https://github.com/jakobzmrzlikar/fake-news-analysis

An analysis of the FakeNewsNet dataset using NLP techniques.
https://github.com/jakobzmrzlikar/fake-news-analysis

data-analysis fake-news ipynb-jupyter-notebook nlp-machine-learning

Last synced: 3 days ago
JSON representation

An analysis of the FakeNewsNet dataset using NLP techniques.

• Host: GitHub
• URL: https://github.com/jakobzmrzlikar/fake-news-analysis
• Owner: jakobzmrzlikar
• License: mit
• Created: 2021-10-29T09:31:09.000Z (about 3 years ago)
• Default Branch: main
• Last Pushed: 2022-03-18T09:51:12.000Z (over 2 years ago)
• Last Synced: 2024-10-12T15:43:48.580Z (about 1 month ago)
• Topics: data-analysis, fake-news, ipynb-jupyter-notebook, nlp-machine-learning
• Language: Jupyter Notebook
• Homepage:
• Size: 16.6 MB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# fake-news-analysis

## This is a python project analyzing the [FakeNewsNet dataset](https://github.com/KaiDMML/FakeNewsNet) using NLP techniques. 

### Description

The Jupyter notebook contains exploratory data analysis of news articles and tweets from the Politifact part of the FakeNewsNet dataset. This includes:

- NLP preprocessing steps including cleaning, tokenization, lemmatization and stemming

- Term-Document and TF-IDF matrix construction

- visualization of PCA and T-SNE projections

- traininng and testing linear and nonlinear SVM classifiers on the data

- and much more...

### Some interesting takeaways

- 95% of news articles and 99% of the tweets containing the word 'transcript' were about real news.

- Tweets talking about presidents are mostly about real news, while tweets talking about vaccines are mostly about fake ones.

- Using SVM models trained on news articles to predict the turstworthiness of tweets is often impractical and produces mediocre results.

- The FakeNewsNet dataset is enormous, and this analysis barely touches the surface of what could be learned from it.