Research On The Construction And Application Of Financial Event Correlation Network

Things in today’s world are growing super connected and complex.After exploration,people gradually realize that complex network is a powerful tool for insight into the world: It can help people understand the natural or social problems behind a particular system by abstracting events into network nodes,mapping the correlation between things into network edges,and then generating a network that describes a particular system.Take the financial field as an example: In the past decades,many scholars have tried to introduce the complex network into the research of financial problems,to build the financial network and carry out the analysis of the network,to understand the existing phenomenon in the financial system more comprehensively,and to discuss the future evolution trend of individuals and even the whole system.At present,it has been widely practiced to construct financial network by using the original data of financial market,but such methods have certain limitations in terms of data collection and network analysis.At the same time,the use of alternative data,represented by news text,not only give more resolutions for financial network building and analysis,but also attract more researchers’ interests who’s originally majoring in other fields that give more eyes on financial networks researches.But this kind of network construction method can often be stumbled by data accessibility and environmental conditions,that makes the construction process less transplantable;For the validity of this kind of network construction method,whether in qualitative or quantitative perspective,the way for carrying out a discussion needs to be further discussed.In view of the above problems,this paper focuses the following three aspects works:(1)By referring to the construction idea of the two-layer event correlation network,with the "Co-occurrence Analysis" idea which originates in text mining is taken as the cut-in point of network construction,while taking the Chinese context as an example,this paper gives a framework for constructing financial network based on event correlation mining.The biggest difference from existing related methods is that when the financial network is constructed with alternative data represented by news text,the dependence on semantic knowledge base is avoided in the process of obtaining event correlation relations.(2)Based on the collected financial news data,this paper uses the above mentioned network construction method to build an actual financial network,and analyzes the variable factors that affect the network topology characteristics and then make interpretations--By adjusting the similarity threshold parameters that change the shape of the generated network,based on several specific topological indicators that analyze the network topology shift with logics behind them;At the same time,select some representative paths in the network,combined with the basic knowledge of finance,to elaborate the effectiveness of the constructed financial network in an intuitive manner.(3)Referring to the existing question of "Predicting the trend of a given stock’s price according to relationships among stocks",a quantitative analysis method is set to evaluate the effectiveness of event correlation network construction.Based on the constructed financial network,the associated network of financial entities is mapped in the first place then predict the trend results of the stock market according to the associated stock market of a certain stock that begins with the stock price prediction,and the results obtained by the existing method are compared with the results obtained by this method with the reference of the existing similar research.The experiment result shows that compared with the existing methods in obtaining the financial entity correlation,the financial entity correlation based on event correlation can better detect the trend of stock price changes under the premise of more emphasis on the reliability of entity correlation,with which quantitatively proves the validity of the constructed financial event correlation network.This research enriches the method of constructing financial network based on alternative data,shows a new analytical perspective for financial network research,and provides new ideas for the construction of decision support system in financial field.