Research On Analysis Method Of Social Network Accounts Based On Knowledge Graph

In recent years,the rapid development of social media such as Twitter,Google,and Facebook has provided people with convenient access to information such as hot events and real-time news,allowing them to participate more freely in online social activities.Through social network accounts,users can easily access content of interest,exchange and share ideas,and express personal opinions.These accounts contain a wealth of valuable information.Therefore,studying the basic materials,behavioral characteristics,and correlation between these accounts can better understand users’ personal preferences and behavior patterns,which is of great significance for users and social platforms.By delving deeper into the key information in social network accounts,resources on social media can be better utilized.At present,most mainstream social network analysis methods extract account characteristics through statistical analysis,embedded representation and other technologies,and use supervised learning and regression analysis to classify accounts.However,these methods have low accuracy in the account classification for two main reasons.Firstly,they lack sufficient consideration of the various key information involved in the account and their correlation,and do not make good use of resources in social networks.Secondly,traditional classification methods are not suitable for social network account classification problems,which require a large number of training samples and the cost of model training.It is not able to accurately capture the deep features of accounts in the social network structure,and some key information is lost during the feature process.Therefore,this article introduces knowledge graph analysis of social network accounts and proposes some innovative analysis methods from two aspects: account representation and classification:(1)A knowledge graph based representation method for the key information of social network accounts was proposed.This method first extracts knowledge from the key information of account data through node extraction and relationship extraction,and then performs knowledge filtering and node merging to remove noise information and duplicate parts in account data information.Finally,the account knowledge graph representation is obtained.The experiment shows that the social network account knowledge graph representation method proposed in this article has achieved good results in key information extraction,key information representation,and other aspects.(2)A social network account classification method based on knowledge graph meta paths was proposed.This method is based on the knowledge graph structure of meta path analysis,designing different meta paths to calculate the correlation between accounts,and learning the weights of different meta paths.The correlation matrix is weighted and combined to transform the account knowledge graph into a weighted account node network.Finally,a transductive classification model is constructed to classify social network accounts.Through actual data testing,this method can fully utilize the key information represented by the account knowledge graph and outperform other common traditional machine learning classification methods in account classification problems.