Date of Award
Spring 5-1-2015
Document Type
Honors Project
Degree Name
Bachelor of Science
Department
Computer Science
First Advisor
Davies, Stephen
Major or Concentration
Computer Science
Abstract
In this thesis, I explore the identification of non-human Twitter users. I am interested in classifying users by behavior into the categories of either bot or human. My goal in this research is to find an accurate and efficient means of identifying and segregating non-human Twitter users from their human counterparts. I use a two-stage data collection process to collect Twitter users suspected of being a bot and then obtain a majority vote on the suspected users to validate the suspicion. I gather, on average, 1000 tweets per user, on which I calculate 40 features characterizing the user. I explore the effectiveness of three different methods to most accurately classify users as either a bot or a human based on these features. The results of this work show that bots can be classified efficiently and with a high degree of accuracy. I show that certain features play a larger role in the classification process than others. The applications of Twitter bot identification include: (i) protecting users from malicious content (ii) spam filtering, and (iii) bot removal from Twitter data for other research.
Recommended Citation
Holster, Bryan, "Twitter Bot Detection" (2015). Student Research Submissions. 35.
https://scholar.umw.edu/student_research/35