Research On Recommendation System Based On Cross-platform Social Network Resolution Analysis And Transfer Learning

With the continuous expansion of the scale of the social network service platform,it is difficult for users to browse and quickly discover the information and services they are interested in.Users not only expect to get information quickly and efficiently,but also expect the system to make intelligent recommendations.In order to improve users'satisfaction and help users realize the requirements of information filtering,filtering and automatic recommendation,various corresponding personalized recommendation algorithms keep emerging.In the field of single-platform social network recommendation,users are usually recommended by extracting their behavior patterns.In order to objectively and comprehensively reflect users' interests,hobbies,social relationships and other attributes,it is necessary to model user behavior data across social platforms.Because users' behaviors are colorful,there are different behavioral models for users across social platforms.At the same time,due to the sparsity of data,new users of a single platform have the problem of the cold startup,and now they usually use the user dialogue to solve the problem of time-consuming and inefficient.Based on the characteristics of cross-platform social network users,this paper proposes a cross-platform SPST algorithm for user entity resolution.Two improved clustering algorithms based on transfer learning(AT-MEC algorithm and TF-KPC algorithm)are proposed,which are more conducive to the processing of clustering and community partitioning tasks in social platforms.A cross-platform social network user recommendation HCPUR model is proposed to more accurately recommend users.This paper proposes a cross-platform LLS recommendation algorithm,which is applied to the business recommendation and solves the problem of cold start and data sparsity.The main research contributions of this paper are as follows:The entity resolution algorithm proposed in this paper provides an effective solution for multi-identity recognition of users in cross-platform social networks and the elimination of information islands between platforms.Based on the deep fusion of users' heterogeneous information sources in a cross-platform social network,a new user recommendation model based on the cross-platform social network is proposed.Compared with the single-platform user recommendation model,the cross-platform user recommendation model can describe user behaviors more comprehensively and accurately,improve the accuracy of social network user recommendation,and improve the stickiness of social platform to users and user experience.The cross-platform personalized recommendation algorithm is applied to the merchant recommendation platform to realize the migration learning and transfer the data from the auxiliary platform to the target platform,to solve the problem of cold start and data sparsity,and improve the user behavior prediction and related recommendation tasks.