Fake News Predictor

A multi-source, RAG-enhanced fake news detection system using ML models, retrieval-augmented generation, and AI-powered analysis. Combines news API verification, knowledge base retrieval, and advanced analytics in a Streamlit web app.

Features

• Multi-API News Verification: Checks news existence across NewsAPI, GNews, CurrentsAPI, ContextualWeb, and Google Fact Check.
• RAG Knowledge Base: Ingests facts from datasets and APIs, supports semantic retrieval (ChromaDB + SentenceTransformers, TF-IDF fallback).
• ML Model Analysis: Uses multiple trained models (Naive Bayes, Logistic Regression, Random Forest, CatBoost) for prediction.
• AI Assessment: Integrates Gemini AI for deep analysis.
• Content Validation: Ensures only news-like content is analyzed.
• Bulk Dataset Ingestion: Supports streaming ingestion from large CSVs (True.csv, Fake.csv).

Setup

1. Clone the repository

   git clone https://github.com/muhammadnavas/Fake_News_Predictor.git
   cd Fake_News_Predictor

2. Install dependencies

   pip install -r requirements.txt

3. Configure API keys
   • Copy .streamlit/secrets.toml.example to .streamlit/secrets.toml and fill in your keys:

     NEWSAPI_KEY = "your_newsapi_key"
     GNEWS_KEY = "your_gnews_key"
     CURRENTS_KEY = "your_currents_key"
     CONTEXTUALWEB_KEY = "your_contextualweb_key"
     GOOGLE_FACTCHECK_API_KEY = "your_google_factcheck_key"
     GEMINI_API_KEY = "your_gemini_key"

   • Or set keys in .env (see .env.example).
4. Run the app

   streamlit run app.py

Usage

• Enter a news headline or article in the input box.
• The app validates content and runs analysis using selected methods (API, RAG, ML, AI).
• View results in tabs: Verification, RAG Analysis, ML Models, AI Assessment, Summary.
• Use the sidebar to ingest datasets and manage the knowledge base.

Troubleshooting

• No news fetched: Ensure API keys are set and match expected names in both secrets and code.
• Analysis blocked: Input must be news-like (headline or article, not personal/casual text).
• Module import errors: Run pip install -r requirements.txt to install all dependencies.
• Secrets parse error: All values in .streamlit/secrets.toml must be quoted strings.

Security

• Never commit real API keys to version control. .gitignore excludes .env and .streamlit/secrets.toml by default.
• Rotate keys if accidentally exposed.

Project Structure

• app.py — Main Streamlit app
• rag_system.py, rag_pipeline.py — RAG logic and pipeline
• ml_analysis.py — ML model loading and analysis
• content_detector.py — Content validation
• fetch_news.py — News API integration
• models/ — Pretrained ML models
• chroma_db/ — ChromaDB persistent storage
• fact_database.json — Knowledge base facts

License

MIT (or specify your license)

Repository

• GitHub: https://github.com/muhammadnavas/Fake_News_Predictor.git
• Owner: muhammadnavas (MrHidey)
• Branch: main

Credits

• Built by MrHidey (muhammadnavas) and contributors
• Uses open-source libraries: Streamlit, scikit-learn, chromadb, sentence-transformers, Google Generative AI, etc.

---

For questions or contributions, open an issue or pull request at the GitHub repository.