Personalized Recommendation Algorithm Integrating Social Relationships

The rapid development of the Internet and artificial intelligence technology provides a timely opportunity for the dissemination and effective use of information,but with the accumulation of a large amount of information,the problem of information overload has become an important problem that plagues people,and personalized recommendation algorithms have emerged.The recommendation algorithm can not only save users' search costs and enhance user experience,but also recommend valuable products to users and enhance the enthusiasm of physical manufacturers.Therefore,how to recommend valuable products to users with real needs is an important topic in the research of recommendation algorithms.The personalized recommendation algorithm fused with trust information can not only transmit information between friends,but also enhance the algorithm's predictive performance.Currently,personalized recommendation algorithms have some problems: First,some algorithms only use rating information and trust information to limit the potential characteristics of users to achieve information transfer between friends,ignoring the importance of product information,that is,they have not thought about restricting products.Potential features are used to achieve information transfer between products;secondly,some users have fewer friends and fewer clicks or purchases of products,resulting in sparse data and severely hindering the recommendation performance of the algorithm;again,explicit social interaction There is a noisy problem in the relationship,that is,the information transmission between the user and the fake friend may hinder the optimization of the algorithm;then,there is also a preference deviation between the user and the trusted friend,that is,the human preference is not single,and It is diverse;in the end,there are dynamic changes between users and friends,and only modeling the static relationship between users and friends does not conform to real life.In order to solve the above problems,this paper conducts research on personalized recommendation algorithm integrating social relationship,and conducts experiments and analysis on real data sets.The specific research content is as follows:1.In order to realize the transfer of information between trusted users and related products,a social recommendation algorithm integrating product information is proposed.This model first builds product relevance on the score information,and then uses random walk and The Skip-Gram method builds deep product relevance,and then uses graph convolutional neural network to learn the deep relevance between scoring information,trust information,and products to obtain the deep features of users and products,and finally use the optimized matrix decomposition method to deliver the product The relationship between the correlation and the trust relationship between users,to achieve user preference prediction.2.In order to solve the problem that matrix decomposition cannot directly express user preferences,this paper further uses Mahalanobis distance to predict user preferences,and integrates a variety of information to realize the relationship between people and products,between people and people.Information transfer between products.Experiments show that the above algorithm can effectively predict the preferences of ordinary users and cold start.3.In order to construct the trust relationship between users in a timely manner,a social recommendation algorithm of fusion-generating adversarial graph convolutional neural network is proposed.This model first builds more reliable potential friends based on rating information and explicit social information Then use the efficient graph convolutional neural network to learn the structural characteristics of the score information,obtain the deep-level characteristics of users and products,and finally use the generative confrontation network to dynamically construct trusted friends with the same preferences as the user,and punish false friends,To achieve dynamic changes between users and friends,and at the same time in order to ensure that the selection of friends is random,an automatic encoder is used to rebuild the friendship.Experiments show that the algorithm can effectively recommend suitable products for ordinary users and cold-start users.