Simultaneous Learning of Nonlinear Manifold and Dynamical Models for High-dimensional Time Series

The goal of this work is to learn a parsimonious and informative representation for high-dimensional time series. Conceptually, this comprises two distinct yet tightly coupled tasks: learning a low-dimensional manifold and modeling the dynamical process. These two tasks have a complementary relationship as the temporal constraints provide valuable neighborhood information for dimensionality reduction and conversely, the low-dimensional space allows dynamics to be learnt efficiently. Solving these two tasks simultaneously allows important information to be exchanged mutually. If nonlinear models are required to capture the rich complexity of time series, then the learning problem becomes harder as the nonlinearities in both tasks are coupled. The proposed solution approximates the nonlinear manifold and dynamics using piecewise linear models. The interactions among the linear models are captured in a graphical model. By exploiting the model structure, efficient inference and learning algorithms are obtained without oversimplifying the model of the underlying dynamical process. Evaluation of the proposed framework with competing approaches is conducted in three sets of experiments: dimensionality reduction and reconstruction using synthetic time series, video synthesis using a dynamic texture database, and human motion synthesis, classification and tracking on a benchmark data set. In all experiments, the proposed approach provides superior performance.

Simultaneous Learning of Non-Linear Manifold and Dynamical Models for High-Dimensional Time Series

The goal of this work is to learn a parsimonious and informative representation for high-dimensional time series. Conceptually, this comprises two distinct yet tightly coupled tasks: learning a low-dimensional manifold and modeling the dynamical process. These two tasks have a complementary relationship as the temporal constraints provide valuable neighborhood information for dimensionality reduction and conversely, the low-dimensional space allows dynamics to be learnt efficiently. Solving these two tasks simultaneously allows important information to be exchanged mutually. If nonlinear models are required to capture the rich complexity of time series, then the learning problem becomes harder as the nonlinearities in both tasks are coupled. The proposed solution approximates the nonlinear manifold and dynamics using piecewise linear models. The interactions among the linear models are captured in a graphical model. The model structure setup and parameter learning are done using a variational Bayesian approach, which enables automatic Bayesian model structure selection, hence solving the problem of over-fitting. By exploiting the model structure, efficient inference and learning algorithms are obtained without oversimplifying the model of the underlying dynamical process. Evaluation of the proposed framework with competing approaches is conducted in three sets of experiments: dimensionality reduction and reconstruction using synthetic time series, video synthesis using a dynamic texture database, and human motion synthesis, classification and tracking on a benchmark data set. In all experiments, the proposed approach provides superior performance.

Continuous Plankton Records - Persistence In Time-Series Of Annual Means Of Abundance Of Zooplankton

Time-series of annual means of abundance of zooplankton of the north-east Atlantic Ocean and the North Sea, for the period 1948 to 1977, show considerable associations between successive years. The seasonal dynamics of the stocks appear to be consistent with at least a proportion of this being due to inherent persistence from year-to-year. Experiments with a simple model suggest that the observed properties of the time-series cannot be reproduced as a response to simple random forcing. The extent of trends and long wavelength variations can be simulated by introducing fairly extensive persistence into the perturbations, but this underestimates the extent of shorter wavelength variability in the observed time-series. The effect of persistence is to increase the proportion of trend and long wavelength variability in time-series of annual means, but stocks can respond to short wavelength perturbations provided these have a clearly defined frequency.

Modelling the impact of antibiotic use and infection control practices on the incidence of hospital-acquired methicillin-resistant Staphylococcus aureus: a time-series analysis

Objectives: Methicillin-resistant Staphylococcus aureus (MRSA) is a major nosocomial pathogen worldwide. A wide range of factors have been suggested to influence the spread of MRSA. The objective of this study was to evaluate the effect of antimicrobial drug use and infection control practices on nosocomial MRSA incidence in a 426-bed general teaching hospital in Northern Ireland.

Methods: The present research involved the retrospective collection of monthly data on the usage of antibiotics and on infection control practices within the hospital over a 5 year period (January 2000–December 2004). A multivariate ARIMA (time-series analysis) model was built to relate MRSA incidence with antibiotic use and infection control practices.

Results: Analysis of the 5 year data set showed that temporal variations in MRSA incidence followed temporal variations in the use of fluoroquinolones, third-generation cephalosporins, macrolides and amoxicillin/clavulanic acid (coefficients = 0.005, 0.03, 0.002 and 0.003, respectively, with various time lags). Temporal relationships were also observed between MRSA incidence and infection control practices, i.e. the number of patients actively screened for MRSA (coefficient = -0.007), the use of alcohol-impregnated wipes (coefficient = -0.0003) and the bulk orders of alcohol-based handrub (coefficients = -0.04 and -0.08), with increased infection control activity being associated with decreased MRSA incidence, and between MRSA incidence and the number of new patients admitted with MRSA (coefficient = 0.22). The model explained 78.4% of the variance in the monthly incidence of MRSA.

Conclusions: The results of this study confirm the value of infection control policies as well as suggest the usefulness of restricting the use of certain antimicrobial classes to control MRSA.

NON-GAUSSIAN STATISTICS OF OIL PRICING TIME-SERIES: A CASE STUDY

The stochastic nature of oil price fluctuations is investigated over a twelve-year period, borrowing feedback from an existing database (USA Energy Information Administration database, available online). We evaluate the scaling exponents of the fluctuations by employing different statistical analysis methods, namely rescaled range analysis (R/S), scale windowed variance analysis (SWV) and the generalized Hurst exponent (GH) method. Relying on the scaling exponents obtained, we apply a rescaling procedure to investigate the complex characteristics of the probability density functions (PDFs) dominating oil price fluctuations. It is found that PDFs exhibit scale invariance, and in fact collapse onto a single curve when increments are measured over microscales (typically less than 30 days). The time evolution of the distributions is well fitted by a Levy-type stable distribution. The relevance of a Levy distribution is made plausible by a simple model of nonlinear transfer. Our results also exhibit a degree of multifractality as the PDFs change and converge toward to a Gaussian distribution at the macroscales.

Detecting sequence dependent transcriptional pauses from RNA and protein number time series

Background: Evidence suggests that in prokaryotes sequence-dependent transcriptional pauses a?ect the dynamics of transcription and translation, as well as of small genetic circuits. So far, a few pause-prone sequences have been identi?ed from in vitro measurements of transcription elongation kinetics.

Results: Using a stochastic model of gene expression at the nucleotide and codon levels with realistic parameter values, we investigate three di?erent but related questions and present statistical methods for their analysis. First, we show that information from in vivo RNA and protein temporal numbers is su?cient to discriminate between models with and without a pause site in their coding sequence. Second, we demonstrate that it is possible to separate a large variety of models from each other with pauses of various durations and locations in the template by means of a hierarchical clustering and a random forest classi?er. Third, we introduce an approximate likelihood function that allows to estimate the location of a pause site.

Conclusions: This method can aid in detecting unknown pause-prone sequences from temporal measurements of RNA and protein numbers at a genome-wide scale and thus elucidate possible roles that these sequences play in the dynamics of genetic networks and phenotype.

Modified Information Criteria and Selection of Long Memory Time Series Models

The problem of model selection of a univariate long memory time series is investigated once a semi parametric estimator for the long memory parameter has been used. Standard information criteria are not consistent in this case. A Modified Information Criterion (MIC) that overcomes these difficulties is introduced and proofs that show its asymptotic validity are provided. The results are general and cover a wide range of short memory processes. Simulation evidence compares the new and existing methodologies and empirical applications in monthly inflation and daily realized volatility are presented.

Structural damage diagnosis of steel truss bridges by outlier detection

This study discusses structural damage diagnosis of real steel truss bridges by measuring trafficinduced vibration of bridges and utilizing a damage indicator derived from linear system parameters of a time series model. On-site damage experiments were carried out on real steel truss bridges. Artificial damage was applied to the bridge by severing a truss member with a cutting machine.Vehicle-induced vibrations of the bridges before and after applying damagewere measured and used in structural damage diagnosis of the bridges. Changes in the damage indicator are detected by Mahalanobis-Taguchi system (MTS) which is one of multivariate outlier analyses. The damage indicator and outlier detection was successfully applied to detect anomalies in the steel truss bridges utilizing vehicle-induced vibrations. Observations through this study demonstrate feasibility of the proposed approach for real world applications.

Bandwidth selection by cross-validation for forecasting long memory financial time series

The paper addresses the issue of choice of bandwidth in the application of semiparametric estimation of the long memory parameter in a univariate time series process. The focus is on the properties of forecasts from the long memory model. A variety of cross-validation methods based on out of sample forecasting properties are proposed. These procedures are used for the choice of bandwidth and subsequent model selection. Simulation evidence is presented that demonstrates the advantage of the proposed new methodology.

Optimal Time Series Momentum

We develop a continuous-time asset price model to capture the timeseries momentum documented recently. The underlying stochastic delay differentialsystem facilitates the analysis of effects of different time horizons used bymomentum trading. By studying an optimal asset allocation problem, we find thatthe performance of time series momentum strategy can be significantly improvedby combining with market fundamentals and timing opportunity with respect tomarket trend and volatility. Furthermore, the results also hold for different timehorizons, the out-of-sample tests and with short-sale constraints. The outperformanceof the optimal strategy is immune to market states, investor sentiment andmarket volatility.

Supervised Aggregative Feature Extraction for Big Data Time Series Regression

In many applications, and especially those where batch processes are involved, a target scalar output of interest is often dependent on one or more time series of data. With the exponential growth in data logging in modern industries such time series are increasingly available for statistical modeling in soft sensing applications. In order to exploit time series data for predictive modelling, it is necessary to summarise the information they contain as a set of features to use as model regressors. Typically this is done in an unsupervised fashion using simple techniques such as computing statistical moments, principal components or wavelet decompositions, often leading to significant information loss and hence suboptimal predictive models. In this paper, a functional learning paradigm is exploited in a supervised fashion to derive continuous, smooth estimates of time series data (yielding aggregated local information), while simultaneously estimating a continuous shape function yielding optimal predictions. The proposed Supervised Aggregative Feature Extraction (SAFE) methodology can be extended to support nonlinear predictive models by embedding the functional learning framework in a Reproducing Kernel Hilbert Spaces setting. SAFE has a number of attractive features including closed form solution and the ability to explicitly incorporate first and second order derivative information. Using simulation studies and a practical semiconductor manufacturing case study we highlight the strengths of the new methodology with respect to standard unsupervised feature extraction approaches.

Optimal alarms systems and its application to financial time series

This thesis focuses on the application of optimal alarm systems to non linear time series models. The most common classes of models in the analysis of real-valued and integer-valued time series are described. The construction of optimal alarm systems is covered and its applications explored. Considering models with conditional heteroscedasticity, particular attention is given to the Fractionally Integrated Asymmetric Power ARCH, FIAPARCH(p; d; q) model and an optimal alarm system is implemented, following both classical and Bayesian methodologies. Taking into consideration the particular characteristics of the APARCH(p; q) representation for financial time series, the introduction of a possible counterpart for modelling time series of counts is proposed: the INteger-valued Asymmetric Power ARCH, INAPARCH(p; q). The probabilistic properties of the INAPARCH(1; 1) model are comprehensively studied, the conditional maximum likelihood (ML) estimation method is applied and the asymptotic properties of the conditional ML estimator are obtained. The final part of the work consists on the implementation of an optimal alarm system to the INAPARCH(1; 1) model. An application is presented to real data series.

Integer-valued APARCH processes in the analysis of time series of counts

The Asymmetric Power Arch representation for the volatility was introduced by Ding et al.(1993) in order to account for asymmetric responses in the volatility in the analysis of continuous-valued financial time series like, for instance, the log-return series of foreign exchange rates, stock indices or share prices. As reported by Brannas and Quoreshi (2010), asymmetric responses in volatility are also observed in time series of counts such as the number of intra-day transactions in stocks. In this work, an asymmetric power autoregressive conditional Poisson model is introduced for the analysis of time series of counts exhibiting asymmetric overdispersion. Basic probabilistic and statistical properties are summarized and parameter estimation is discussed. A simulation study is presented to illustrate the proposed model. Finally, an empirical application to a set of data concerning the daily number of stock transactions is also presented to attest for its practical applicability in data analysis.

Analysis of temperature time-series: embedding dynamics into the MDS method

Global warming and the associated climate changes are being the subject of intensive research due to their major impact on social, economic and health aspects of the human life. Surface temperature time-series characterise Earth as a slow dynamics spatiotemporal system, evidencing long memory behaviour, typical of fractional order systems. Such phenomena are difficult to model and analyse, demanding for alternative approaches. This paper studies the complex correlations between global temperature time-series using the Multidimensional scaling (MDS) approach. MDS provides a graphical representation of the pattern of climatic similarities between regions around the globe. The similarities are quantified through two mathematical indices that correlate the monthly average temperatures observed in meteorological stations, over a given period of time. Furthermore, time dynamics is analysed by performing the MDS analysis over slices sampling the time series. MDS generates maps describing the stations’ locus in the perspective that, if they are perceived to be similar to each other, then they are placed on the map forming clusters. We show that MDS provides an intuitive and useful visual representation of the complex relationships that are present among temperature time-series, which are not perceived on traditional geographic maps. Moreover, MDS avoids sensitivity to the irregular distribution density of the meteorological stations.

Short run and long run causality in time series: Inference

We propose methods for testing hypotheses of non-causality at various horizons, as defined in Dufour and Renault (1998, Econometrica). We study in detail the case of VAR models and we propose linear methods based on running vector autoregressions at different horizons. While the hypotheses considered are nonlinear, the proposed methods only require linear regression techniques as well as standard Gaussian asymptotic distributional theory. Bootstrap procedures are also considered. For the case of integrated processes, we propose extended regression methods that avoid nonstandard asymptotics. The methods are applied to a VAR model of the U.S. economy.