NameBot

Its All in a Name: Detecting and Labeling Bots by Their Name

---

• authors: David M. Beskow and Kathleen M. Carley
• link: https://arxiv.org/pdf/1812.05932.pdf
• file structure:

├── preprocess.py  # extract features and convert raw dataset into standard format
└── LRclassification.py  # train model and classify Bot

• implement details:

  The original paper did not have the specific practice of Feature Engineering, so we designed feature engineering myself

  • When calculating the parameter of large (small) letters, take the percentage of large (small) letters in the length of the user name as the standardized parameter
  • When calculating TF-IDF, the feature dimension is recorded as the length of all Bi grams
  • The effect of logistic regression in the original paper is the best, so this method is used for calculation

How to reproduce:

1. In line 161 of the file LRclassification.py, change the name of dataset to what you want to train.

2. train Logistic Regression model by running:

   python LRclassification.py

Result:

dataset		acc	precison	recall	f1
botometer-feedback-2019	mean	0.6981	0.4545	0.3333	0.3846
botometer-feedback-2019	std	0.0000	0.0000	0.0000	0.0000
Cresci-2015	mean	0.7701	0.7681	0.9112	0.8336
Cresci-2015	std	0.0000	0.0000	0.0000	0.0000
cresci-2017	mean	0.7679	0.8039	0.9179	0.8571
cresci-2017	std	0.0003	0.0003	0.0000	0.0002
cresci-rtbust-2019	mean	0.6324	0.6500	0.7027	0.6753
cresci-rtbust-2019	std	0.0000	0.0000	0.0000	0.0000
cresci-stock-2018	mean	0.5584	0.5834	0.6413	0.6110
cresci-stock-2018	std	0.0000	0.0000	0.0000	0.0000
gilani-2017	mean	0.6081	0.5821	0.3645	0.4483
gilani-2017	std	0.0000	0.0000	0.0000	0.0000
midterm-2018	mean	0.8510	0.8693	0.9682	0.9161
midterm-2018	std	0.0000	0.0000	0.0000	0.0000
Twibot-20	mean	0.5722	0.5872	0.7047	0.6506
Twibot-20	std	0.0000	0.0000	0.0000	0.0000
Twibot-22	mean	0.7061	0.6773	0.0003	0.0050
Twibot-22	std	0.0000	0.0000	0.0000	0.0000

baseline	acc on Twibot-22	f1 on Twibot-22	type	tags
NameBot	0.7061	0.0050	F	Logistic Regression