Research On Product Personalized Recommendation Algorithm And Its Application Based On User Portrait

In recent years,with the rapid development of the Internet,video websites,ecommerce websites and live broadcast websites have sprung up,and all walks of life have begun to promote their products on the Internet.On many e-commerce network platforms,many users will leave a lot of evaluation and purchase data for various products.In addition,websites such as Taobao and Tiktok also have many new advertising methods of live video with goods,and users have bought many goods under the influence of these shopping methods.Therefore,how to judge user preferences based on user social data has become an urgent need for personalized recommended algorithm.In view of the current situation of diversified user data,this thesis proposes a recommended algorithm based on a new user profile model:(1)By combining sociological theory,this thesis creates a knowledge map that integrates user social network data,product purchase behavior,product evaluation behavior,and anchor user product recommendation;(2)In order to make up for the problem that the traditional user portrait based on user data can not meet the needs of heterogeneous user data and can not integrate user relationships such as social networks,a user portrait model based on user commodity knowledge atlas and multi-layer graph neural network system is proposed.This model can effectively combine user social data,user focus data and user commodity data,The training efficiency of the model is greatly improved and the training cost is reduced by random sampling;(3)A recommendation algorithm based on a new user profile model is proposed to solve the problem that the recommended algorithm is separated from the relevance between users and products and other users.Finally,experiments based on multiple data sets show that the new user profile model is more comprehensive and the new recommended algorithm is feasible in industrial practice.