THE CLASSIFICATION OF INTERNET MEMES THROUGH SUPERVISED AND UNSUPERVISED MACHINE LEARNING ALGORITHMS**

Authors

William H. Little, Kennesaw State University, Marietta, GA 30060Follow
Pengcheng Xiao, Kennesaw State University, Marietta, GA 30060Follow

Abstract

Memes—internet phenomena—provide entertainment online and in a way that is easily transmissible. Through time-series data collected by Google Trends on memes, their trend of popularity can be graphed, and patterns ascertained. The rate and spread in which these memes become popular has been modeled in a previous study, establishing four general patterns on the post-peak popularity behavior of a meme: “smoothly decaying”, “spikey decaying”, “leveling off”, and “long-term growth”. In said study, a subset of memes was chosen from KnowYourMeme.com, with the memes’ keywords (phrases that serve as a pertinent identifier of a meme) inputted into Google Trends and the subsequent time-series data added to a dataset. The paper found that it was possible to model a meme’s popularity trend via ordinary differential equations. In this study a programmatic approach utilizing machine learning classification algorithms, supervised and unsupervised, will be employed on the time-series data. The dataset will be expanded upon from around 100 elements to over 2000 due to the necessity by ML. The unsupervised algorithm applied will be the KNN (K-Nearest Neighbors) algorithm, whose resulting clusters will then serve as labels for the supervised SVC (Support Vector Classification) algorithm. Four clusters will be used in the KNN classification, with each cluster expected to take shape to the four general patterns defined above. The resulting model trained with the SVC algorithm should be able to classify with reasonable accuracy a meme’s popularity pattern from its given time-series data. In addition to this, each meme in the dataset will be mapped to one or more categories (catchphrase, viral video, etc.). An analysis can then be done on the relationship between a meme’s category and its resulting popularity pattern.

Acknowledgements

KSU Office of Undergraduate Research

Recommended Citation

Little, William H. and Xiao, Pengcheng (2024) "THE CLASSIFICATION OF INTERNET MEMES THROUGH SUPERVISED AND UNSUPERVISED MACHINE LEARNING ALGORITHMS**," Georgia Journal of Science, Vol. 82, No. 1, Article 112.
Available at: https://digitalcommons.gaacademy.org/gjs/vol82/iss1/112