Fake News Detection System Using Django and Machine Learning

• Staff dashboard with cross-user visibility in this final year project
• Staff overview with all users’ prediction data
• Staff history view with all visible records
• Staff reports with cross-user filters
• Extra User column in staff tables
• Filter reports by user ID
• Staff can delete records visible in their queryset
• Django admin access for superusers/staff
• Manage all prediction results from Django admin
• List prediction entries
• Search prediction entries
• Filter by user
• Filter by input type
• Filter by prediction
• Filter by date
• Add prediction result manually from admin
• Change prediction result from admin
• Delete prediction result from admin
• Standard Django superuser support
• Create superuser using command line
• Management command for model training
• Management command for seed data
• Standard Django migrate command support

• User registration in this final year project
• User login and logout
• Email-based password reset
• Auto login after registration
• Public landing page
• Detection form for logged-in users
• Paste news text for detection
• Submit article URL for detection
• Optional country field
• Optional India state field
• Optional news category field
• Text input validation
• URL article fetching
• HTML extraction using BeautifulSoup
• URL authenticity heuristic support
• Country hint inference from URL
• Fake/Real prediction display
• Confidence score display
• Hybrid ensemble probability display
• Logistic Regression probability display
• Random Forest probability display
• Cleaned-text preview
• URL authenticity information display
• Store each prediction in database
• User dashboard overview
• View total detections
• View fake vs real breakdown
• View model trained/not trained status
• View latest prediction records
• View own prediction history
• Search history by text, URL, prediction, country, state, and category
• Delete own prediction records
• Dashboard reports with charts
• Filter reports by date
• Filter reports by country
• Filter reports by category
• Filter reports by prediction
• Filter reports by input type
• View label distribution chart
• View detections over time
• View category/country breakdown where data exists
• Access About page
• Access Contact page
• Health check endpoint support

Machine Learning Features in this Final Year Project

• Fake news classification using machine learning
• Hybrid ML prediction pipeline
• TF-IDF vectorization
• Logistic Regression model
• Random Forest model
• Probability averaging ensemble
• Fake threshold set at averaged fake probability ≥ 0.4
• Stores per-model probabilities
• Stores hybrid probability
• Trains from labelled CSV dataset
• Supports Kaggle-style text + label dataset
• Text preprocessing using NLTK
• Stopword removal with negation words retained
• WordNet lemmatization
• Train/test split support
• Accuracy score calculation
• Confusion matrix generation
• Classification report generation
• Model artifacts saved using joblib

• Saved vectorizer:

  ml_artifacts/tfidf.joblib

• Saved Logistic Regression model:

  ml_artifacts/logreg.joblib

• Saved Random Forest model:

  ml_artifacts/rf.joblib

• Saved metadata:

  ml_artifacts/meta.json

URL Detection Features in this Final Year Project

• Article URL input support
• Fetches HTML using requests
• Extracts article text using BeautifulSoup
• Uses lxml parser support
• Validates extracted article length
• Supports fallback by pasting text when URL extraction fails
• URL authenticity heuristic support
• URL country hint support
• Stores URL-based prediction records
• Displays URL authenticity details when available

Dashboard and Report Features in this Final Year Project

• User dashboard overview
• Detection totals
• Fake vs Real distribution
• Latest detection records
• Model training status indicator
• Detection history table
• Search and filter history
• Delete own detections
• Report dashboard with charts
• Label distribution chart
• Detection trend over time
• Country-wise breakdown where data exists
• Category-wise breakdown where data exists
• Input type based filtering
• Prediction based filtering
• Date range filtering
• Staff-level report visibility
• Matplotlib and Seaborn chart generation

Other Features in this Final Year Project

• Complete fake news detection source code
• Suitable for final year project, major project, and minor project
• Project report content can be prepared from included modules
• Django web application
• User authentication system
• Django admin support
• Staff dashboard support
• Text-based fake news detection
• URL-based fake news detection
• NLP preprocessing
• Hybrid machine learning ensemble
• Prediction history storage
• Dashboard analytics
• Reports and visual charts
• Password reset email support
• Email fallback folder support
• Environment configuration using .env
• Health check endpoint
• Management commands included
• Dataset training command included
• Seed data command included
• No default admin shipped in repository
• Useful for viva, source code review, project report writing, and project demonstration

• Open terminal in the project root folder containing:

  manage.py

• Create virtual environment:

  python -m venv .venv

• Activate virtual environment on Windows:

  .venv\Scripts\activate

• Activate virtual environment on Linux/macOS:

  source .venv/bin/activate

• Install dependencies:

  pip install -r requirements.txt

• Optional: copy environment file:

  .env.example

• Rename or copy it as:

  .env

• Run database migrations:

  python manage.py migrate

• Place labelled dataset at:

  data/news.csv

• For quick demo training, run:

• python manage.py train_model --csv data/sample_news.csv

• For full training, run:

• python manage.py train_model

• Optional: insert demo rows:

• python manage.py seed_data

• Optional: attach demo rows to a specific registered user:

• python manage.py seed_data --user YOUR_USERNAME

• Create Django admin account:

• python manage.py createsuperuser

• Start development server:

• python manage.py runserver

• Open landing page:

• http://127.0.0.1:8000/

• Open register page:

• http://127.0.0.1:8000/register/

• Open login page:

• http://127.0.0.1:8000/login/

• Open dashboard:

• http://127.0.0.1:8000/dashboard/

• Open Django admin:

• http://127.0.0.1:8000/admin

This repository does not include default usernames or passwords.

Django Admin / Superuser:
Create locally using:

python manage.py createsuperuser

Then login at:

http://127.0.0.1:8000/admin/

Normal App User:
Register from:

http://127.0.0.1:8000/register/

Password Reset:
Configure SMTP or file-based email backend using .env.example. Real email secrets should not be committed.