Research And Analysis On The Fusion Degree Of Virtual Space And Corresponding Transaction Space Based On Link Similarity

The group's activities in the virtual space are the more stable norms and spirits formed by members of the group in their direct and recurrent joint activities such as group interests,group needs,group opinions,group goals,community norms,etc.Group is the mediator of individual participation in social activities,group has a direct impact on individual behavior.With the rapid development of the Internet and social media,the Internet provides a convenient basis for people to express their opinions.The trend of the network public opinion information and the actual transaction behavior is becoming more and more highly integrated.The Internet has evolved from a simple information release technology platform to The main carrier of social media has become an interactive social network of information dissemination,sharing,exchange and collaboration,which has greatly changed the way people look both socially and economically.Group activities occur in a certain public space.Without the existence of public space,there will be no group activities.As the direct foundation of group activities,with the continuous improvement of the group participation,people no longer accept the knowledge passively,but actively publish various opinions and comments.These ideas and comments can not only express people's true thoughts in real time,but also change the real-world activities by influencing the psychology of the audience.The social activities of virtual network are mainly manifested in the aspects of group cognition,group psychology,group interaction and group attitude.At present,researchers mainly make use of the degree of attention index in network public opinion information as a reflection of group activities in the virtual space,while the content hidden in network public opinion information and transaction behavior is not taken into account enough.In addition,the fusion degree of network public opinion information in the virtual space and the transaction behavior in the corresponding transaction space is repeatedly mentioned,there is seldom empirical research.In order to solve the above problems,this paper selects the actual stock trading behavior data and the network public opinion information associated with the stock in the virtual space as the research object on the background of the financial market.In order to study the fusion degree of the virtual space and the corresponding transaction space in different time periods,based on the group attention of virtual space we build network public opinion space and real transactional behavior data sematic space based on the data sematic,respectively.On the one hand,we extract the network public opinion information and the group attention behavior published in the virtual space,construct the network public opinion space,analyze the sentiment tendency of the network public opinion information,extract the group's attention activity in the network public opinion space,vectorize the stock in network public opinion space according to the group's attention,give the representation method of the stock in network public opinion space and construct the public opinion space complex network.Based on the characteristics of information and groups' activities in the network public opinion space,we construct the group's sentiment tendencies sequence according to a certain time granularity.On the other hand,we extract data semantic of real-world transactional data.Based on the semantic space of the real-world transactional behavior data,a complex network of transactional behaviors is constructed to analyze the fusion degree with the network public opinion space complex network in different time periods,in order to confirm the consistency of the group's activities in the network public opinion space and the transactional behaviors in the real world.The main contribution and innovation of the thesis are as follows:(1)For the sentiment tendencies of the information contained in the network public opinion space,we propose a character-based reccurrent convolution neural network.The method uses convolutional network and recurrent network to determine the relationship between characters without the need to consider the syntax and semantic features of the information itself,and improve the efficiency of analyzing the information sentiment tendencies in the network public opinion space.(2)Based on the group's attention activity in the network public opinion space,we study the representation of the stock,and propose a stock vectorization representation method based on the word vector model of natural language processing.This method uses the group's attention sequence to construct and analyzes the vectorized representation of stock in the network public opinion space based on time correlation and spatial consistency.(3)On the basis of the data space of the real transactional behavior,the group's activities in the real society are further divided into the information space of the real transactional behavior and the data semantic space of the real transactional behavior.This paper proposes a semantic extraction model of real transactional behavior data based on the contextual relationship.The algorithm maps the real transactional data to high dimensional semantic space.The experimental results show that the algorithm effectively guarantees the temporal and spatial relationship of the original data in the mapping process.(4)We systematically confirm the consistency of the virtual space and the corresponding transactions space in China's A-share market.Based on the stock representation in network public opinion space and the semantic space of real transactional behavior data,this paper studies the construction of network public opinion complex network and transaction behavior complex network.On the basis of this,the relationship between the group's transaction behavior in the real society and the activity in the network public opinion space is transformed into the similarity measure between the complex network in the network public opinion space and the complex network in the corresponding transactions space.The experiment confirms the consistency of the network public opinion space and the real transaction behavior in the Chinese A-share stock market.(5)In view of the time-varying and unpredictable of real-world transactions,we propose a group behavior prediction model based on long short term memory and an adaptive segmentation method based on data driven.Experimental results show that the proposed method can effectively predict population behavior,and the segmentation results are stable and can effectively find the mapping relationship between network public opinion information and real transaction behavior.