Thread by @jtk_III

As you may know I have been scraping & compiling tweets about Donald Trump weekly

#Vader Sentiment Analysis says that this latest batch of tweets, based on the same keyword list, classified the majority of the tweets as POSITIVE

pos 61608
neg 53446

This bucks the trend

The averages are interesting, a sort of inverted bell curve where most tweets have a few re-tweets, views etc. and then the opposite end of the spectrum where a minority have a TON of views, likes, etc. or none at all

(mean)Average

Likes: 12
Retweets: 3
Replies: 1
Views: 2,828

The replies seems weirdly low, I'll double check that. Views are easily the data point with the highest volatility.

most replies: 1,334
most retweets: 21,321
most likes: 72,795
most views: 133,790,475

So it still looks likes replies are the rarest interaction maybe.

I was wondering how there were so many tweets that had 1 view, no likes, etc. - and I think the answer is deleted tweets :)

So 5.3% of the tweets came from verified users.

I thought something was wrong with my data collection techniques. No, but there are many accounts like this that all tweet *exactly* the same thing on the same day, etc.

Most popular locations?

United States 3172
USA 1746
#Florida 952
Washington, DC 878

And then there are accounts that quote tweet and / or reply the exact same thing repeatedly...

In case anyone wanted to remove all the emojis from a #pandas dataframe:

df = df.astype(str).apply(lambda x: x.str.encode('ascii', 'ignore').str.decode('ascii'))

I was wondering how good the Vader sentiment analysis was, as it was hard to tell from the 10,000 ft. view. But it's easy to see how good it is when you sort by it: