2026-05-20T23:29:14Zhttps://keep.lib.asu.edu/oai/request

oai:keep.lib.asu.edu:node-1981522025-02-11T20:39:20Zoai_pmh:alloai_pmh:repo_items

198152 https://hdl.handle.net/2286/R.2.N.198152 http://rightsstatements.org/vocab/InC/1.0/ http://creativecommons.org/licenses/by-nc-sa/4.0 2024-12 28 pages Le, Justin Motsch, Sebastien Brust, Johannes Barrett, The Honors College School of Mathematical and Natural Sciences School of Mathematical and Statistical Sciences Text We introduce the diffusion model as a means to generate new samples in imbalanced datasets. Diffusion models (and more broadly, generative models), have recently been adopted for tasks such as art generation (Stable Diffusion, Dall-E) and text generation (ChatGPT). Our aim is to train diffusion models that generate synthetic data on a credit card fraud dataset to help train a classifier to distinguish fraudulent credit card transactions from legitimate transactions. Before discussing the model itself, we will briefly review the theory behind diffusion models, namely, the process of reversing the dynamics of a stochastic differential equation. Machine learning Differential Equations Mathematics Data Science Probability Statistics Diffusion Models to Alleviate Class Imbalance