Research On Time Series Data Based On Visibility Graph And Structure Entropy

Time series data research is one of the hot spots of current academic research.Among them,the use of relevant knowledge of complex networks to analyze and research time series data is an important method.There are two key issues in interpreting time series data with the help of complex network theory: one is how to convert time series data into complex networks.The second is to select suitable indicators in the obtained complex network to interpret and analyze the time series data.Mapping a time series into a complex network,in terms of conversion methods,the current viewable method presents intuitive and effective characteristics,but the similarities and differences in the application of several viewable methods to time series remain to be summarized.In the selection of indicators,the entropy of the network has been widely used,but the discrimination of several existing entropy indicators is not obvious in many networks.For this reason,the main research contents of this article are:1.Comparing the two methods of no-traversing and limited-traversing visualization in the study of the application of time series data,the conclusion shows that both methods can effectively map the time series into a complex network,but the construction of the limited-traversing visualization method needs to select a suitable cross-view distance.Different cross-line of sight leads to changes in the network connection.Based on the degree distribution,betweenness and kernel indicators,it is found that the two methods exhibit different characteristics.Compared with the limited crossing view method,the fluctuation characteristics and actual time series data of the degree distribution and betweenness and kernel are shown in the non-traversing viewable method more consistent.2.A new structure entropy is constructed based on the similarity of neighborhoods.The new metric breaks through the limit of connection by considering the relationship between the intersection and union of neighbor nodes.It is confirmed by several networks with special structures that the new structural entropy can distinguish More types of network diagrams.Then apply it to linearly increasing time series data,periodic time series data and a few simple actual time series data,and found that the new entropy can distinguish the network structure of these types of time series data.In this paper,a new structural entropy is constructed.Its purpose is to effectively obtain the associated information between nodes and their neighbor nodes,avoiding certain limitations in obtaining network global information,and is more conducive to the application and promotion of time-varying complex networks.3.Finally,based on the non-traversing view method and the structural entropy proposed in this paper,the stock time series data of six countries(the United States,Malaysia,Switzerland,Belgium,China,and Italy)and the stock time series data and2020.01.23-2020.04.08 Hubei province and the whole country(except Time series data of new coronary pneumonia(outside Hubei Province).Research has shown that the complex networks converted from non-traversable views have power-law characteristics and small-world characteristics.Network statistical indicators also reflect these two data characteristics.Based on the calculation of entropy in this paper,it is found that in the stock network,the corresponding entropy values of the two largest and second largest economies in the world,the United States and China,are the largest and the second largest respectively.There is a big difference in the entropy value of the new crown pneumonia data(according to Hubei Province,excluding Hubei Province).The research in this article points out several issues that should be paid attention to in the research of time series data by visual views,and constructs a new measurement method for studying time series data through complex networks.At the same time,a more effective analysis is carried out on the current actual time series data.