Latent Dirichlet Allocation and Predatory Pricing Online Data

Journal Title
Journal ISSN
Volume Title
Department of Applied Economics and Statistics, University of Delaware, Newark, DE.
In this paper, we study Latent Dirichlet Allocation (LDA; Blei et al., 2012) for topic modeling of Amazon unfair pricing data during Covid-19. A topic model is designed to capture topics relating to words in text document or corpus. LDA is a generative probabilistic model with helping to collect topics from discrete data, like text & corpora. It is also known as a three-level hierarchical Bayesian model, where each item of the collection is modeled as a nite mixture over an underlying set of topics. For each topic, it is modeled as an in nite mixture on an underlying set of basic topic probabilities in turn. We conduct analy- sis of unfair pricing data by sellers from Amazon during the Covid-19 period using LDA. Speci cally, we perform topic modeling and generate topics under Amazon product description. Our goal is to capture information and topics on what kind of surgical masks and products are in unfair pricing during Covid- 19. Finally, we conclude that N95 is the most unfairly priced product under the topic modeling. By generating graphical illustrations with the Python pyL- DAvis package, we are able to summarize and provide more detailed information based on Predatory Pricing Online model.
Topic modeling, Natural language processing, Latent Dirichlet Allocation, Predatory pricing online data, COVID-19