Research On Technology Of Commodity Recommendation System Based On Deep Learning

The advent of the highly information age of human society is inseparable from the rapid development of various Internet technologies,and people often rely on the Internet to obtain all kinds of information.With the continuous progress of the times,the Internet world has developed rapidly.At the same time,the data generated on the Internet is fast and large in quantity,and the huge resources have caused the overload of information.The emergence of recommender systems has provided people with a new way of thinking and played a vital role in solving this problem.In the era when the field of e-commerce is gradually evolving,the recommendation system can not only solve the problem of massive information difficult to handle,but also recommend a summary of their preferred products for users,and at the same time bring more benefits to the operation of the enterprise.This paper studies the characteristics of the e-commerce platform system and the relevant technical principles of the current product recommendation system,and conducts a comprehensive analysis of the characteristics and difficulties of the current recommendation system.In view of the insufficient shallow learning capabilities of traditional recommendation algorithms,they can usually only capture users' general static interests and preferences,and lack the interpretability of recommendations.This further leads to the improvement of integrating deep learning into product recommendation,and uses the power of deep learning.The ability to process data provides a reliable source of data for the recommendation system,while improving the accuracy,interpretability and personalization of recommendations.The main work of this paper is as follows:(1)Use graph neural network to model the user's behavior sequence and apply it to the product recommendation scenario.Introduce the gated graph neural network used in this article in detail,and use its own structural advantages to capture complex conversions between sequences to increase the interpretability of recommendation;(2)Further novelly introduce Transformer into the sequence recommendation based on graph nerves,and use its multi-head self-attention mechanism to better capture the user's interest preferences,and further enhance the accuracy and interpretability of the recommendation.Combine gated graph neural network and Transformer to build a product recommendation system model based on deep learning to provide users with more accurate and personalized product recommendations;(3)Then,on the two real e-commerce data sets of Yoochoose and Diginetica,select HR and MRR as evaluation indicators,and conduct multi-group comparison experiments with the three sequence recommendation models of GRU,SR-GNN,and SR-GNN variants,to verify the effectiveness and advantages of this model;(4)In addition,in order to better simulate and demonstrate the application of recommendations in actual e-commerce business scenarios,carry out the system design of the e-commerce platform,and visually display the recommendation results.