Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{198061,
        author = {Dr. Shwetambari Pundkar and Vedant N. Bhuskat and Kshitij M. Kadu and Parth R. Kale and Prerna M. Dudhe},
        title = {Fake News Detection System},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {9226-9233},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=198061},
        abstract = {The problem of fake news is a huge nightmare in the digital world nowadays, and in India, the impact is much closer and more threatening. This was evident with the COVID-19 pandemic, as fake or pseudo Science health information was spreading on WhatsApp significantly faster than even the real information, and the majority of the population had no good option of finding out whether what they were reading was even real or just another fake news. The biggest issue is that a majority of the detection tools available in market are build on the basis of Western datasets. They simply do not work with Indian news which is a combination of English and Hindi, local politics and some specific viral forwards. The current project is an endeavor to correct that A system was built using three models—LinearSVC, Logistic Regression, and Gradient Boosting—trained on the Indian Fake News Dataset (IFND). This has over 56,000 fact-checked articles that actually represent the messy data seen in real life. The system does not simply scan the text but checks such aspects as the sensationality of the headline and the possibility of the source to be credible. All these signals are inputted into a model whose decision-making system is based on the majority-vote system. This method achieved 88.4% accuracy of the 11, 343 articles in testing. I also applied SMOTE in the training process to ensure that the model was not biased on real news simply because of the sheer number of such in the data. It also has a manual override in case of health misinformation, and a gibberish check such that the system does not give a sure answer to random typing on the keyboard. Everything is put into a web app with a React frontend and a Flask backend. A user just pastes an article and gets a prediction with a simple reason why it’s flagged as real or fake.},
        keywords = {},
        month = {April},
        }