Ask HN: How would you go about doing a cluster analysis on Tweets?

I'm trying to do some analysis on which Tweets with which features.

For example, I'd like to know if a Tweet gets more engagement if it has hashtags or gifs or movies or mentions, etc. There's been things like this before: https://www.quicksprout.com/2014/03/05/what-type-of-content-gets-shared-the-most-on-twitter/ But their methodology isn't shared and might not be an accurate analysis. I'm also not sure if its results are valid 4-5 years later.

I'm curious how others would go about this. Would you just take the top 10% of a bunch of tweets and just do some basic counts on a single variable: has media, does not have media. And just visualize which variable makes the impact?

Or would you explore how maybe multiple variables interact? Would you do some kind of k-means cluster analysis? All pointers to education, methodologies, helpful tools/software and previous attempts at this are very welcome.