Research on Clothing Recommendation Technology Integrating Multimodal Data
Rong Lin, Jiale Tang, Yi He, Zhihang Tang · 2024
Due to the rapid increase in the volume of text data, traditional recommendation systems that rely solely on textual information are struggling to keep up with the evolving fashion trends and the personalized needs of users. In today's age of information overload, users are no longer satisfied with simple text descriptions when making clothing choices. They seek a comprehensive evaluation of information through various means such as text descriptions, images, and user comments. This paper introduces a clothing recommendation algorithm that utilizes a multimodal data mining technique, integrating it with a multi-vector semantic segmentation method for a clothing product recommendation system. By conducting in-depth multimodal data analysis and introducing topic words, the algorithm generates personalized recommendations that align with users' needs. The experimental results demonstrate the algorithm's effectiveness in reducing confusion between different topics. With the LOP parameter set to 20, the Precision and MAP values of the model proposed in this paper are better than the comparison model. This algorithm can effectively mine potential information within multimodal data, identify users' potential points of interest, and accurately predict user preferences, boosting merchandise sales and user growth on the apparel e-commerce platform.