Theoretical Models And Empirical Research On Information Dissemination And Sentiment Diffusion In Investor Network

Investors in stock market are not in isolation, while they self-organize into the investor network. They obtain stock market information, exchange ideas of stock market, and vent their emotions through the investor network. Investors in the network have the highly dynamic, self-organization, and heterogeneity characteristics that make their activities become more and more complicated. So, the traditional models and research methods of behavioral finance are difficult to accurately describe not only the microscopic interaction behavior between users but also the macroscopic phenomenon of information dissemination and sentiment diffusion. In nowdays, the investor network often works through the Social Network Service (SNS). In view of this, we use the interdisciplinary ideas and methods to study information dissemination and sentiment diffusion models in SNS to describe how the information dissemination and sentiment diffusion in the investor network affect the stock price.In the paper, we focus on the information dissemination and sentiment diffusion in investor network. The research on information dissemination in investor network includes two aspects:Firstly, we think about the theory of information dissertation in investor network. We propose a new model for information dissemination in social networks. The dynamic equation of information dissemination is modified, in which the infectious probability is defined as a function of the homogeneity and heterogeneity between nodes. Moreover, we investigate numerically the behavior of the model on a real scale-free social site. We find that initial spreaders with big out-degree of can accelerate information dissemination. Weak ties between nodes play an important role in information dissemination process. Specially, selecting weak ties preferentially can make information spread faster and wider, and the efficiency of diffusion will be greatly affected after removing them. Then, we think about the competition of rumor and counter-rumor from two ways. We construct the competition model of rumor and counter-rumor based on SIR. The quantitative analysis with the support of the dynamic mathematics model is proposed. Using the theory of differential dynamical system, the status and stability of the rumor spreading are analyzed. Then, we construct the competition model of rumor and counter-rumor based on complex networks, in which more status of nodes are considered. Moreover, we investigate numerically the behavior of the model on a real scale-free social site. We find that the lag between rumor and counter-rumor play an important role in information competition process. Initial spreaders of counter-rumor with big out-degree of can not only accelerate counter-rumor dissemination, but also enlarge the range of counter-rumor.Secondly, three hypotheses are proposed based on the theory of information dissertation in investor network. Then we empirical analyses the size and the influential factors of stock price shocking caused by rumor and counter-rumor.The research on sentiment diffusion in investor network includes two aspects:As the carrier of the sentiment diffusion in investor network is information, we construct the model of SNS information flow (MSIF). For analysing the change characteristics of MSIF caused by the internet information that investors are interested in, a wavelet method is proposed for detecting and estimating jumps and cusps of a semiparametric regression model with heteroscedastic variance. We construct the test statistics which can be used to detect change-points in the model, and the asymptotic distributions of the test statistics are established. We also utilize the test statistics to construct the estimators for the numbers, locations and jump sizes of the change-points. The asymptotic properties of these estimators are derived. Existing studies are mostly concerned with the estimation of change-points of nonparametric regression model, this paper is concerned with the estimation of change-points of semi-parametric regression model which could expand the range of the applications. Some simulation studies are conducted to assess the finite sample performance of the proposed test statistics and estimators. The results from these simulation studies show that the new test has excellent power of the test statistics and the very accurate estimators. We empirical analyses the SNS information flow by a semiparametric regression model, and detect the jumps and cusps of this model by wavelet methods. The relationship between internet public opinion and the jumps and cusps of the model of SNS information flow.Then we construct the investors’sentiment model in investor network based on the MSIF, while we proposed a general method of extracting investors’sentiment from investor network. We extract the orient of investors’sentiments on SNS from the information flow based on the How Net framework, and construct the index of investors’sentiments on SNS. The index is proved to be raional after correlation analysis.So we use it as a proxy variable to analyze the relationship between investors’sentiment and stock price through granger causality test and impulse response functions of random shocks.This research includes the following contributions and innovations:(1) A new model for information dissemination in social networks that exist directed edges is proposed, in which the infectious probability is defined as a function of the homogeneity and heterogeneity between nodes. By numerically experiment, we find that weak ties between nodes play an important role in information dissemination process. Specially, selecting weak ties preferentially can make information spread faster and wider, and the efficiency of diffusion will be greatly affected after removing them.(2) The competition model of rumor and counter-rumor is constructed based on both SIR and complex networks, which makes the research in the competition model of rumor and counter-rumor more comprehensive.(3) We empirical analyses the size and the influential factors of stock price shocking caused by rumor and counter-rumor under the framework of information dissemination in social networks, which enrich the research method of the area of information asymmetry.(4) We develop wavelet methods for detecting and estimating jumps and cusps of a semiparametric regression model with heteroscedastic variance. We construct the test statistics which can be used to detect change-points in the model, and the asymptotic distributions of the test statistics are established. We also utilize the test statistics to construct the estimators for the numbers, locations and jump sizes of the change-points. The asymptotic properties of these estimators are derived. Existing studies are mostly concerned with the estimation of change-points of nonparametric regression model, this paper is concerned with the estimation of change-points of semi-parametric regression model which could expand the range of the applications.(5) We construct the model of investor’s sentiment in investor network. Based on the model we extract investor’s sentiment index from the SNS. Then, we use the index as a proxy variable to analyze the relationship between investors’sentiment and stock price through Granger causality test and impulse response functions of random shocks.