Skip to content

We created a classifier for human vs. bot Twitter accounts, for which input metrics are based on attributes of likes-, mentions-, and follower-based local networks.

Notifications You must be signed in to change notification settings

fitzgerald-kyle/twitter-bot-detection

 
 

Repository files navigation

Twitter-Bot-Detection

This repo contains scripts to query the Twitter API and construct edgelists for networks representing follow, likes, and @mentions interactions. These scripts are:

  • code/download-mentions-script.ipynb
  • code/download-likes.ipynb
  • code/download-follows-script.ipynb

The data analysis and machine learning code is at code/data-analysis. These files are named:

  • code/data-analysis/compute-network-metrics.ipynb/
  • code/data-analysis/visualizations.ipynb
  • code/data-analysis/classification.ipynb

Other code files for data cleaning, scratchwork exploring, and fusing follows with likes/mentions are:

  • code/aggregate_weighted_edges.ipynb
  • code/read_edgelists.ipynb

Figures from exploration and network visualization are at code/data-analysis/figures.

Finally, edgelists are in the directories: follows/, likes/, mentions/. The botometer dataset for labelled human and bot accounts is at botometer-labelled-dataset/ginali-2017/.

About

We created a classifier for human vs. bot Twitter accounts, for which input metrics are based on attributes of likes-, mentions-, and follower-based local networks.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 100.0%