Year: 2022
Type: article
Abstract: With the rise of social e-commerce and live broadcast e-commerce, real-time recommendations based on consumers’ browsing behavior are becoming more and more important. Traditional utility recommendation has some problems such as cold start and subjectivity. And traditional sequential recommendation relies on behavior sequences. However, these data contain some irrelevant data of products, which will lead to wrong dependencies and affect the accuracy of recommendation. To solve such problems, this research builds a Commodity Utility-Behavior Sequence (CUBS) dual-utility model. CUBS consists of two models: Commodity Utility (CU) and Behavior Sequence (BS). The Commodity Utility can evaluate the psychological motivation of consumer and transform it into commodity utility. The Behavior Sequence can predict preference by calculating the behavior sequences. CUBS combines the advantages of the Commodity Utility and Behavior Sequence. At the same time, it makes up for the shortcomings of only using a single model. A total of 22,417 records of 214 customers are randomly selected from the JD.com database as the test set. These customers are divided into 4 groups and calculated by CU, BS and CUBS respectively. The results show that the accuracy of CUBS model is the highest. This also confirms that customers’ click behaviors and behavior sequences have an important influence on purchase intention prediction.
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,200
OpenAlex ID: W4281290731