Twitter-bot-detection

Machine learning system for detecting bot accounts using temporal patterns, tweet embeddings, and user metadata.

Models

ML Models: XGBoost with Optuna optimization, Multi-input Neural Network

NLP: RoBERTa-based sentence transformers for tweet embeddings, BERT-base sentence transformers for profile descriptions embeddings

Features: 47 temporal/behavioral features + 384D description embeddings + 768D tweet embeddings

Temporal Analysis: Posting patterns, inter-tweet intervals, burst detection, Fourier analysis of timing gaps

Content Processing: Mean embeddings from top-10 recent tweets per user

Class Imbalance Handling: Weighted training and threshold optimization

git clone <repository-url>
cd twitter-bot-detection

pip install -r requirements.txt

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
Comparisons.ipynb		Comparisons.ipynb
NN_final.ipynb		NN_final.ipynb
README.md		README.md
Twibot_top10.ipynb		Twibot_top10.ipynb
requirements.txt		requirements.txt