Research On Internet Person Relationship Analysis Method Based On Person Similarity

As the world's largest professional social networking platform,LinkedIn plays an important role in people's career and has become one of the important ways for users to communicate.In LinkedIn,users interact with each other by perfecting materials,sharing experiences,and expanding contacts,thus enabling the LinkedIn social platform to contain a large amount of real user information.Using this information to analyze the relationship between LinkedIn users and mining the information behind user data will help to grasp the distribution of talents in various fields of society and achieve targeted information on talent demand.Based on the user dataset of LinkedIn social network platform,this thesis proposes a user relationship analysis method based on character similarity,which is applied to the topic of character relationship analysis.The main work and contributions of this thesis are as follows:(1)A network of similarity relationships of users is constructed.The LinkedIn user has various attributes.For each different attribute,the similarity between the users is calculated.Then,the user is used as the node,and the similarity between the attributes of the users is used as the edge to construct the user similarity relationship network,which transforms the problem of LinkedIn user relationship analysis into complex network analysis problems,laying the foundation for subsequent research.(2)In order to divide user groups more effectively and improve the classification effect of users,this thesis proposes a user classification method based on graph embedding.This method selects the user's word attribute and text attribute as the attribute feature.On this basis,the graph embedding algorithm is introduced to mine users' topology attributes in users' similarity relation network,which is used as users' topology features and combine it with attribute features to jointly represent users.Experiments were conducted by using the user data of LinkedIn.The five-fold cross-validation method was used to train the classifier and the corresponding experimental results were obtained,which proved the effectiveness of the method.(3)In order to make full use of the similarity between users on different attributes,this thesis proposes a user community partitioning method based on multi-layer network aggregation,which applies multi-layer network aggregation to user group mining.This thesis defines a multi-layer network system,in which each layer network represents the similarity relationship of users in one attribute;multi-layer network aggregation methods based on topology,based on analytic hierarchy process and based on inter-layer similarity is proposed,and the aggregated network is used in the user's group mining.The aggregation method based on topology considers the topological characteristics of each layer of network,and the aggregation method based on the analytic hierarchy process measures the importance degree between each layer of networks.The aggregation method based on the interlayer similarity considers the interlayer similarity factors,and considers that if one layer network has a high degree of similarity with the rest of the network,the more it reflects the user relationship in the entire multi-layer network.Finally,it is proved by experiments that the multi-network aggregation method proposed in this thesis perform more effectively in user community division with LinkedIn user dataset than the traditional method.