Fake News Detection

A full-stack project to detect whether news content is likely REAL or FAKE using text classification and image OCR.

Live Demo

• Website (GitHub Pages): https://ankit-xo.github.io/FakeNewsDetection/home

Website Screenshots

Home	Text Check

Image Check	About

Project Highlights

• Text news prediction with confidence score
• Image-based prediction using OCR + model inference
• Why-this-result explanation + fact-check tips
• API health indicator and retry UX
• Context-aware retry flow for both prediction and feedback failures
• Recent checks history (last 5 via localStorage)
• Feedback collection and contact email
• Mobile-optimized header/footer and responsive layout

Model Performance Snapshot

Metric	Value
Accuracy	98.5%
Precision	99.7%
Recall	96.9%
F1 Score	98.3%
Dataset Size	38.8K+ records
Last Trained	7 Apr 2026

Confusion matrix (validation split, positive class = FAKE):

Actual \ Predicted	FAKE	REAL
FAKE	3470	110
REAL	9	4176

Architecture

Frontend sends text/image input to FastAPI. Backend preprocesses text/OCR output, runs the model, and returns result + confidence + insights.

Tech Stack

• Frontend: React + Vite
• Backend: FastAPI
• ML: scikit-learn
• OCR: pytesseract + Pillow

Project Structure

FakeNewsDetection/
├── backend/
│   ├── core/
│   │   ├── main.py              # FastAPI app + API routes
│   │   └── train.py             # ML training pipeline
│   ├── models/                  # trained model files
│   └── feedback/                # user feedback logs
├── dataset/                     # datasets (Git LFS)
├── frontend/
│   ├── public/assets/           # logo, favicon, architecture diagram
│   ├── src/
│   ├── package.json
│   └── vite.config.js
├── docs/
│   └── screenshots/             # README website screenshots
├── Dockerfile                   # Render backend container
├── render.yaml                  # Render Blueprint config
├── Testing.md                   # QA checklist and test cases
├── requirements.txt
└── README.md

Prerequisites

• Python 3.9+
• Node.js 18+
• npm
• Tesseract OCR installed

Clone Repository

git clone https://github.com/ankit-xo/FakeNewsDetection.git
cd FakeNewsDetection

Run Locally

One Command Start

After installing Python dependencies and frontend packages once, start both backend and frontend together from the project root:

python3 run_dev.py

This starts:

• Frontend: http://localhost:3000/home
• Backend docs: http://localhost:8000/docs

If you already have a local venv/, the script will use it automatically.

1) Backend

macOS / Linux:

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
uvicorn backend.core.main:app --host 0.0.0.0 --port 8000 --reload

Windows PowerShell:

python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
uvicorn backend.core.main:app --host 0.0.0.0 --port 8000 --reload

2) Frontend (new terminal)

cd frontend
npm install
npm run dev

3) Open App

• Frontend: http://localhost:3000/home
• Backend docs: http://localhost:8000/docs

4) Quick Health Check

• API health: http://localhost:8000/api/health
• Frontend health pill should show: API Online • Model Ready

Quick Demo Flow (For Evaluator)

1. Open Text Check and click Load Sample News.
2. Click Predict and review:
   • result badge,
   • confidence bar,
   • Why this result?,
   • Fact-check tips.
3. Open Image Check, upload sample image, run prediction.
4. Test Feedback buttons (Real / Fake).
5. Disconnect backend briefly to verify error + retry UX.

Environment Variable (Frontend)

Set backend base URL before deploy:

macOS / Linux:

cd frontend
VITE_API_BASE_URL=https://your-backend-url.com npm run deploy

Windows PowerShell:

cd frontend
$env:VITE_API_BASE_URL="https://your-backend-url.com"
npm run deploy

Deploy Frontend on GitHub Pages

cd frontend
npm install
npm run deploy

Deploy Backend on Render (Docker)

This repository already includes Dockerfile and render.yaml.

1. Push latest code to GitHub.
2. In Render, create a new Blueprint or Web Service.
3. Use:
   • Runtime: Docker
   • Branch: main
   • Health Check Path: /api/health
4. Set env var:

CORS_ALLOW_ORIGINS=https://ankit-xo.github.io,http://localhost:3000,http://127.0.0.1:3000

5. Deploy and verify:
   • https://fake-news-detection-api-8zp1.onrender.com

API Endpoints

• GET /api/health
• POST /api/predict
• POST /api/predict-image
• POST /api/feedback

Notes

• Threshold for REAL classification: 0.50 (PROB_THRESHOLD)
• Dataset CSV files are tracked with Git LFS
• Render free plan may sleep after inactivity

Author

Ankit Anand

• GitHub: @ankit-xo
• Feedback email: ankitanand.works@gmail.com