What Is the Role of Data Science in the Context of Web3? | by Rohith Teja | Mar, 2022

two bitcoins and a chart behind them
Photo by Pierre Borthiry on Unsplash

Supervised Problem

  1. Web Scraping— Bitcoin users sometimes openly share their public key on the crypto forums. It can be for several reasons like hoping to receive a donation, asking a question, etc. This makes our job easy, and we can use this information to match the public key with the user ID of the forum.
  2. Transaction Fingerprinting — Here, we use transactional metadata to pinpoint the user with the highest accuracy possible. For instance, over the forums we find some chat discussing the transfer of X number of Bitcoins. We then move to our transactional database and cross-verify the approximate time and date of the transfer. The public key which matches with the transaction is picked this way.

Unsupervised Problem

  1. Features relating to currency — amount sent, received.
  2. Features relating to network — In-degree, Out-degree, clustering coefficient.
TSNE Visualization of Bitcoin transactional sub-graph
***** RF MODEL *****
ACC: Train: 1.0 Test: 0.987
ROC: Train: 1.0 Test: 0.935
F1: Train: 1.0 Test: 0.993
***** GCN MODEL *****
ACC: Train: 0.98 Test: 0.975
ROC: Train: 0.9 Test: 0.894
F1: Train: 0.99 Test: 0.986
TSNE Visualization of Bitcoin transactional sub-graph (Unknowns predicted)

Leave a Comment