Skip to content

readme.md

Latest commit

 

History

History
executable file
·
44 lines (36 loc) · 2.2 KB

readme.md

File metadata and controls

executable file
·
44 lines (36 loc) · 2.2 KB

Social Bot Detection

Social bot detection based on our FediData dataset leverages user profiles, posts, images, and social connections to accurately identify social bots. We have implemented several representative baselines.

Files

  • data_split.py : Script for splitting the original dataset into training, validation, and test sets.
  • data_vis.py : Script for visualizing data distributions and feature analysis.
  • feature_extract.py : Script for extracting numerical, categorical, text, and image features from raw data.
  • final_results.py : Script for aggregating the final results from different models.
  • top_bottom3_results.py : Script for summarizing the top and bottom 3 performing instances for each model.

Folders

  • figs/ : Stores various visualization charts and analysis result images.
  • models/ : Contains model-related code.
    • BECE.py : Implementation of the BECE model.
    • BECE_train.py : Training and evaluation workflow for the BECE model.
    • BotRGCN.py : Implementation of the BotRGCN model.
    • BotRGCN_train.py : Training and evaluation workflow for the BotRGCN model.
    • sgbot_model.py : Training and evaluation for the SGBot model (Random Forest based).
    • xgboost_model.py : Training and evaluation for the XGBoost model.
  • results/ : Stores test results (CSV files) for each model.
    • bece_test_results.csv : BECE model test results.
    • botrgcn_test_results.csv : BotRGCN model test results.
    • sgbot_test_results.csv : SGBot model test results.
    • xgboost_test_results.csv : XGBoost model test results.

Usage

Data Preparation

  1. Use data_split.py to split the raw dataset into training, validation, and test sets.
  2. Extract features using feature_extract.py.
  3. Visualize data distributions and feature statistics with data_vis.py.

Model Training and Evaluation

  • Train and evaluate the BECE model using BECE_train.py.
  • Train and evaluate the BotRGCN model using BotRGCN_train.py.
  • Train and evaluate the SGBot model using sgbot_model.py.
  • Train and evaluate the XGBoost model using xgboost_model.py.

Results Analysis

  • Aggregate final model results with final_results.py.
  • Summarize top and bottom 3 performing instances per model using top_bottom3_results.py.