Research On Cross-social Network User Identification Based On User Nickname And Friend Relationship

In recent years,social network applications are becoming more and more extensive,Internet users will generally register multiple social network accounts to meet different social needs,if the same user of different social network accounts can be identified and information integrated,can get more user information and then the user for a more comprehensive portrait,used in intelligent recommendation,network security and other fields.Therefore,cross-social network user identification has become a research hots-pot in academia in recent years.Aiming at the problems of high difficulty in obtaining data and unsatisfactory recognition accuracy in the field of cross-network user identification,this paper proposes an identification method based on user nickname and friend relationship information.This method uses the user’s nickname and friend relationship to determine whether two accounts belong to the same user.On the one hand,nicknames and friend relationships are publicly available basic data on English social networking sites,which can overcome the problem of difficult data acquisition caused by the website’s focus on privacy protection.On the other hand,the comprehensive use of user nicknames and friend relationship information will produce more information redundancy and higher recognition than the use of one information alone.This paper takes Foursquare,Facebook and Twitter websites as research objects,uses the function of linking users’ Facebook and Twitter homepages provided by Foursquare website,designs a crawler to obtain the nickname and friend list of the same user in Facebook and Twitter on Foursquare as a positive data set,and obtains a negative example datasets through the random combination of positive case data.Data analysis is carried out on the nickname and friend relationship datasets,the characteristics and rules of the data are obtained,and the feature processing program based on Python is designed to extract the characteristics of user nickname and friend relationship in batches,and find the machine learning model Light GBM with the best classification effect for feature training and analysis results.Finally,the characteristics of similar users in the friend relationship are further extracted through analysis,the model training and result analysis are completed,and the recognition results of other scholars are compared.Through the experimental results,it can be concluded that the recognition effect of comprehensive user nickname and friend relationship is greatly improved compared with the use of single information,especially after adding similar user pair characteristics in friend relationship,the recognition accuracy is further improved.The experimental results show that the accuracy,recall rate,F1 value and accuracy rate of user identification when combining user nickname and friend relationship reach94.20%,88.30%,91.15% and 88.67%,respectively,which are increased by 6.13%,7.31%,6.77% and 8.56% compared with the indicators of user nickname recognition only,and11.28%,7.44% and 9.28% respectively compared with the indicators of using only friend relationship to identify users and 12.85%,with a significant improvement.After adding the characteristics of similar users in the friend relationship,the indicators of the recognition results reached 95.95%,91.31%,93.57% and 91.68%,respectively,which further improved the recognition accuracy to a certain extent.