A framework for association rule learning with social media networks. (1st March 2022)

• Record Type: Journal Article
• Authors: Kruse, Ryan; Lokukatagoda, Tharindu; Alkhushayni, Suboh
• Is Part Of: IOP SciNotes. Volume 3:Number 1(2022)
• Publication Date: 2022-03-01
• Subjects: association rule learning -- Twitter -- social media -- market basket analysis -- data mining; 500
• DOI: 10.1088/2633-1357/abe9be
• Languages: English
• ISSNs: 2633-1357

Abstract

Abstract: We present an application of association rule learning to analyze Twitter account follow patterns. In doing so, we develop a basic framework and tutorial for future researchers to build on, which takes advantage of the Twitter API. To demonstrate the method, we take samples of Twitter accounts following Joe Biden and Donald Trump. For each account in our sample population, we pull the account's 100 most recently followed accounts. This data is cleaned and formatted for use with Python's apyori package, which uses the well-known apriori algorithm to learn association rules for a given dataset. This work has two objectives: (1) demonstrate the application association rule learning to social media networks and (2) perform exploratory analysis on the resulting association rules. We successfully demonstrate association rule learning in a Jupyter-notebook environment with Python. The resulting association rules indicate some interesting similarities and differences in the networks of Biden's and Trump's Twitter followers. The demonstrated method can be generalized to any non-private Twitter account(s). Extensions of our work can apply the method to larger datasets, with a focus on analyzing the learned association rules. Our study demonstrates an innovative application of association rule learning outside of the traditional use cases, which suggests similar opportunities in fields such as politics, education, public health, and more.