We’ve accumulated a decent-sized library of tweet datasets over the last three years, and decided to use them to take a first stab at training a classifier to identify different categories of Twitter activity.

cc: @ZellaQuixote
@ZellaQuixote For the first round of this experiment, we chose three categories into which to classify accounts and manually selected ~200 of our existing datasets belonging to said categories as training/test data.

The 3 categories:
- #MAGA
- #Resist/#BlueWave
- spambots
@ZellaQuixote Next, we selected 24 features with which to score the tweet datasets. Examples of metrics used for scoring:

- repetition of content
- how many times retweeted
- hashtag usage
- sites linked
- trains/lists/followback tweets
- percentage of non-original tweets
- apps used
@ZellaQuixote Next, we trained classifiers (neural networks) to identify each of the three categories, i.e.:

- #MAGA or not #MAGA
- #Resist/#BlueWave or not #Resist/#BlueWave
- spambots or not spambots
@ZellaQuixote We repeated the training until we had 1000 classifiers for each category (maga, resist, spambots) with at least 80% accuracy and a lower false positive than false negative rate.
@ZellaQuixote To classify a dataset, we ran all the classifiers on it, categorizing it as whichever category it was most frequently identified as (i.e., a dataset identified 37 times as resistance, 916 times as maga, and 137 times as spambots would be categorized as maga.)
@ZellaQuixote How well does it work on data other than the training/test data used to build the classifiers? We gave it a spin on some of the projects we’ve done over the last month, and results are encouraging - 13 of the 17 classifications seem reasonable.
Missing some Tweet in this thread? You can try to force a refresh.

Enjoying this thread?

Keep Current with Conspirador Norteño

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!