The Traveling Hacker

Chronicles of an international programmer

Make Data Analysis Great Again

Identifying millions of bots in Donald Trump's Twitter account

Unless you live under a rock, you might know that Donald Trump's Twitter account is one of the most famous and influential accounts in the world, and that he uses it very often. Whether you like him or not, you cannot deny the fact that Donald Trump's Twitter account is incredibly powerful, and that he would not have won the elections without it. But what makes it so powerful? What are its particularities?

I decided to extract information about his account, his tweets, and his followers. Yes, his followers; it is very easy to get data about anybody on Twitter, and yes, it is very creepy. In this article, I will expose the results I found, but I will not describe in detail how I did it. I will write a separate technical article describing all the details on how to extract data using Twitter's API.

So, let us get to it.

Data Analysis on Donald Trump's account

When was Donald Trump's Twitter account created?

His account was created on Wednesday, March 18, 2009 at 13:46:38.

How many followers does he have?

As of today (Saturday, May 19, 2018), he has 52 050 779 followers.

How many times has he tweeted?

He has tweeted (or retweeted) a total of 37 578 times.

Which is his most liked tweet?

With 607 116 likes and 262 223 retweets, this is his most liked tweet:

Not bad for a world leader, huh?

Which is his most retweeted tweet?

With 573 688 likes and 348 069 retweets, this is his most retweeted tweet:

Really not bad for a world leader, huh?

What is his account's background image?

Yes, Twitter accounts have a background image, probably used in older versions of the app...

Yes... Donald Trump's Twitter account background is... a picture of Trump Golf in Scotland.

Which words does he use the most in his tweets?

From his last 2825 tweets, these are the words he uses the most:

Here is the bar chart representation:

And the word cloud, because word clouds are still cool, right? Right?!

By far, the words he uses the most are: great, will and I. Well, no surprise there, I guess...

Now, let us take a look at his followers.

Data Analysis on Donald Trump's followers' accounts

Where do they come from?

Users can add a location to their Twitter accounts, it is not very common though:

The feature not being very popular, the bar chart is disproportioned:

Let us get rid of the None value to visualize the rest better:

I was really not expecting to see Egypt, Nigeria, or even Bangladesh in the top 20!

Who are his most followed followers?

It is funny to see that none of his followers have more followers than him.

How many of his followers are verified?

Only 50 374 of his followers are verified. This is normal, because in order to get verified, you need to be a famous brand or a celebrity. However, things start to get fishy in the next point.

How many of his followers seem fake (bots)?

We all know there are tons of bots on Twitter, so I wanted to make a quick check on Donald Trump's Twitter account. Like I mentioned before, I extracted data from 50 941 578 Donald Trump's followers and injected it into a database (more on that on my next article, which will be much more technical).

First, I took a look at his followers page, and tried to find accounts that look fake. Just by looking at the first row of followers, you can recognize a pattern:

tricia's and Zion's accounts look extremely fake, and their usernames follow the same pattern: they both finish in 8 digits: tricia15588470 and Zion00290666. By the way, since I started writing this part of the article, tricia15588470's account has been flagged already:

Zion00290666's account has not been flagged yet, but it still looks extremely fake to me:

These two accounts seem to be bots, but if you continue scrolling down Donald Trump's followers page, you will find tens, even hundreds of accounts following this 8-digit pattern at the end of their usernames.

Remember I said I injected all of his followers data into a database? Well, using a simple SQL query, I can count all the accounts that follow the 8-digit pattern:

5 196 452 of Donald Trump's followers accounts follow this pattern. That is more than 10% of all his followers.

From these 5 196 452 accounts, 4 354 251 have less than 10 followers.

Now, I am not saying these are bots, I am just saying these are bots.

Also, do notice that this is just a tiny fraction of all the bot-like usernames in Donald Trump's followers list. Some other patterns are more difficult to express in a SQL query with regular expressions, because they are a little bit more random than the 8-digit one. For instance, you have:


User 赵宇 with username sLFMiysIsdWxj6n and user ساهر سعيد with username nZ9DC2knhReJL7d look pretty fake to me, but there is no clear pattern here, only the length: 15 characters, but that is not enough to identify a bot because real usernames could also be 15-character long. Maybe there is something about the small amount of vowels or something like that? If anyone has a nice regular expression that would match this kind of usernames, feel free to share it.

When were most of these accounts created?

Most of them where created in 2017 and 2018, but we can also see that some 645 623 of them were created in 2016. Election-year, anybody?

And here is the same data, truncated to month:

Some final thoughts

Twitter is an extremely powerful tool, and Donald Trump knows how to use it. It was fun to get some insights from his Twitter account, especially the bots parts, but do not think he is the only one doing it, many politicians and stars like Barack Obama, Hillary Clinton, Kim Kardashian or Justin Bieber also have millions of fake followers.

The millions of accounts identified in this article seem fake, but there is no deterministic way to be sure, you can never be 100% sure that a user is a bot, and that is why social networks are extremely careful when trying to ban fake accounts. You really do not want to start banning real accounts by mistake, users will get very mad, Instagram tried in 2014 and failed miserably.

Also, I am not convinced that these huge companies want to get rid of all the bots, because they generate traffic like posts, likes, shares and comments, and these are key when trying to show shareholders that the company is doing well.


So you like traveling? Hacking? Squirrels? Subscribe to my newsletter!

blog comments powered by Disqus