分数布朗运动时间序列的基于递归图的网络分析-应用数学专业论文.docx

I I 摘 要 非线性时间序列分析旨在解释随机或混沌过程的动力学行为。近年来发展 了许多将时间序列转化为复杂网络的方法，如递归图法等。因此我们可以通过 研究网络的拓扑性质来研究原始时间序列的性质。本文用递归图方法将分数布 朗运动(F BM ) 时间序列转化为复杂网络，并研究其对应递归网络的拓扑性质。 我们发现，对固定的Hurst 指数 H ，在网络连通率首次增长到 1 之前，随着 递归图的参数阈值的增大，网络的平均路径长度 L 也随之递增，之后反而递减。 我们也发现由 F BM 时间序列转化得到的网络是无标度网络。我们采用节点覆 盖盒计数法分析发现 F BM 的递归网络为分形网络，具有自相似特性，其分形 维数 dB 随 Hurst 指数 H 的增大而减小，特别当 H ≥ 0.4 时，有近似关系 dB = H2 ? 2.1 × H + 2。 关键词：分数布朗运动；Hurst 指数；递归图；复杂网络；分形维数；网络拓 扑特性 I II Abstract Nonlinear time series analysis aims at revealing the dynamics of stochas- tic or chaotic processes. In recent years, many methods including recurrence plot method have been proposed to convert a single time series to a complex network so that the properties of the original time series can be understood by investigating the topological properties of the network. In this paper，we con- vert fractional Brownian motion (FBM) time series to complex networks using recurrence plot method, then we investigate the topological properties of the cor- responding recurrence networks. We find that for fixed Hurst exponent H, before the connectivity rate of corresponding networks increases to about 1 for the first time, the average length L of the shortest paths increases with the increase of pa- rameter threshold in the recurrence plot, and then decreases. It is also found that the converted networks from FBM time series are scale-free. The analysis using the node-covering box-counting method shows that the recurrence networks are fractals with self-similarity property, and the fractal dimension dB decreases with Hurst index H, especially when H ≥ 0.4, there is a approximately relationship dB = H2 ? 2.1 × H + 2. Key Words: Fractional Brownian motions; Hurst exponnet; recurrence plot; complex networks; fractal dimension; topological property of network 目 录 第一章 综 述 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 研究背景及现状 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .