Latent Dirichlet Allocation for Uncovering Fraud Cases on Twitter

Abstract

Fraud is a phenomenon that continues to exist in society with a modus operandi that continues to evolve with the times. The mode of operation of fraud is continually evolving with technological advancements, globalization, and consumer behavior shifts. In today's digital age, social media is important in spreading information regarding fraud. Twitter is a social media platform that is widely used. Twitter provides easy and fast access to relevant information. As a result, to raise fraud awareness, it is critical to study the mode of operation of fraud spread on social media, particularly on Twitter. The Latent Dirichlet Allocation (LDA) approach is used in this work to classify and identify fraud issues often addressed by Indonesian Twitter users. By applying LDA modeling, this study aims to understand more comprehensively the fraudulent topics that often appear on Twitter. The research found that seven fraud topics are most commonly discussed by Twitter users in Indonesia, with the highest cohesion value of 0.491899.