Change-point detection with recurrence networks

Iwayama, Koji; Hirata, Yoshito; Suzuki, Hideyuki; Aihara, Kazuyuki

doi:10.1587/nolta.4.160

Cited by 6 publications

(4 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A theoretical relation is established between measures of the recurrence networks and phase space properties. ε ‐recurrence networks were also widely used to determine changes in the dynamics of theoretical (Iwayama, Hirata, Suzuki, & Aihara, 2013; Marwan et al, 2009) and real world (Donges, Heitzig, et al, 2011; Fukino, Hirata, & Aihara, 2016) systems.…”

Section: Mapping Univariate Time Series Into Complex Networkmentioning

confidence: 99%

Time series analysis via network science: Concepts and algorithms

Silva

Ribeiro

et al. 2021

WIREs Data Min & Knowl

View full text Add to dashboard Cite

There is nowadays a constant flux of data being generated and collected in all types of real world systems. These data sets are often indexed by time, space, or both requiring appropriate approaches to analyze the data. In univariate settings, time series analysis is a mature field. However, in multivariate contexts, time series analysis still presents many limitations. In order to address these issues, the last decade has brought approaches based on network science. These methods involve transforming an initial time series data set into one or more networks, which can be analyzed in depth to provide insight into the original time series. This review provides a comprehensive overview of existing mapping methods for transforming time series into networks for a wide audience of researchers and practitioners in machine learning, data mining, and time series. Our main contribution is a structured review of existing methodologies, identifying their main characteristics, and their differences. We describe the main conceptual approaches, provide authoritative references and give insight into their advantages and limitations in a unified way and language. We first describe the case of univariate time series, which can be mapped to single layer networks, and we divide the current mappings based on the underlying concept: visibility, transition, and proximity. We then proceed with multivariate time series discussing both single layer and multiple layer approaches. Although still very recent, this research area has much potential and with this survey we intend to pave the way for future research on the topic. This article is categorized under: Fundamental Concepts of Data and Knowledge > Data Concepts Fundamental Concepts of Data and Knowledge > Knowledge Representation

show abstract

Section: Mapping Univariate Time Series Into Complex Networkmentioning

confidence: 99%

Time series analysis via network science: Concepts and algorithms

Silva

Ribeiro

et al. 2021

WIREs Data Min & Knowl

View full text Add to dashboard Cite

show abstract

“…In [11], Marwan et al used the clustering coefficient to detect changes in dynamics using the proximity network paradigm as applied to the logistic map and paleo climate data. More recently degree variance was shown to track changes in dynamics along the bifurcation spectrum of the Rössler system [14], and in [16] Iwayama et al developed a spectral clustering measure for proximity networks and used this in conjunction with surrogate data for change point detection in both Rössler and Lorenz systems.…”

Section: Proximity Networkmentioning

confidence: 99%

“…While the above highlights the potential of proximity networks for use in practical time series analysis, the primary drawback of most of these methods is that the resulting networks are invariant to the relabelling of nodes and hence do not explicitly preserve temporal information. Some exceptions do exist, however, including in [16] where the authors have deliberately added links based on temporal adjacency, in contrast to the general approach where temporally adjacent nodes are sometimes even deliberately disconnected [1]. A new proximity method was recently introduced where nodes (times series points) are connected based on a threshold of relative amplitude [13].…”

Section: Proximity Networkmentioning

confidence: 99%

Time lagged ordinal partition networks for capturing dynamics of continuous dynamical systems

McCullough

Small

Stemler

et al. 2015

Chaos: An Interdisciplinary Journal of Nonlinear Science

144

121

View full text Add to dashboard Cite

We investigate a generalised version of the recently proposed ordinal partition time series to network transformation algorithm. Firstly we introduce a fixed time lag for the elements of each partition that is selected using techniques from traditional time delay embedding.The resulting partitions define regions in the embedding phase space that are mapped to nodes in the network space. Edges are allocated between nodes based on temporal succession thus creating a Markov chain representation of the time series. We then apply this new transformation algorithm to time series generated by the Rössler system and find that periodic dynamics translate to ring structures whereas chaotic time series translate to band or tube-like structures -thereby indicating that our algorithm generates networks whose structure is sensitive to system dynamics. Furthermore we demonstrate that simple network measures including the mean out degree and variance of out degrees can track changes in the dynamical behaviour in a manner comparable to the largest Lyapunov exponent. We also apply the same analysis to experimental time series generated by a diode resonator circuit and show that the network size, mean shortest path length and network diameter are highly sensitive to the interior crisis captured in this particular data set. 1 arXiv:1501.06656v1 [nlin.CD] 27 Jan 2015Within the last ten years a novel approach to time series analysis has emerged whereby data is transformed into a complex network and then analysed using various measures from network science. The choice of transformation algorithm is critical in this process as different methods are inherently more effective at capturing certain aspects of dynamics and less effective at capturing others. In this paper we investigate a recently proposed algorithm known as the method of ordinal partitions. This computationally simple algorithm explicitly embeds temporal information in the network structure by partitioning the time series into a set of symbolic states which become network nodes, and then connecting these nodes based on the transition sequence present in the data. New in this work, we generalise the algorithm by introducing a time lag parameter for the elements in each partition, as is done in traditional methods of time delay embedding. Our results demonstrate that this new approach generates networks which are measurably sensitive to the dynamics present in the source time series, and has the potential to be useful as a tool for change point detection in continuous chaotic systems.

show abstract

“…Rapp, Darmon and Cellucci [59] use a similar notion and propose the 'quadrant scan' method to detect transitions, which has been recently also applied by Zaitouny et al [58] to detect transitions in various kinds of real-world data. Iwayama et al [136] proposed a change point detection approach based on the detection of 'communities' in recurrence networks using spectral clustering. More recently, Goswami et al [137] use a similar notion and propose that communities can be used as an indicator of abrupt transitions in time series.…”

Section: Detecting Dynamical Regimes Using Recurrencesmentioning

confidence: 99%

A Brief Introduction to Nonlinear Time Series Analysis and Recurrence Plots

Goswami

2019

Vibration

View full text Add to dashboard Cite

Nonlinear time series analysis gained prominence from the late 1980s on, primarily because of its ability to characterize, analyze, and predict nontrivial features in data sets that stem from a wide range of fields such as finance, music, human physiology, cognitive science, astrophysics, climate, and engineering. More recently, recurrence plots, initially proposed as a visual tool for the analysis of complex systems, have proven to be a powerful framework to quantify and reveal nontrivial dynamical features in time series data. This tutorial review provides a brief introduction to the fundamentals of nonlinear time series analysis, before discussing in greater detail a few (out of the many existing) approaches of recurrence plot-based analysis of time series. In particular, it focusses on recurrence plot-based measures which characterize dynamical features such as determinism, synchronization, and regime changes. The concept of surrogate-based hypothesis testing, which is crucial to drawing any inference from data analyses, is also discussed. Finally, the presented recurrence plot approaches are applied to two climatic indices related to the equatorial and North Pacific regions, and their dynamical behavior and their interrelations are investigated.

show abstract

Change-point detection with recurrence networks

Cited by 6 publications

References 41 publications

Time series analysis via network science: Concepts and algorithms

Time series analysis via network science: Concepts and algorithms

Time lagged ordinal partition networks for capturing dynamics of continuous dynamical systems

A Brief Introduction to Nonlinear Time Series Analysis and Recurrence Plots

Contact Info

Product

Resources

About