss" by Antonov et al.

This research is used to evalueate methods of node attributed network generation. The generation is done based on the real world dataset: https://drive.google.com/drive/folders/1SDWWiOQ0mOtXqoEat5FRIgVNzKsE2D8v?usp=sharing

Currently we have evaluated the following methods:

• "MAG" (https://arxiv.org/abs/1009.3499, https://github.com/snap-stanford/snap/tree/master/examples/maggen)
• "CABAM" (http://nshah.net/publications/CABAM.MLG.2020.pdf, https://github.com/nshah171/cabam-graph-generation)
• "acMark" (https://www.semanticscholar.org/paper/General-Generator-for-Attributed-Graphs-with-Maekawa-Zhang/b113704d38cc9109ac4f2b24c2b897c82686d67e, https://github.com/seijimaekawa/acMark)
• "Attributed netwok generation algorithm" by C. Largeron (https://www.semanticscholar.org/paper/General-Generator-for-Attributed-Graphs-with-Maekawa-Zhang/b113704d38cc9109ac4f2b24c2b897c82686d67e, https://perso.univ-st-etienne.fr/largeron/ANC_Generator/#download)

Each method is in it's own way making a copy of a real-world network, so we evaluate each method's ability to replicate the initial attributed network. We then calculate metrics to understand the quality of method's performance. We plan to improve this research and then use it to evaluate a novel method of generation of attributed networks.

Current results: