Fake News Detection Using Machine Learning

Fake News Detection Using Machine Learning

The proliferation of generative adversarial networks has democratized the creation of hyper-realistic but entirely fabricated video and imagery, threatening information integrity. This defensive system trains a binary classifier to distinguish authentic media from deepfaked content by analyzing artifacts invisible to human perception. The feature extraction pipeline identifies inconsistencies in facial boundary blending, unnatural eye gaze dynamics, irregular pulse patterns in remote photoplethysmography signals, and compression artifact distributions inconsistent with natural capture pipelines. Ensemble methods aggregate weak signals across spatial, temporal, and frequency domains to produce a calibrated authenticity score. The framework is evaluated against publicly available deepfake datasets and emerging generation architectures to ensure resilience against adversarial evolution.

Components

Python 3.8+
Scikit-learn
NLTK/spaCy
Pandas/NumPy
Flask

Components	Hexkart	Flipkart
Python 3.8+
Scikit-learn
NLTK/spaCy
Pandas/NumPy
Flask

Note: The components listed are software tools and frameworks required for development.

NLP Pipeline

The NLP pipeline begins with text preprocessing: converting to lowercase, removing URLs, special characters, and numbers. Tokenization splits text into words. Stop words (common words like 'the', 'is') are removed. Lemmatization reduces words to their base form. The cleaned text is transformed using TF-IDF (Term Frequency-Inverse Document Frequency) vectorization, which converts text into numerical features weighing word importance across the document corpus.

Machine Learning Classifiers

Multiple classification algorithms are implemented and compared using Scikit-learn. Naive Bayes provides a strong baseline using probabilistic text classification. Support Vector Machine (SVM) finds the optimal decision boundary between real and fake news. Random Forest uses ensemble learning with multiple decision trees for robust prediction. The best performing model is selected based on accuracy, precision, recall, and F1-score on the test set.

Flask Web Interface

A Flask web application provides the user interface for the fake news detection system.
Key Flask features:

• Simple web form where users paste news article text for analysis.
• Real-time prediction returning Real or Fake label with confidence percentage.
• Explanation of prediction factors (key words/patterns that influenced the result).
• Batch upload option for analyzing multiple articles via CSV.

System Architecture

Key Functionalities

Data Preprocessing

• Load labeled datasets (LIAR, FakeNewsNet, Kaggle Fake/Real News)
• Text cleaning: lowercase, remove URLs, punctuation, numbers
• Tokenization, stop word removal, and lemmatization
• TF-IDF vectorization for numerical feature extraction

Model Training

• Train/test split (80/20) for model evaluation
• Train Naive Bayes, SVM, and Random Forest classifiers
• Hyperparameter tuning using GridSearchCV
• Compare models and select best performer based on F1-score

Prediction Interface

• Flask web app with input form for text paste or CSV upload
• Model prediction returning Real/Fake classification
• Confidence score and contributing keywords display