| The securities market is an important part of the financial market,and the explanation of securities market price volatility has always been a key concern in financial research.There are many factors affecting the price volatility of the securities market,and information serves as an important vehicle for eliminating stochastic uncertainty.Financial studies have verified the economic principle of the impact of information on securities market prices and explored the mechanism of the role of information on the securities market,starting from the fact that information in the securities market has an explanatory role in the price volatility of the securities market.However,with the rapid development of contemporary information technology,information in the securities market presents new characteristics.The information structure is characterized by multiple sources and heterogeneity,the speed of dissemination is more rapid,and there is a large amount of redundant information in the huge amount of information.Based on the new characteristics of information,the flow and diffusion of multi-source heterogeneous information have more significant impact on the pricing of assets in the securities market,and the interaction of different types of information forms a comprehensive impact on the price of securities market.The fusion,processing,mining and analysis techniques of multi-source heterogeneous information are still in the exploratory stage,which limits the explanation of multi-source heterogeneous information on the price fluctuation of the securities market.The latest economics research results show that the securities market as a complex dynamic system is influenced by both rational and irrational factors,and information,as an important content carrier of the influencing factors,becomes the main research object,which has an explanatory role for the securities market price fluctuations.However,the research on securities market price volatility based on market efficiency hypothesis and behavioral finance mostly starts from single information elements such as securities market news and pricing factors,and studies the impact on securities market brought by the flow or diffusion of single information,trying to explain securities market price volatility from single information.As the securities market is a complex dynamic system,this type of research has gradually shown limitations.The effect of complex data information on the securities market on the price fluctuations of the securities market and the explanation of the role of the securities market has become the latest research hotspot.Although researchers have explored complex data fusion from the perspective of the correlation network of the securities market and conducted a large number of fruitful studies.However,in view of the multi-source and heterogeneous characteristics of complex data information,how to construct complex networks in the securities market,explore the correlation relationships in complex data,and data fusion based on the correlation perspective still need to be further explored.For complex data in securities market,traditional financial measurement methods are limited by dimensionality and linear analysis characteristics,and it is difficult to effectively present the interactive superposition impact of multi-source heterogeneous data information on asset price fluctuations in securities market,and cannot effectively interpret the flow or diffusion process of complex information data,thus the traditional financial measurement research framework cannot effectively verify the explanatory role of fused multi-source heterogeneous information data on market price fluctuations and In order to overcome the limitations of the traditional financial econometric research frameworks,we have been able to verify the explanatory role and dynamic impact of the fused heterogeneous information data on market price fluctuations.To overcome the limitations of the traditional research framework,this paper adopts a modern financial research approach to fuse multi-source heterogeneous data infographics based on a correlation perspective and constructs a deep learning model to explain the impact of the fused multi-source heterogeneous information on asset price volatility in the Chinese securities market.Based on the market efficiency hypothesis and behavioral finance theory,this paper is based on the premise that securities market is a dynamic and complex system affected by multiple factors.First,based on previous studies,we explore the correlations in single information elements and construct sub-graphs of different types of information to verify the effectiveness of the graph embedding method and the rationality of research based on the correlation perspective.Secondly,on the basis of the theoretical research on the correlation network and information spillover effect in the securities market,this paper innovatively proposes the construction method of the heterogeneous graph of the securities market based on the correlation perspective to realize the fusion of subgraphs of single information elements.Finally,in order to verify the impact of multi-source heterogeneous information flow and diffusion on the asset prices in the securities market,this paper designs multiple attention network algorithms to process and analyze the fused heterogeneous graphs,in which the nodal attention mechanism is used to focus on the comprehensive impact of multi-source heterogeneous information on the price fluctuations in the securities market,and the semantic attention mechanism interprets the correlation relationship of multi-source heterogeneous information.Based on the correlation perspective,this paper innovatively proposes a graph embedding method to construct the heterogeneous graph data of Chinese securities market by subgraph fusion.In view of the characteristics of the constructed heterogeneous graph data,multiple attention graph neural network is used to focus on capturing the information spillover of multisource heterogeneous data,aiming to accurately interpret the comprehensive impact of multi-source heterogeneous information spillover on the asset prices of securities market.This paper uses deep learning techniques as the implementation basis to explore the impact of multi-source heterogeneous information flow and spillover on the price volatility of the securities market,and the main contributions and innovations of this paper’s research include.First,this paper extends the theoretical study of information spillover from multiple sources of heterogeneous data in the securities market and argues for the combined effect and explanatory role of multiple sources of heterogeneous information flow and diffusion on asset prices in the securities market in China.The theoretical extension of the spillover effect in the securities market is extended from the impact of a single information element on asset pricing across market quotations to the dynamic impact and integrated effect of multi-source heterogeneous data information spillover on the price volatility of the securities market.The correlation perspective is used to interpret the information spillover from multiple sources of heterogeneous data in the securities market,and the correlation relationship between multiple sources of heterogeneous information is explored to verify the explanatory role of information spillover from multiple sources of heterogeneous data on the price volatility of the securities market.Second,based on the perspective of correlation of multi-source heterogeneous data information in the securities market,we propose the method of embedding and analyzing the multi-source heterogeneous information graph in the securities market.Combining the characteristics of different types of information in the securities market,the graph embedding method is introduced to construct a single information element subgraph,which effectively solves the problems of multi-scale and multitext information mining of time series and unstructured index feature extraction.We break through the tensor and vector heterogeneous data fusion methods,introduce the relationship dimension to represent the interrelationship of the influencing factors of the studied securities market,and creatively adopt the graph fusion method to construct the multi-source heterogeneous graph data of the securities market,which can more fully represent the propagation and interaction of multisource heterogeneous information in the securities market.Third,on the basis of multi-source heterogeneous information fusion,the proposed attention graph neural network model achieves heterogeneous graph analysis and algorithmic innovation by combining the small-world characteristics,law degree distribution characteristics and community structure characteristics of the constructed multi-source heterogeneous graph data.It provides new research perspectives and ideas for studying classical propositions in finance.The multiple attention mechanism graph neural network consisting of nodes and semantics is introduced,focusing on interpreting the key nodes and association relationships of multi-source heterogeneous information,and mining the implied semantics of multisource heterogeneous graphs,which helps to explore the dynamic influence and explanatory role of interrelated formation of multi-source heterogeneous information on securities market prices. |
PDF Full Text Request