APPLICATION OF ARTIFICIAL INTELLIGENCE IN FASHION RECOMMENDATION SYSTEMS FOR SUSTAINABLE APPAREL DEVELOPMENT

Abstract

This paper presents a conceptual architecture for an artificial intelligence-based fashion recommendation system oriented towards sustainable apparel development. The main functional components of the system are analysed, including data acquisition, garment detection and segmentation, visual and semantic feature extraction, user and context modelling, multi-criteria ranking, virtual try-on, and continuous feedback processing. The analysis shows that convolutional neural networks are effective for clothing recognition and classification, whereas hybrid and ensemble architectures provide broader opportunities for combining visual characteristics, user preferences, anthropometric information, and contextual factors. However, high classification accuracy on a simplified dataset should not be interpreted as evidence of effective real-world personalisation. The proposed approach treats fashion recommendation as a multi-objective task combining preference relevance, fit compatibility, contextual suitability, sustainability attributes, and predicted return risk. Such systems may support more informed product selection and demand-oriented apparel development, but their environmental contribution depends on reliable product data, responsible recommendation objectives, user privacy protection, and integration with production planning.