Product personalized combination recommendation method based on dobby machine algorithm

Abstract

The invention discloses a product personalized combination recommendation method based on a dobby machine algorithm. The method comprises the following steps: 1, obtaining information and images of all commodities in a shop and demographic information and historical behavior records of a current user to be recommended; 2, initializing feature representations of commodities and current to-be-recommended users to form clusters; 3, constructing a relationship between the to-be-recommended users and the commodities, forming a commodity cluster set, forming commodity groups from the commodity cluster set, calculating corresponding estimated return values, and selecting the commodity group with the maximum estimated return value for recommendation; 4, after the to-be-recommended user receives the recommendation, real feedback is made; and 5, updating the pre-estimation return model and clustering based on real feedback, and adjusting model parameters for re-recommendation. According to the method, the recommendation diversity and accuracy can be improved, so that personalized combined recommendation is realized.

Description

technical field [0001] The invention belongs to the field of product recommendation, and specifically relates to a product personalized combination recommendation method based on a multi-armed slot machine algorithm. Background technique [0002] The rapid development of the Internet and e-commerce has led to a sharp increase in the number of users and products of online shopping platforms. The situation of information overload has plagued users and shopping platforms. The recommendation system has effectively alleviated this problem. [0003] However, the current recommendation algorithm blindly uses user historical data to make recommendations, which makes users fall into the situation of information cocoons, making it difficult to meet the diversity of recommendations and to complete the task of exploring users' potential preferences. If we can balance the use of user behavior data and the exploration of users' potential preferences, we can make the recommendation results...

AI Technical Summary

Problems solved by technology

However, the existing multi-armed bandits often ignore the context information, and most of them are single recommendation

Images