Date: October 23, 2025
Time: 12.30 – 13.30
Place: MA -205
“Integrating Customer Shopping Behaviour into Predicting Online Return:
A Data-Driven Approach”
by
Vefa Övünç Özer
(Advisor : Assoc. Prof. Fehmi Tanrısever)
Abstract:
The continued expansion of online retail, particularly in the fashion sector, is accompanied by the significant operational and financial challenge of high product return rates. These returns impose substantial costs related to logistics, inventory management, and sustainability. To address this challenge, this study introduces an optimized clustering framework that employs optimization technique to systematically determine the ideal number of clusters and feature weights for segmenting users, products, and suppliers. By constructing high-order interaction features from these optimally predictive segments, the framework captures the non-linear relationships that drive return behavior. The results demonstrate that this approach yields a significant improvement in predictive performance over a baseline model, quantifying the value of modeling these interaction-based behavioral patterns.