Cross-social Networks User Identification Model Based On Multi-dimensional Information

In the past decade,online social networks at home and abroad have flourished,more and more people register accounts on multiple social networks.Cross-social network account identification has received more and more attention.Realizing the identification of different social platform accounts plays an important role in practical applications such as cross-platform recommendation systems and user portraits.However,there are some problems with the current account identification algorithms:most user identifications rely on one or more attribute information,and lack of in-depth analysis of user behavior patterns.In addition,most studies focus on one-to-one account identification,failing to achieve many-to-many account identification make the algorithm lack generalization.In view of these problems,this paper selects user attribute information and user behavior information to construct a similarity vector,uses random forest binary classification for account matching,and performs secondary confirmation based on the weighted similarity vector to realize account identification of two social networks.The main work of this article is as follows:1.This paper analyzes the integrity of user attribute data of well-known social networks in China,and selects features with high data integrity as the basis for account identification,which improves the generalization ability of the model.2.This paper proposes a topic model with a length factor which is combined with a dynamic topic model to measure user interest and evolution.3.This paper refines the mining of user behavior patterns,such as using the number of followers and fans to form a vector to measure the user's dating pattern,using the number of likes and articles to measure the quality of the articles,and the user activity model is divided into two dimensions:holidays and weekdays for measurement4.This paper proposes weight-based filtering.First,the user name is used to screen the accounts to be matched,then,the account matching is performed by determining whether the comprehensive similarity meets the threshold.It reduce the calculation cost and improve the accuracy of the model5.Based on the above research content,the ADTM-AI model is proposed,and the well-known social networking sites in China are selected to conduct experiments.The experimental results fully proved the validity and rationality of the model in terms of accuracy,recall rate,and comprehensive index F1.