NeuroFit: enhancing fashion recommendations through multi-modal graph neural networks with cold-start user handling

Abstract

E-commerce has transformed consumer shopping habits significantly, enhancing accessibility yet introducing complexity in navigating extensive product catalogs. Personalized fashion recommendations have emerged as a crucial element in improving customer experience and boosting sales. However, traditional recommendation systems (RS) often struggle with cold-start scenarios, where new users or items possess limited interaction histories, and face challenges effectively utilizing multimodal data, including visual, textual, price and demographic information. To overcome these limitations, this research presents a novel multi-modal Graph Neural Network (GNN) recommendation system, NeuroFit, designed explicitly for enhancing personalization and managing cold-start issues.