Twitter Bot Detection

What is this Project about

In an Undergrad Research Project, we were tasked with creating a program to find bots injected in a large dataset of tweets by our peers.

What did we do

To find the bots, we implemented many features, such as clustering based on cosine similarity, removing profanity, detecting emojis and checking grammar mistakes. After a long time of deliberating and brainstorming with my team, we found that those features were the ones that stood out the most between bots and real users.

What results did we find

For the second dataset, we had the following results:

We have found: 58 bots

There are 64 users in the users_df

90.625% of the users in users_df are bots

89.23076923076924% bots have been found

For the first dataset, we had the following results:

We have found: 2 bots

There are 2 users in the users_df

100% of the users in users_df are bots

100% bots have been found