ARS Session 03 - Sumit Chahar

Title: Effect of Training-Data Size on Rare Customer-Segment Fidelity and Privacy in Synthetic E-commerce Data Course: DS687 Description: This project evaluates how training-data size influences the fidelity and privacy of synthetic e-commerce data, with a particular focus on rare customer segments. Using a public e-commerce dataset, the study compares Gaussian Copula, CTGAN, and TVAE to evaluate their ability to preserve uncommon purchasing patterns while limiting potential privacy risks.