A Time Series Weighted Complex Network Construction Method Based on Viewing Angle Measurement

Abstract

The invention discloses a time series weighted complex network construction method based on visual angle measurement, firstly, data points of a time series are defined as network nodes, then, a connection side between the nodes is determined through analysis on whether a visibility criterion is satisfied between every two data points, finally, size of a visual angle of the data points with the connection side is calculated as a connection side weight among the nodes, and thus, a time series weighted complex network is constructed. According to the time series weighted complex network construction method based on visual angle measurement, provided by the invention, a relation among the time series data points can be described more accurately, and network characteristics of the time series weighted complex network can reflect subtle difference among different signals more accurately and sensitively.

Description

technical field [0001] The invention relates to a complex network analysis method based on time series. In particular, it relates to a time series weighted complex network construction method based on viewing angle measurement. Background technique [0002] In recent years, researchers have discovered that time series-based complex network analysis methods can not only dig out important information hidden in nonlinear time series, but also facilitate the study of complex nonlinear systems that are difficult to accurately describe in theoretical models, providing a basis for different types of complex dynamics. It provides a new perspective for the study of scientific systems, so it has received extensive attention from researchers in various fields. After the pioneering work of small-world networks and scale-free networks, different network construction methods for time series complex networks have been proposed and have been successfully applied to the research of many com...

AI Technical Summary

Problems solved by technology

Phase space complex network is to transform the original time series into a set of multi-dimensional vectors by selecting the appropriate embedding dimension and delay time, and use these vectors as the nodes of the complex network, and then determine the connection edges of the network nodes according to the similarity between the vectors , so as to determine the adjacency matrix of the network. This method has unstable factors in the selection of embedding dimensions and the estimation of delay time, and it is difficult to determine the optimal threshold for connecting edges; Sequence fragments are used as nodes, and the correlation strength between the fragments is used to determine the connection between nodes, thereby constructing a dynamic complex network. This method is also difficult to accurately determine the optimal connection threshold; the visual network construction method is to define data points is a network node, and the connection between data points that meets the visibility criterion is defined as a network connection edge. Compared with the previous two methods, the visual graph network construction method avoids the problem of threshold selection, and the algorithm is more robust, but This network construction method simply maps the relationship between system elements, that is, associated and unrelated, to "1" and "0" in the network, that is, connected and unconnected, ignoring the relationship between the elements of the system. The difference in the degree of correlation among them makes this method unable to accurately describe the internal characteristics of the dynamic system

[0004] The classic recursive complex network construction needs to go through a cumbersome process to determine the optimization parameters such as delay time, embedding dimension and threshold, and the complex network construction based on signal fragments has threshold optimization problems

[0005] The classic time series complex network construction method simply expresses the complex relationship between elements in the dynamical system as a "0" or "1" relationship, ignoring the difference in the degree of correlation between elements, so that its network characteristic parameters cannot accurately reflect the complex system. Intrinsic properties of , leading to poor time series classification and recognition performance

Images