Independent research paper · 2024

A Minimal Approach to Fake News Detection

A six-month research project testing whether five article-level features could distinguish fake news from real news.

Outcome

The five-feature XGBoost model reached 71% test accuracy, and the paper was published in 2024.
A machine-learning wordmark beside a network shaped like a light bulb.
Role
Researcher and author
Year
2024
Team
Independent
Tools
Machine learning, Natural language processing
6 months

spent researching and writing the paper

01

A deliberately small feature set

Over six months, I studied how other researchers had applied machine learning to fake-news detection. I wanted to find out whether a useful model could work with as few textual features as possible.

The resulting paper reviews established approaches and presents my own minimal method. The five-feature XGBoost model reached 71% test accuracy.

02

Learning through publication

Researching, testing and writing the paper gave me practical experience with natural-language processing and machine learning.

Next

What came next.

The paper is published and available on ResearchGate.

Project gallery