AI-Driven Fake News Detection Platform: A Machine Learning Approach Using Passive Aggressive Classifier and Tf-Idf Vectorisation

The proliferation of digital misinformation across news platforms, social media, and messaging applications has created urgent demand for scalable, automated credibility assessment systems. This paper presents the AI-Driven Fake News Detection Platform, a full-stack Django web application that integrates a Passive Aggressive Classifier (PAC) trained on Term Frequency-Inverse Document Frequency (TF-IDF) features with a real-time browser interface and a persistent analytics dashboard. Upon receiving a text submission via AJAX POST, the system applies a clean_text preprocessing pipeline, transforms the normalised text into a sparse TF-IDF feature vector via a corpus-fitted TfidfVectorizer, classifies it as FAKE or REAL using the trained PAC, and derives a confidence percentage from a sigmoid transformation of the decision function score. Every prediction is persisted to SQLite through Django's ORM and aggregated in a SystemStats singleton that powers the analytics dashboard without requiring aggregate SQL queries at render time. Evaluation on a held-out test set achieves 92.1% overall classification accuracy, 96.4% precision on high-confidence fake articles, and a mean inference latency of 3.2 msper request, confirming the system's viability for real-time deployment on commodity hardware..