User Identity Recognition Technology On Cross-Plat Form Social Networks

In recent years,various online social networks have been developing at an alarming rate and have gradually become an Indispensable part of people's online life,such as Facebook,Twitter,Linkedln and so on.However,there is often no direct connection between individual social network accounts,and as a result,a complete social user network map cannot be obtained.T he problem of cross-platform social network user identification is to identify the physical users behind each social network.At present,there are still some shortcomings concerning the problem of identity recognition.On the one hand,most user identifications rely on th e matching of user profile information and lack of personalized user behavior analysis.On the other hand,many matching algorithms of machine-learning implement one-to-one user identification,failing to achieve many-to-many user identification.T his paper focuses on cross-platform user identification of social networks.In terms of user pre-matching,this paper proposes a personalized user behavior analysis algorithm based on frequent pattern mining to analyze user blog post data for the user's blog writi ng behavior,so that the problem of user identification does not rely solely on user profile data.For the distribution of user attribute weights,this paper proposes an information entropy weight assignment algorithm based on posterior probability to assi gn corresponding weights to user's different user attribute information and improve the low accuracy rate caused by the empirical weight distribution method for user identification..In the aspect of user matching,a random forest verification algori thm base on stable marriage is proposed to improve the accuracy of user matching and achieve many-to-many user matching problems.T his article is inspired by the stable marriage matching algorithm and achieve many-to-many user matching.In order to ensure the a ccuracy of the user's matching,the random matching of the candidate matching pair is trained for a second time and the final matching pair is generated.The experimental results show that the proposed algorithm is improved compared with the existing relat ed algorithms in terms of accuracy,accuracy,and AUC.Finally,this paper builds a cross-platform social network user identity recognition system using Spark distributed computing framework,Sping,Hive and other engineering technologies.