Compound method of time series classification
Łukasz Korus2015, Nonlinear Analysis-Modelling and Control
Sign up for access to the world's latest research
checkGet notified about relevant papers
checkSave papers to use in your research
checkJoin the discussion with peers
checkTrack your impact
Abstract
Many real phenomenona preserves the properties of chaotic dynamics. However, unambiguous determination of belonging to a group of chaotic systems is difficult and complex problem. The main purpose of this paper is to present compound method of time series classification which is basically directed to the detection of chaotic behaviors. The method has been designed for differentiation of three types of time series: chaotic, periodic and random. Our approach assumes, that more reliable information about the dynamics of the system will provide the compilation of several methods, than any individual. This paper focuses on choosing a good set of methods and analysis of their results. In our investigation, we used the following methods and indicators: time delay embedding, mutual information, saturation of system invariants, the largest Lyapunov exponent and Hurst exponent. We checked the validity of the methods applying them to three kinds of basic systems which generate chaotic, periodic and random time series. As a summary of this paper, all selected methods and indicators computed for generated times series have been summarized in the table, which gives the authors a possibility to conclude about type of observed behavior.
Related papers
Some Aspects of Chaotic Time Series Analysis
2001
We address two aspects in chaotic time series analysis, namely the definition of embedding parameters and the largest Lyapunov exponent. It is necessary for performing state space reconstruction and identification of chaotic behavior. For the first aspect, we examine the mutual information for determination of time delay and false nearest neighbors method for choosing appropriate embedding dimension. For the second aspect we suggest neural network approach, which is characterized by simplicity and accuracy.
Recent Developments in Chaotic Time Series Analysis
International Journal of Bifurcation and Chaos, 2003
In this paper, two issues are addressed: (1) the applicability of the delay-coordinate embedding method to transient chaotic time series analysis, and (2) the Hilbert transform methodology for chaotic signal processing.A common practice in chaotic time series analysis has been to reconstruct the phase space by utilizing the delay-coordinate embedding technique, and then to compute dynamical invariant quantities of interest such as unstable periodic orbits, the fractal dimension of the underlying chaotic set, and its Lyapunov spectrum. As a large body of literature exists on applying the technique to time series from chaotic attractors, a relatively unexplored issue is its applicability to dynamical systems that exhibit transient chaos. Our focus will be on the analysis of transient chaotic time series. We will argue and provide numerical support that the current delay-coordinate embedding techniques for extracting unstable periodic orbits, for estimating the fractal dimension, and f...
A proposal for regime change/duration classification in chaotic systems
Proceedings of the 2015 Conference of the International Fuzzy Systems Association and the European Society for Fuzzy Logic and Technology, 2015
In order to to predict regime duration in a given chaotic system, for a set of output prototypes are available, we propose to use a clustering technique for the definition of classes of regime duration, which are then used by a chosen classifier. In this way, the exact boundaries between classes are allowed to emerge from the data, as long as prototypical values fall in distinct classes. We investigate the use of both unsupervised and semi-supervised fuzzy clustering techniques FCM and ssFCM, as well as the traditional k-Means technique. To classify the data, we use neuro-fuzzy system ANFIS and two decision trees (J48 and NBTree). We apply the procedure on the well-known Lorenz strange attractor, having bred vector counts as input variables.
On the Calculation of Chaotic Features for Nonlinear Time Series
Recent developments in the theory of nonlinear dynamics have paved the way for analyzing signals generated from nonlinear biological systems. This study is aimed at investigating the application of nonlinear analysis in differentiating between patients and healthy persons, as well as in investigating the relation between ergodicity and stationarity in the dynamics of the heart. The nonlinear analysis in this work includes attractor reconstruction, estimation of the correlation dimension, calculation of the largest Lyapunov exponent, the approximate entropy, the sample entropy, and a Poincaré plot. Four groups of electrical cardiograph (ECG) signals have been investigated. Our results, obtained from clinical data, confirm the previous studies; this allows one to distinguish between a healthy group and a group of patients with more confidence than the standard methods for heart rate time series. Furthermore we extended our understanding of heart dynamics using entropies and a Poincaré plot along with the correlation dimension and the largest Lyapunov exponent. We have also obtained the results that stationarity and ergodicity are related to each other in heart dynamics.
Chaotic time series. Part I. Estimation of some invariant properties in state-space
Modeling, Identification and Control: A Norwegian Research Bulletin, 1994
Certain deterministic non-linear systems may show chaotic behaviour. Time series derived from such systems seem stochastic when analyzed with linear techniques. However, uncovering the deterministic structure is important because it allows for construction of more realistic and better models and thus improved predictive capabilities. This paper describes key features of chaotic systems including strange attractors and Lyapunov exponents. The emphasis is on state space reconstruction techniques that are used to estimate these properties, given scalar observations. Data generated from equations known to display chaotic behaviour are used for illustration. A compilation of applications to real data from widely di erent elds is given. If chaos is found to be present, one may proceed to build non-linear models, which is the topic of the second paper in this series.
Identification of suitable embedding dimensions and lags for time series generated by chaotic, finite-dimensional systems
Physical Review E
In the field of nonlinear dynamics, many methods have been proposed to tackle the issue of optimally setting embedding dimension and lag in order to analyze sampled scalar signals. However, intrinsic statistical uncertainties due to the finiteness of input sequences severely hinder a general solution to the problem. A more achievable approach consists of assessing sets of dimension and lag pairs that are equivalently suitable to embed a time series. Here we present a method to identify these sets of embedding pairs, under the hypothesis that the time series of interest is generated by a chaotic, finite-dimensional dynamical system. We first introduce a "distance gauge transformation" based on the analytical forms of correlation integrals corresponding to a Gaussian white noise source. We show that in this new distance gauge, correlation integrals generated by chaotic, finitedimensional systems are characterized by distinctive features, whose absence is a marker of the unsuitability of the underlying embedding choice. By means of a new estimator of the correlation dimension that relies on the new distance gauge, sets of suitable embedding pairs are finally identified by looking at the uniformity of the estimation. The method is completely automatic and was successfully tested on both synthetic and experimental time series. It also provides a tool to estimate the redundance and irrelevance timescales of the system that underlie the time series as well as a lower constraint to the sampling rate. The method is suitable for applications in research fields where a chaotic behavior has to be identified, such as neuroscience, geophysics, and economics.
Chaotic and random dynamics discrimination in financial time series
2013
One of the richest fields for the application of econophysics methods is Finance. In particular, financial markets produces a large amount of ready-to-use time series, which could be subject to statistic scrutiny. The analysis of financial time series by means of permutation information quantifiers derived from Information Theory resulted of great value in order to distinguish random and chaotic paths. In this paper we describe the permutation entropy and permutation statistical complexity. Both metrics form a locus where each planar realization reveals a particular statistic characteristic of the time series under study. We give several successful applications of this methodology. We perform an econophysic application to the sovereign fixed income market.
Distinguishing chaotic and stochastic dynamics from time series by using a multiscale symbolic approach
Physical Review E, 2012
In this paper we introduce a multiscale symbolic information-theory approach for discriminating nonlinear deterministic and stochastic dynamics from time series associated with complex systems. More precisely, we show that the multiscale complexity-entropy causality plane is a useful representation space to identify the range of scales at which deterministic or noisy behaviors dominate the system's dynamics. Numerical simulations obtained from the well-known and widely used Mackey-Glass oscillator operating in a high-dimensional chaotic regime were used as test beds. The effect of an increased amount of observational white noise was carefully examined. The results obtained were contrasted with those derived from correlated stochastic processes and continuous stochastic limit cycles. Finally, several experimental and natural time series were analyzed in order to show the applicability of this scale-dependent symbolic approach in practical situations.
A brief methodological note on chaos theory and its recent applications based on new computer resources
ENERGEIA. Sociedad Ibero-Americana de Metodología Económica (ISSN 1666-5732), 2020
Chaos theory refers to the behaviour of certain deterministic nonlinear dynamical systems whose solutions, although globally stable, are locally unstable. These chaotic systems describe aperiodic, irregular, apparently random and erratic trajectories, i.e., deterministic complex dynamics. One of the properties that derive from this local instability and that allow characterizing these deterministic chaotic systems is their high sensitivity to small changes in the initial conditions, which can be measured by using the so-called Lyapunov exponents. The detection of chaotic behaviour in the underlying generating process of a time series has important methodological implications. When chaotic behaviour is detected, then it can be concluded that the irregularity of the series is not necessarily random, but the result of some deterministic dynamic process. Then, even if such process is unknown, it will be possible to improve the predictability of the time series and even to control or stabilize the evolution of the time series. This article provides a summary of the main current concepts and methods for the detection of chaotic behaviour from time series.
Related papers
Analyses of transient chaotic time series
Physical Review E, 2001
We address the calculation of correlation dimension, the estimation of Lyapunov exponents, and the detection of unstable periodic orbits, from transient chaotic time series. Theoretical arguments and numerical experiments show that the Grassberger-Procaccia algorithm can be used to estimate the dimension of an underlying chaotic saddle from an ensemble of chaotic transients. We also demonstrate that Lyapunov exponents can be estimated by computing the rates of separation of neighboring phase-space states constructed from each transient time series in an ensemble. Numerical experiments utilizing the statistics of recurrence times demonstrate that unstable periodic orbits of low periods can be extracted even when noise is present. In addition, we test the scaling law for the probability of finding periodic orbits. The scaling law implies that unstable periodic orbits of high period are unlikely to be detected from transient chaotic time series.
Chaotic time series part II: System identification and prediction
Arxiv preprint chao-dyn/ …, 1994
This paper is the second in a series of two, and describes the current state of the art in modelling and prediction of chaotic time series.
A New Test for Chaotic Dynamics Using Lyapunov Exponents
SSRN Electronic Journal, 2000
Los Documentos de Trabajo se distribuyen gratuitamente a las Universidades e Instituciones de Investigación que lo solicitan. No obstante están disponibles en texto completo a través de Internet:
Prediction of chaotic time series using computational intelligence
Expert Systems with Applications, 2011
In this paper, two CI techniques, namely, single multiplicative neuron (SMN) model and adaptive neurofuzzy inference system (ANFIS), have been proposed for time series prediction. A variation of particle swarm optimization (PSO) with cooperative sub-swarms, called COPSO, has been used for estimation of SMN model parameters leading to COPSO-SMN. The prediction effectiveness of COPSO-SMN and ANFIS has been illustrated using commonly used nonlinear, non-stationary and chaotic benchmark datasets of Mackey-Glass, Box-Jenkins and biomedical signals of electroencephalogram (EEG). The training and test performances of both hybrid CI techniques have been compared for these datasets.
Identification of Variable Relations from Multivariate Chaotic Time Series Data
2007
Abstract| Identifying relations among di erent variables becomes a very practical issue as it could determine values of the variables, which are hardly or expensively measured, from values of the variables, which can be measured easily. One of the key problems related is to determine embedding dimensions from multivariate data. A simple and e ective method is proposed to nd embedding dimensions in this paper. Several examples are provided.
Detection of nonlinearity and chaoticity in time series using the transportation distance function
Physics Letters A, 2002
We propose a systematic two-step framework to assess the presence of nonlinearity and chaoticity in time series. Although the basic components of this framework are from the well-known paradigm of surrogate data and the concept of short-term predictability, the newly proposed discriminating statistic, the transportation distance function offers several advantages (e.g., robustness against noise and outliers, fewer data requirements) over traditional measures of nonlinearity. The power of this framework is tested on several numerically generated series and the Santa Fe Institute competition series. (S. Basu). mension follows, but there exists no converse theorem . Like dimensions, a number of other measures (e.g., entropies, Lyapunov exponents) have been developed based on concepts from nonlinear dynamics and theory of deterministic chaos, which possess similar problems .
Fuzzy clustering approach for accurate embedding dimension identification in chaotic time series
Integrated Computer-Aided Engineering, 2003
Xiaomo had been heavily involved in real-time construction monitoring of a 450-meter cablestayed concrete bridge, assessing integral-lift steel scaffolds in high-rise building construction under wind loadings, and simulating nonlinear responses of space truss structures under wind excitations. As a research assistant in University of Wyoming from 2000 to 2001, Xiaomo Jiang had participated in investigating the effects of multiple moisture susceptibility testing, freeze-thaw cycles on mechanical properties and strength of Hot Mix Asphalt mixtures.
Behavioural analysis of a time series–A chaotic approach
Sadhana, 2014
Out of the various methods available to study the chaotic behaviour, correlation dimension method (CDM) derived from Grassberger-Procaccia algorithm and False Nearest Neighbour method (FNN) are widely used. It is aimed to study the adaptability of those techniques for Indian rainfall data that is dominated by monsoon. In the present study, five sets of time series data are analyzed using correlation dimension method (CDM) based upon Grassberger-Procaccia algorithm for studying their behaviour. In order to confirm the results arrived from correlation dimension method, FNN and phase randomisation method is also applied to the time series used in the present study to fix the optimum embedding dimension. First series is a deterministic natural number series, the next two series are random number series with two types of distributions; one is uniform and another is normal distributed random number series. The fourth series is Henon data, an erratic data generated from a deterministic non linear equation (classified as chaotic series). After checking the applicability of correlation dimension method for deterministic, stochastic and chaotic data (known series) the method is applied to a rainfall time series observed at Koyna station, Maharashtra, India for its behavioural study. The results obtained from the chaotic analysis revealed that CDM is an efficient method for behavioural study of a time series. It also provides first hand information on the number of dimensions to be considered for time series prediction modelling. The CDM applied to real life rainfall data brings out the nature of rainfall at Koyna station as chaotic. For the rainfall data, CDM resulted in a minimum correlation dimension of one and optimum dimension as five. FNN method also resulted in five dimensions for the rainfall data. The behaviour of the rainfall time series is further confirmed by phase randomisation technique also. The surrogate data derived from randomisation gives entirely different results when compared to the other two techniques used in the present study (CDM and FNN) which confirms the behaviour of rainfall as chaotic. It is also seen that CDM is underestimating the correlation dimension, may be due to higher percentage of zero values in rainfall data. Thus, one should appropriately check the adaptability of CDM for time series having longer zero values.
Estimating the Lyapunov exponents of chaotic time series: a model based method
In this paper, the problem of Lyapunov Exponents (LEs) computation from chaotic time series based on Jacobian approach by using polynomial modelling is considered. The embedding dimension which is an important reconstruction parameter, is interpreted as the most suitable order of model. Based on a global polynomial model fitting to the given data, a novel criterion for selecting the suitable embedding dimension is presented. By considering this dimension as the model order, by evaluating the prediction error of different models, the best nonlinearity degree of polynomial model is estimated. This selected structure is used in each point of the reconstructed state space to model the system dynamics locally and calculate the Jacobian matrices which are used in QR factorization method in the LEs estimation. This procedure is also applied to multivariate time series to include information from other time series and resolve probable shortcoming of the univariate case. Finally, simulation results are presented for some well-known chaotic systems to show the effectiveness of the proposed methodology. 0 2 1 J J J
Characterization of normality of chaotic systems including prediction and detection of anomalies
Accurate prediction and control pervades domains such as engineering, physics, chemistry, and biology. Often, it is discovered that the systems under consideration cannot be well represented by linear, periodic nor random data. It has been shown that these systems exhibit deterministic chaos behavior. Deterministic chaos describes systems which are governed by deterministic rules but whose data appear to be random or quasi-periodic distributions. Deterministically chaotic systems characteristically exhibit sensitive dependence upon initial conditions manifested through rapid divergence of states initially close to one another. Due to this characterization, it has been deemed impossible to accurately predict future states of these systems for longer time scales. Fortunately, the deterministic nature of these systems allows for accurate short term predictions, given the dynamics of the system are well understood. This fact has been exploited in the research community and has resulted in various algorithms for short term predictions. Detection of normality in deterministically chaotic systems is critical in understanding the system sufficiently to able to predict future states. Due to the sensitivity to initial conditions, the detection of normal operational states for a deterministically chaotic system can be challenging. The addition of small perturbations to the system, which may result in bifurcation of the normal states, further complicates the problem. The detection of anomalies and prediction of future states of the chaotic system allows for greater understanding of these systems. The goal of this research is to produce methodologies for determining states of normality for deterministically chaotic systems, detection of anomalous behavior, and the
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.