Store recommendation algorithm based on user-store interactive behaviors in network-physical spaces

Abstract

The invention relates to a store recommendation algorithm based on user-store interactive behaviors in network-physical spaces. The store recommendation algorithm comprises 1) an off-line user-store preference relation learning stage, in which user-store interactive behaviors in the physical space are extracted from Wifi logs generated by the users, a User-Store preference relationship is obtainedaccording to the Wifi logs, and modeling is performed on the relation between stores and store attributes; and 2) a tripartite graph and preference relationship based store recommendation stage, in which a tripartite graph about the users, stores and store attributes is built so as to predict the relationship strength between a predicted user u3 and an accessed store s1. The store recommendationalgorithm has the beneficial effect that the store recommendation model can obtain the best recommendation performance for all kinds of store recommendation, and has an advantage in digging shopping preferences of the users by comprehensively considering the interactive behaviors between the users and stores in the physical space and the network space.

Description

technical field [0001] The present invention relates to an algorithm for recommending k stores to a user based on the user's visit to the store, in particular to a store recommendation algorithm based on the network-physical space interaction between the user and the store. Background technique [0002] Online stores can mine user shopping preferences and habits from customer click logs and transaction records, while retailers in brick-and-mortar stores still lack an effective way to dig into user shopping preferences. Retailers in traditional brick-and-mortar stores mainly use manual questionnaires to understand users' shopping preferences. These manual questionnaires lack scalability due to the need for labor-intensive manpower and material resources. [0003] In order to mine the preference relationship between users and shops in this case, there are user collaborative filtering based on location co-occurrence (UCF-LC); recommendation algorithm based on matrix factorizati...

AI Technical Summary

Problems solved by technology

However, these methods only consider whether the user has visited the store, or the user's stay in the store, and only consider the interaction between the user and the store in the physical space, and this consideration is not comprehensive.

Images