Date of Award
2-10-2023
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Information Science
First Advisor
Nitin Agarwal
Abstract
Online social networks have been a target of misinformation and manipulative campaigns in recent years. Although, recent literature has been developed to identify individual suspicious accounts that are connected to coordinated manipulative campaigns on platform such as Twitter, there is a need to detect and characterize large-scale coordinated efforts within other contexts. YouTube is the one of the most popular and visited websites in the world and has also been a target of misinformation and online campaigns in recent years, making analysis of suspicious behaviors within the platform vital for the research community and policymakers at large. In this research, I assess YouTube data (comprising user engagement statistics such as total views, total subscribers, total comments and total videos) from 39 channels, combining techniques including rolling window correlation analysis, anomaly detection, peak detection, rule-based classification, unsupervised clustering and forecasting. The results show that channels that exhibit suspicious behaviors within their user engagement statistics are characterized by a relatively lesser number of peaks in their anomaly patterns, compared to less suspicious channels. However, more suspicious channels have a larger distribution of peaks with higher magnitudes.
Recommended Citation
Adeliyi, Oluwaseyi David, "Detecting and Characterizing Suspicious Behaviors Within YouTube Channels" (2023). Theses and Dissertations. 1117.
https://research.ualr.edu/etd/1117
