Event Space Model Based On Virtual And Real Society And Its Application In Stock Field

The social media, such as Twitter, Facebook, Sina Microblog has been widely popularized. The real-time interaction between online and offline has great influence on the virtual cyberspace and the actual space. The virtual cyberspace with hundreds of millions of users has become a world that almost parallel with the real world, and these two worlds are interactive and interdependent. The data of virtual cyberspace has the characteristics of expression freedom, large scale and fast increment, and its influence and impact on the real world is more and more. However, the real data with multi-source and heterogeneity feature can lead to a large number of virtual network space data loss and noise when mapped to virtual network space. The incompleteness and magnanimity of cyberspace data inhibit the further development of event value mining, such as event influence time definition, event influence space definition, event multiple influence calculation.In this paper, we combine Virtual Opinion Data Space(OS) and Actual Behavior Data Space(BS) to describe and measure an event from various angles. The Event Space Model(ESM) is constructed with event trajectories mapping in the OS and BS two spaces. Based on this model we make the further research of event. The general research content and innovation spot of this paper are listed as follows:(1) This paper effectively finds the starting point of the events by multiple members and multiple attributes clustering fusion method, that is a more comprehensive and effective method from the OS and BS spaces to describe and measure. In addition, due to the lack of research on the nonlinear linkage of multi-dimensional time series, this paper uses random matrix theory and sliding window technology to calculate the similarity of multidimensional time series, and multidimensional time series data of nonlinear correlation model is established to find relationship between members of the stock event.(2) This paper calculates the event affect scope through the method of clustering and multidimensional attribute that is a blend of the event influence in the distribution of the OS and BS spaces overlap degree. Based on event affect scope calculation in the life cycle of event, we find that the affect scope gradually converge to a certain interval. According to the event affect scope gradual convergence to trace the source of the event in the time series, which is verified with the news at the same time.(3) Based on the memory and forgetting curve of cognitive psychology, the memory and forgetting features of fusion events are used to calculate the multiple influence of events from multiple angles. In addition, the observable influence and potential influence of stock events show different trajectories in the actual situation. We found that observable influence and potential influence in different spatial stages were different in different stages of the event, and the OS space after the fusion influence and BS spatial correlation.The experiments show that this paper focuses on the establishment of the ESM, further development and perfection in theory the event starting point detection, event effect scope detection, event influence calculation based on the memory and forgetting curve. In addition, from the aspects of practicality and feasibility the research results can be applied in the event correlation discovery, source of the event and the relationship mining between observable influence and latent influence and so on.