https://github.com/wyldelarue/tweetpredictor

Is it possible to predict the political party of a twitter user? A model to classify and make inferences on the political party of politicians' twitters.
https://github.com/wyldelarue/tweetpredictor

data-science r

Last synced: about 1 year ago
JSON representation

Is it possible to predict the political party of a twitter user? A model to classify and make inferences on the political party of politicians' twitters.

• Host: GitHub
• URL: https://github.com/wyldelarue/tweetpredictor
• Owner: WyldeLaRue
• Created: 2017-11-27T18:37:09.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2018-10-18T23:22:49.000Z (over 7 years ago)
• Last Synced: 2025-02-28T10:55:15.235Z (over 1 year ago)
• Topics: data-science, r
• Language: R
• Homepage:
• Size: 109 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# TweetPredictor

Can we predict the political party of a twitter user?        

We fit a logistic model to classify and make inferences on the political party of politician's twitters. 

## PCA & Clustering

![](presentation_files/figure-markdown_strict/unnamed-chunk-2-1.png)

## Native Logistic Model

![](presentation_files/figure-markdown_strict/unnamed-chunk-3-1.png)

    ## [1] 0.05140878

Consideration

-------------

Skree plot - are there exactly 2 parties?

![](presentation_files/figure-markdown_strict/unnamed-chunk-5-1.png)

Consideration Continued

-----------------------

![](presentation_files/figure-markdown_strict/unnamed-chunk-6-1.png)

Weaknesses of Previous Model

----------------------------

-   Naive

Improved Model (Trees)

----------------------

![](presentation_files/figure-markdown_strict/unnamed-chunk-7-1.png)

    ## [1] 506 253 126  63  32  16   8   4   1

Visualize Improved Model

------------------------

![](presentation_files/figure-markdown_strict/unnamed-chunk-8-1.png)

Outliers & Interesting Things

-----------------------------

-   Who are these users scoring extremely low on PC1 and PC2?

-   What are the 5-clusters encoding? What’s up with the density on

    cluster 3?

Method Continued

----------------

To expand upon this research question, “Can you predict the political

party of a twitter user using their tweet history,” some improvements

could be made:

-   Training and testing on members of the population to make a model

    that works more generally, instead of on politicians due to charged

    language.

-   Considering different scalings of the data such as quantile-based

    instead of ratio-based numerics.

-   

Fun Takeaways

-------------

![](images/all_words_dot.png)

Fun Takeaways

-------------

![](images/hash_tags_dot.png)

Fun Takeaways

-------------

![](images/hash_tags_bar.png)