Quantitative Analysis Of Stock Market Based On Multi-dimensional Market Information From The Perspective Of Big Data And Artificial Intelligence

As China's market economy has developed continuously,China's stock market has become an essential part of the economic system.Healthy of the China stock market prosperity is fundamental to the Chinese economy's stability,a critical point that affects the nation's economy.As of February 2020,more than 3,800 listed companies in Shanghai and Shenzhen stock markets with a total market value of61.44 trillion Yuan,an approximate of 62% of 2019 annual GDP.The market has achieved its capital accumulation and circulation,provides a channel for the jointstock company to raise funds,and promotes community resources optimizat ion.Significantly,the market has expanded capital investment methods for investors where the health and stability of the market environment are beneficial for protecting investors' fundamental interest;maintain the safety of social order and prosperity has provided a reliable guarantee.Therefore,stock market researches not only related to the future development of China's economy.It can optimize the rules and regulation of the capital market by exploring the fluctuation pattern of the stock price,protecting the market participant's interest,and ensuring social harmony and long-term stability of China.In the stock market research field,the stock market impact analysis from market information-driven is always mainstream in academic research.From a theoretical and empirical perspective,researchers have fully proved the impact of market information on the stock market and how this key factor,market information,captures stock price fluctuation and measure market future operation trends.In recent years,globalization and the country's opening-up have become the central theme of Chinese stock market development;the reform's continuous improvement has made China stock market no longer a "lonely capital island." The opening of capital channels such as the Shanghai-Hong Kong channel,Shenzhen-Hong Kong channel,London-Hong Kong channel has made the China stock market and the international stock market to become more closely connected and strengthen the relationship.Simultaneously,with the rapidly changing internet technology,channels for disseminating market information and dissemination speed have been greatly enhanced,enabling investors to promptly receive and search for the latest market information.Scholars also began to analyze from different market information perspectives to unfold the stock market's correlation with international market information,macro market information,and micro market information.However,there are serious flaws in current quantitative research of the stock market based on market information.Since fluctuations of stock market prices are caused by multi-dimensional market information,the correlation analysis method of market information of a single dimension cannot effectively and accurately quantify the future trend of the stock market,but no scholar has studied the comprehensive effect of market information of multi-dimensional on the stock market,therefore research on how market information of multi-dimensional can affect the stock market is still undiscovered.At the same time,there are many challenges in the research on the effect of market information of multi-dimension.Firstly,from data acquisition,with rapid development and popularity of internet technology,the research on the effect of market information on multi-dimension on the stock market is facing mass data collection.Secondly,during data processing,market information of multi-dimension data has contained different data structure(numerical and textual),a method to achieve precise measurement for textual market information is a key issue that needs to be addressed.At the same time,market information of multi-dimension needs to be processed for data fusion before correlation analysis,the method to consider and preserve the interrelation of market information of different multi-dimension during the process of data fusion is an important difficulty in the research process.Finally,during data decision making,time-series of market information of multi-dimension(continuous and discontinuous)can distort feature space and reduce the accuracy of the correlation analysis model.Nevertheless,the release of stock market information is a dynamic process and real-time trading decisions need to be sensitively and timely,therefore,how to use market information of multi-dimension to improve efficiency and accuracy of decision making is also one of the important challenges of the research.For that reason,as we facing a research gap in market information of multidimension,this research shall propose a concept and idea for quantitative analysis of the stock market based on market information of multi-dimension and applies artificial intelligence technology to solve difficulties and challenges in this research.It also provides guidance and suggestions to participants of various stock markets,present a theoretical basis for the formulation of market policies,and protect the health and stability of the stock market.Specifically,this research mainly includes four aspects of research contents and contributions.(1)Based on the quantitative method of sentiment factors in deep neural networks for attention mechanisms.This research is aimed at quantifying the unstructured text information in market information of multi-dimension,which proposed a novel attention mechanism(Attention)for the emotion quantification method in a convolutional neural network(CNN).A deep neural network model can abstract and analyze low hierarchical concepts,enhance machine learning methods for feature extraction,and complete sentiment features classification with high accuracy.Specifically,this research uses a multi-levels attention mechanism network to extract the hierarchical features in the text content of market information of different dimensions.To a certain extent,it can alleviate the heterogeneity of the quality of text content and obtain a high-level semantic analysis of the quality of text in different dimensions.Secondly,the multi-levels attention mechanism can unify and integrate market information of different dimensions;it can effectively extract the joint sentiment between market information of different dimensions.Thereby,its conduct can enhance the downstream association analysis model to capture the synergistic effect of market information of multi-dimension on the stock market.(2)Based on the modeling of feature space for market information of multidimension with the tensor algorithm.The stock price is a comprehensive expression from the interaction between heterogeneous market information from multi-source.When analyzing stock price fluctuation,it is necessary to comprehensively consider heterogeneous market information from multi-source to achieve the purpose of capturing market risk accurately.With traditional research,when handling the combination of information of multi-dimension,information from different sources is often spliced into a super vector.This combination has ignored the interrelationship between market information of multi-dimension and caused some loss in the internal value of interactive information.In this study,we use a tensor model to model feature space of market information of different dimensions and use tensor decomposition and reconstruction to capture interaction,extract interactive effect and common effect,and effectively achieve the purpose of preventing the loss of valuable information.(3)Based on the time-series neural network model driven by market information of multi-dimension.In the stock market,the space in market information of multi-dimension not only contains interaction between the information of various dimensions,but it also has time-series characteristics.For the media information dimension of the stock market,it is a discrete-time series characteristic of changeable length.For the market transaction data dimension,it has a continuous timing characteristic of equal distance.For the fundamental data dimension of listed companies,most of them have discrete time-series characteristics of fixed length.Moreover,in the real stock market,the effect of market information on the different dimensions of the stock price is a dynamic process.It exists not only a mutual relationship between each other,but also eliminate each other.Therefore,the adjustment of key influencing factors is an important step in optimizing and guiding the updating of the model parameter for correlation analysis.Based on the above two points,this study proposes a time series neural network model based on market information of multi-dimension to eliminate the reducing effect of dense continuous time-series data on the characteristics of sparse discrete time series data.Based on the key influencing factor of the dynamic change of the trading time window,the intelligent analysis of market information of multi-dimension has laid a solid foundation.(4)Based on the trend of stock market price research from the three-way decision framework.In the actual stock market trading process,the investors' decisions will be based on comparison of market information being obtained and personal knowledge,where three-way decisions will be made: buy,sell,and hold.Among this,for investors,holding is a way to avoid market risk.When market information is insufficient or lacking analytical capability,the fuzzy analysis area has been set up to trade risk caused by direct buying or selling.This study combines the actual situation of the market transaction and is the first to introduce “multi-level”and ‘multi-perspective” for granular computing ideology in the field of stock price trend research by building a three-way decision framework and use granular computing to deal with the uncertainty of complex problem.Specifically,this research is novel in realizing a deep integration of the granular computing three-way decision research framework and time-series neural network model.We set the granular analysis rules in time and space level and apply fusion algorithm to the stock market price research,by which improve efficiency and ability of the correlation analysis of three-branches computing framework to dynamically capture the future stock prices and achieve the purpose of resolving fuzzy space analysis,reducing the decision cost and learning cost.