This project detects "fake social media accounts" using Machine Learning (Random Forest classifier) and a "Streamlit-based UI". It classifies accounts as "fake or real" based on features like:
- Report count
- Follower/Following ratio
- Verification status
- Engagement patterns
- Comment text analysis
The app supports "CSV uploads" for batch detection.
- Python (Scikit-Learn, TensorFlow, Pandas, NumPy)
- Machine Learning (Random Forest)
- NLP (Comment text analysis)
- Streamlit (User Interface)
Algorithm: Random Forest Classifier Features Used:
- follower_following_ratio
- report_count
- verification_status
- comment_sentiment_score
- engagement_rate
Dataset split: 80% training 20% testing
- Accuracy
- Precision
- Recall
Ensure you have Python 3.8+, and install dependencies: bash pip install -r requirements.txt
bash streamlit run app.py
Upload a CSV file → View fake account usernames
- app.py # Streamlit UI
- model_training.py # ML Model Training Script
- Final.csv # Training Data
- testing_dataset.csv # Testing Data
- requirements.txt # Required Python Packages
Upload "testing_dataset.csv" to verify that the model correctly flags fake accounts.
- Improve dataset quality by collecting real-world data.
- Optimize the ML model to achieve "higher than 60% accuracy".
- Experiment with advanced models like transformers (BERT/GPT) for better NLP analysis.
- Develop a "mobile-friendly version" of the UI
This project is open-source under the "MIT License".
For questions or contributions, reach out via GitHub Issues.