Neural Market Trends

Identifying Viral Bots and Cyborgs in Social Media

You can read this in one minute.

My notes:

  • Bots and sock puppet accounts influenced USA and French elections in 2016 and 2017 respectively
  • Spreading of propoganda and misinformation
  • Collected 60 million tweets containing “Trump”, “Russia”, “FBI”, and “Comey”
  • Used information entropy (avg amount of info produced by a probabilistic stochastic source)
  • BotorNot system used random forest model with over 1,150 features (meta data, etc)
  • Calculated two entropy scores
  • One score is based on distributions of time lags between successive posts
  • Second score is the ordering of words in posts
  • Then caluclated z-scores for time and text entropy
  • Calculated K-core

  • Network analysis using Louvain Community Detection algo
  • Most bots are unsuccessful
  • Anti Trump groups were more successful re: Manafort house raid
  • Applied to NFL kneeling related bots
Read the entire source article here.