Demand Analytics in E-Commerce Leveraging Computer Vision Algorithms

作者：Jing Wang, Anning Li, Jiaxin Zhang, Yancong Deng · 年份：2023 · DOI：10.1145/3603781.3603881 · 研究领域：Consumer Retail Behavior Studies、Color perception and design、Consumer Market Behavior and Pricing

In e-commerce marketing, consumer choice is usually an uncertain process that depends on various factors. This paper introduces deep learning models, such as YOLOv3 and MASK-RCNN, for e-commerce image content detection by utilizing e-commerce sales data. The extraction of HSV color features, identification of promotional information by CTPN + CRNN + CTC, and the use of Spearman's correlation coefficient to analyze the relationship between variables provided valuable data for understanding the relation between sales and images in the e-commerce market. Furthermore, Automl and Catboost Regressor were used to determine consumers' preferences in images, and 2SLS regression analysis was performed on the influencing factors. Additionally, we established the BLP model along with GMM structure estimation to analyze the impact on consumers' purchasing decisions. In this work, an end-to-end machine vision system was designed to turn image information into quantitative coordinates in consumer choice models, and we proposed that the best image layout with deep learning models can improve consumers' willingness to buy.