Thread by @fs0c131y

<Thread>

Few days ago, I captured more than 600K tweets on trendy topics on the Indian Twitter. I was looking at this graph yesterday night and asked myself: How can I found spammy Twitter accounts in this dataset?

1/

In my definition, a spammy account is a Twitter account with:
1) a high degree out but a low degree in. It is tweeting a lot but nobody interact (RT, like, comment) with his tweets
2) a high "tweet speed". tweet speed = total of tweets / twitter account age

2/

This graph represents a capture of tweets on an time interval. Moreover, I captured only the tweets which contain a particular hashtag. So this view is very incomplete.
If a spammy account was not active on this topic or during the time interval I captured, I will miss it.

3/

Due to the nature of the dataset we can redefine the initial question to:

How can I found spammy Twitter accounts which were active on the topic I listened, when I listened?

4/

The 1st critera allows us to filter the dataset and extract a list of potential spammy Twitter accounts. You have to calculate the ratio degree_in / degree_out.

5/

https://twitter.com/fs0c131y/status/1217390479647629313?s=20

As an example a well known Twitter account will have:
- a high degree in: people interact a lot with his tweets
- a low degree out: he is not tweeting a lot

The spammy account will have:
- a low degree in: nobody interact with him
- a high degree out: he is tweeting a lot

6/

So all the spammy Twitter accounts will have a ratio degree in / degree out < 1

7/

So we have a list of potential spammy accounts. Then I downloaded the number of tweets of each account and their twitter age. With this new data, I calculated their "tweet speed" = status_count / age (in hours)

8/

I obtained a list of 287 Twitter accounts. Initialy, the dataset contains 176837 Twitter accounts

9/

Look who are in the top of the ranking my two friends @Ravi78161144 and @Ravi17121038! Our love story started few days ago

10/

https://twitter.com/fs0c131y/status/1216649761492738048?s=20

@Ravi78161144 @Ravi17121038 Number 3 is @RickySH7707. An INC supporter with a tweet speed of 28 tweets per hour. In 1 month he tweeted more than me in 2 years (and I'm tweeting a lot)...

11/

@Ravi78161144 @Ravi17121038 @RickySH7707 Number 4 is @SuvarnaVeerappa. He joined Twitter in December 2019 and tweeted 18.7K times...

12/

@Ravi78161144 @Ravi17121038 @RickySH7707 @SuvarnaVeerappa The fun part. @rajkshroff is the number 30 of the list. He joined Twitter 10 years ago, in December 2009. He tweeted 723,3K times which is equivalent to 8 tweets every hour, 24/7, during 10 years...

12/

@Ravi78161144 @Ravi17121038 @RickySH7707 @SuvarnaVeerappa @rajkshroff Number 15 is @puriyash41
- 630.5K tweets
- Joined in March 2014
- 12 tweets per hour

@Ravi78161144 @Ravi17121038 @RickySH7707 @SuvarnaVeerappa @rajkshroff @puriyash41 The cool part is to play with the data and compare things. It's interesting to analysed the Twitter accounts with the higher ratio age / tweet speed. They probably developed nice strategies to bypass the Twitter detection

14/

@Ravi78161144 @Ravi17121038 @RickySH7707 @SuvarnaVeerappa @rajkshroff @puriyash41 In this category we have @satirist_indian (522561 tweets), @tarungarg87 (395380 tweets), @puriyash41 (630515 tweets), @jd32768519 (169492 tweets), @Arif74890482 (71669 tweets), @satirist_indian (522561 tweets), @sanjai2july (538186 tweets), @ThirupathiINC (65358 tweets)

15/

@Ravi78161144 @Ravi17121038 @RickySH7707 @SuvarnaVeerappa @rajkshroff @puriyash41 @satirist_indian @tarungarg87 @jd32768519 @Arif74890482 @sanjai2july @ThirupathiINC You can find the full list here pastebin.com/wv6GzY6a

16/16

Note 1: I didn’t use the word “bot” on purpose in this thread. I prefer to called this type of account “spammy”.

Update: @Ravi78161144, @Ravi17121038 and @RickySH7707 has been suspended by Twitter