Research On Opinion Features Regulated Network Embedding

With the rapid development of Internet technology,online social network platforms,such as Facebook,Sina Weibo,and Tencent WeChat,become more and more popular.Online social network platforms provide users with a more convenient,faster,and more private space.These characteristics make online social network platforms gradually become the most extensive information diffusion platform.Different from the traditional information diffusion media,the data flow in the online social network platform is more abundant,and the propagation speed is faster,the propagation range is more abundant,thus forming a lot of Internet Public Affairs.All of these require a healthy social network platform.It can effectively predict and prevent the Internet Public Affairs,and avoid the social turbulence caused by the wanton spread of some bad Internet Public Affairs,thus causing some adverse events endangering social security.Network embedding is a technology that maps the nodes in the network to the low dimensional vector space.It plays a crucial role in many machine learning applications,such as node classification,community discovery,and link prediction.Nowadays,most of the network embedding methods map nodes to low-dimensional vector space by calculating the topological structure similarity between nodes.In contrast,the topological structure of nodes in a social network mainly describes the social relationship of nodes.Therefore,there are two problems when applying the network embedding method to social network research.First,the social network based on an Internet Public Affair has a very sparse social relationships.Correspondingly,the quality of network embedding is affected.Second,there are only social connections between users in such a social network.The opinion information we have will be ignored,which will make the network embedding result inaccurate.In order to solve the above problems,this paper redesigns an opinion network embedding model--OFNE model.The main research methods of the OFNE model are as follows.(1)Establish an opinion feature network: for an Internet Public Affair in the social network,first establish a social network.Nodes in the social network represent users,while edges represent social relationships between users.The OFNE model analyzes the text information of nodes in the network,obtains the opinion feature vectors of nodes through the vectorization technology,and then compares the opinion feature vectors between nodes to find the node pairs with high similarity,and connects them with an opinion feature edge.In this way,we can get a new triple network.In this paper,the triple network is defined as the opinion feature network.Because of the opinion feature edge,the opinion feature network is denser than the ordinary social network,and the embedding results of such networks will be better.(2)Opinion feature Network Embedding: after we established the opinion feature network,the network is embedded based on the connection relationship of nodes in the network.Since there are two types of edges in the opinion feature network,the OFNE model designs a weighted objective function to control the contribution proportion of social relations and opinion features.In different types of networks,this proportion can be flexibly deployed to achieve a better embedding effect.The OFNE model introduces opinion features into social networks,which changes the sparse social network into a denser opinion feature network.In the process of network embedding,social relationship and opinion features are considered at the same time to achieve a better embedding effect.We have done many experiments on real network datasets,and the experimental results show that the OFNE model has a good improvement in performance and efficiency.