News Authenticity Classification System

Summary

An end-to-end ML system for news authenticity classification, built on 20,000+ news samples with real-time prediction access through a Telegram chatbot.

Feature List

• Built data ingestion, preprocessing, TF-IDF feature extraction, model training, and deployment workflow
• Evaluated multiple algorithms and selected Logistic Regression based on performance analysis
• Achieved 91.2% accuracy on the classification task
• Deployed real-time predictions via Telegram API, reducing manual verification effort by about 70%
• Wrote modular Python code with clear separation of concerns for future data-source expansion