Tweeple-DNA

Tweeple DNA is a FastAPI based project that classifies twitter-people's account into bot or human, as well as assigns probabilty for being a bot ranging from 0-1.

The project is based on DNA compression method from Detecting Bot Behaviour in Social Media using Digital DNA Compression, where we assign activity of the user throughout its timeline forming an activity class string, which is then converted to bytes and compressed using zlib package. To compare we use features we got from this method, and we try LightGBM and XGBoost on the features.

Dataset

Dataset was collected from Bot Repository of Botometer, an OSoMe project.

Experiment

For experimenting the project, do as the following:

• Clone the repo

git clone https://github.com/AmbiTyga/Tweeple-DNA.git

• Move to the Directory

cd Tweeple-DNA

• Go to main.py, at line 18, replace the Twitter tokens. To get the tokens refer to Developer Portal
• Create Virtual Environment. I recommend using pipenv

pipenv shell

• Install the required packages

pipenv install -r requirements.txt

• Move to /app directory

cd app

• Run the app into local system

uvicorn main:app

• Copy the host link, in my case its http://127.0.0.1:8000, you can also use http://localhost:8000. For more description you can refer to http://localhost:8000/docs.
• Try POST using Python, do the following:

>>> import requests
>>> url = "http://localhost:8000/predict"
>>> id = {'id':'ambi_tyga'}
>>> requests.post(url,json=id).json()
{'Bot Probabilty': 0.43629276266830935, 'Label': 'human'}