Real-Time Fake News Detection: An AI-Powered Web Application with Trust Scores | IJCT Volume 12 – Issue 6 | IJCT-V12I6P36
International Journal of Computer Techniques
ISSN 2394-2231
Author
Abhishek Kumar, Chandra Shekhar, Rohit, Nitesh
Abstract
In order to slow down the rapid spread of false information that erodes public confidence and decision-making in digital media contexts, this work proposes a full-stack AI system for real-time categorization of news headlines and articles as authentic or fraudulent. A Python inference service containing transformer-based models (e.g., BERT) and classical (TF IDF + logistic regression) models with optional ensembles, a Node.js/Express REST API, a React-based web interface, and a MongoDB datastore for predictions and user feedback are all integrated into the platform. The system aims for at least 90% accuracy, provides immediate predictions with an interpretable trust score, and includes a feedback loop to constantly enhance detection performance through recurring dataset refreshes and retraining. In addition to operational metrics like API latency to guarantee interactive responsiveness and user engagement indicators to measure impact, evaluation makes use of Accuracy, Precision, Recall, F1, and AUC ROC. Data minimization, anonymization, and open communication of model constraints to minimize damage from misclassifications are examples of security, privacy, and ethical measures. A production-ready, cloud-deployed web application for credibility assessment is the result of a staged roadmap that addresses requirements, dataset curation and preprocessing, model development, full-stack implementation, deployment, and continuous monitoring for drift and bias.
Keywords
Misinformation, Text Classification, TF IDF, Logistic Regression, BERT, Ensemble Models, Trust Score, Real-Time Inference, AUC ROC, F1 Score, Node.js/Express, React, MongoDB, REST API, Feedback Loop, Model Retraining, Model Drift, Privacy, and Fake News Detection.
Conclusion
Countries and towns can effectively quantify, prioritize, and control food waste toward SDG 12.3 by using the EPA’s prevention-first Wasted Food Scale in conjunction with credible estimation based on the FLW Standard and UNEP’s Food Waste Index advice. Due to the present global baseline of 1.05 billion tonnes wasted, mostly by homes, and the 8–10% GHG contribution from food loss and waste, prevention, donation, and circular valuation must be scaled up immediately by policy, PPPs, and infrastructure that is specifically targeted at urban systems.
References
Food Waste Index Report 2024, United Nations Environment Programme: methodology, global estimates, sector guidelines, PPP models, and integration with NDCs and SDG 12.3. Key conclusions of the UNEP Food Waste Index 2024 In brief: 631/290/131 Mt by sector; 79 kg/capita household average; 1.05 billion tonnes overall; 19% of consumer-available o=food lost; 8-10% GHG from loss and waste; one billion meals wasted daily in households. The scope, definitions, destinations, quantification techniques, normalization/scaling, uncertainty, and reporting requirements of the FoodLoss ans Waste Accounting and Reporting Standard(FLW Standard), version 1.0. Environmental justification for deprioritizing landfill, incinerator, and down-the-drain pathways; revised hierarchy and pathway descriptions for the US EPA Wasted Food Scale that prioritize prevention, donation, and upcycling.
How to Cite This Paper
Abhishek Kumar, Chandra Shekhar, Rohit, Nitesh (2025). Real-Time Fake News Detection: An AI-Powered Web Application with Trust Scores. International Journal of Computer Techniques, 12(6). ISSN: 2394-2231.
Related Posts:
