Design And Implementation Of Clothing Recommendation System Based On Attention Mechanism And Graph Neural Network

With the development of network technology and people’s increasing material needs,most people choose to buy clothes online in the new consumption era.However,there are countless clothes on the Internet,and how consumers can quickly and accurately find the clothes they like is a new challenge.Clothing recommendation is an effective way to solve this problem.Most of the current clothing recommendation systems have a single method and seldom consider the matching function,which can no longer meet the individual needs of users.Based on this,this paper researches,designs and implements a clothing recommendation system based on attention mechanism and graph neural network.The main contents are as follows:(1)A clothing group recommendation model based on attention mechanism and category information CGAC is proposed.The research unit of this model is clothing group,that is,multiple clothing is a group.Learn the internal relationship of a shopping behavior through the correlation matrix,learn the external relationship between multiple shopping behaviors through the recurrent neural network,simulate the user’s real shopping intention based on the multi-intent attention mechanism,use the clothing category information to learn the long-term stable category of the user preferences and different preferences influenced by objective factors,and finally use the attention mechanism to adjust the feature weights.The model was tested on the Taobao dataset,and the three evaluation indicators of recall rate,F1 and HLU were effectively improved,(2)A single-piece clothing recommendation model based on graph neural network and target attention CGTA is proposed.The research unit of this model is a piece of clothing.Aiming at the scene where the user wants to buy a piece of clothing each time,two kinds of graph neural network structures,session graph and global graph,are used to encode and learn serialized user historical shopping behavior.Meanwhile,a target attentive mechanism is used to activate user-specific clothing preferences.The model was tested on two data sets of Taobao and POG,and the two evaluation indicators of recall rate and MRR were effectively improved.(3)A personalized clothing matching model based on multimodality and popularity PCMP is proposed.This model aims to match the most suitable bottoms for a certain top selected by the user.Calculate the matching degree of tops and bottoms through the visual information and text information of the clothing,combine the user’s purchase history of bottoms to understand their preference for bottoms,and grasp the popularity information of clothing from three angles of time,frequency and hot words,so as to realize personalized matching.The model was tested on the IQON3000 dataset,which effectively improved the AUC evaluation indicator.(4)Finally,a clothing recommendation system based on attention mechanism and graph neural network is implemented and tested.This system comprehensively utilizes three models of CGAC,CGTA,and PCMP,and provides users with three application scenarios:buying multiple pieces of clothing at one time,buying a single piece of clothing,clothing matching.