Abstract

This study tries to solve the problem of changing product tags on e-commerce sites in real time based on customer preferences, marketing campaigns, and internet trends, with the goal of getting more people to buy things and interact with the site. The study uses deep learning methods, such Convolutional Neural Networks (CNN), to make it easier to generate tags for recognizing product images and comparing their similarities. We created and tested a recommendation system by combining dynamic product tags with data on how people use them. The research looked at sales data for 3,132 best-selling cartoon goods from June 1, 2023, to January 31, 2024, in partnership with a large Taiwanese e-commerce site. From March 3 to August 17, 2024, the recommendation system was put to the test. The recommendation algorithm made a big difference in how often consumers interacted with the site over the 24-week testing period. In the last four weeks of the experiment, there were 36.06% more clicks, 22.91% more views, 32.29% more cart additions, 28.26% more orders, and 30.41% more payment transactions than in the first four weeks. This study adds to the field of e-commerce by showing that dynamically produced product tags powered by machine learning may improve consumer engagement and buying behavior. This is a new strategy that is different from typical manual tagging approaches.