Analysis Of Oil Prices' Interaction In Different Regions Of Us Based On Complex Networks

Crude oil is an important energy basis for the development and production of the whole society.and has important economic value for its market research.In the existing academic achievements,the research methods based on time series are often used to predict trend of the market economy in the future.However,these methods can not explore the interaction between different objects.In this paper,the sliding window algorithm is introduced,which can transform the high dimensional time series into a complex network with dynamic evolution of the topology,and the network nodes represent the internal variables of the system.By putting multiple variables in the same system,the sliding window algorithm can study the interaction between variables from the macro level.This paper first combs the existing knowledge of the complex network theory and introduces the characteristic parameters of the network topology,including the node degree,the aggregation coefficient and the betweenness.The node degree and the number of betweenness examine the importance of the nodes in the whole network diagram from different aspects.Secondly,the principle of sliding window algorithm is introduced in detail.When the high dimensional time series is transformed into a complex network by sliding window algorithm,the time series is first divided into several data windows,and the relation coefficient matrix of each data window is calculated.The matrix represents the correlation between the variables in the system.Then,the adjacency matrix is established by selecting suitable threshold,which is the final basis for building complex network.In addition,the analysis results show that window parameter setting has an important impact on system analysis.In order to obtain as many network graphs as possible,so as to observe the structural changes of the system more carefully,it is generally necessary to reduce the length of the window,but when the window length is too long,the amount of data contained in a single window will be reduced,and the risk of the isolated sample sampling point exists in the observation system.Meanwhile,the selection of window length also directlyaffects the computation cost of the algorithm.Therefore,this paper gives a comprehensive scheme to determine window setting parameters.Finally,based on the monthly data of crude oil prices in 23 regions of the United States,the paper builds a complex network map of the crude oil price market based on the sliding window algorithm,and analyzes the status of the individual region in the energy market at the same time,as well as the synchronization of the whole energy market.It also discusses the similarities and differences between individuals and the whole world.The results show that the evolution trend of average degree and average clustering coefficient is greatly influenced by the historical average value in the top five regions.In addition,the area with small node degree may play an important role in the whole network,for example,the historical average node degree of the north slop area does not enter the first five.However,the average number of historical betweenness is ranked first,indicating that the area is the middle connection point of different node groups in the network graph.When the sample data is collected,transferred and copied,the original data are often changed,and the results of this paper also show that the network topology does not change with the fluctuation data when using the sliding window algorithm to study the system features,and it still does not affect the dynamic observation of the system from the macro level.The robustness of the method is studied.