Social Network Topic Evolution And Prediction Based On Knowledge Graph

The Internet today has become a major channel for people to obtain information,such as current news,hot topics,learnning material and so on.Those information in social networks can be instantly transmitted on the Internet.So,it is necessary to build an efficient and reasonable topic model to manage these topics which involve a series of concepts,elements,and complex connections.In my opinion,effective topic models will enable the modeling,storage and management of topics and support for analysis and early warning applications.Knowledge graph is a knowledge base with semantic processing ability,which has attracted much attention in the fields of knowledge extraction,knowledge fusion and knowledge reasoning.Firstly,Microsoft concept graph is taken as the research object and constructed into a large concept graph network.Then the relevant theories are used for analysising the feature of this concept network.Complex network theory is an abstract description and analysis tool for complex systems.It can explore the inherent characteristics of Microsoft concept graph and deeply understand its essence.Complex network theory focuses on the interaction or relationship of various factors in complex systems.The complex network characteristics such as node degree distribution,network average shortest distance,clustering coefficient and degree correlation of the largest connected subnet of Microsoft concept graph network has been calculated.Then the concept graph was introduced into the short text classification problem.A short text semantic extension representation method based on concept graph is realized.Firstly,the correlation degree between the text feature words and the concepts in the concept graph is calculated.According to the results of correlation degree,the most relevant concept is selected to form a conceptual dictionary of texts.Then,the concept dictionary is used to expend the feature words to generate a semantic extended representation of the short text.The short texts from Twitter were adopted to evaluate the semantic representation.The results show that the conceptual semantic extension can improve the classification effect of short texts.So as to prepare for the subsequent topic evolution relationship analysis.Finally,the topic evolution model is implemented.The model is divided into two parts.The first part is the topic discovery and tracking.The semantic representation method based on concept graph and the topic detection method based on improved incremental clustering algorithm are introduced.The experiment is carried out through Weibo data show that the validity of the method was verified.Following the topic evolution relationship mining,the correlation and evolution relationship between topics are calculated by cosine similarity and KL divergence.Finally,the topic evolution prediction model is designed to predict the topic to a certain extent.