Search results for "Time serie"
showing 10 items of 261 documents
Prediction of Highly Non-stationary Time Series Using Higher-Order Neural Units
2017
Adaptive predictive models can use conventional and nonconventional neural networks for highly non-stationary time series prediction. However, conventional neural networks present a series of known drawbacks. This paper presents a brief discussion about this concern as well as how the basis of higher-order neural units can overcome some of them; it also describes a sliding window technique alongside the batch optimization technique for capturing the dynamics of non-stationary time series over a Quadratic Neural Unit, a special case of higher-order neural units. Finally, an experimental analysis is presented to demonstrate the effectiveness of the proposed approach.
Online topology estimation for vector autoregressive processes in data networks
2017
An important problem in data sciences pertains to inferring causal interactions among a collection of time series. Upon modeling these as a vector autoregressive (VAR) process, this paper deals with estimating the model parameters to identify the underlying causality graph. To exploit the sparse connectivity of causality graphs, the proposed estimators minimize a group-Lasso regularized functional. To cope with real-time applications, big data setups, and possibly time-varying topologies, two online algorithms are presented to recover the sparse coefficients when observations are received sequentially. The proposed algorithms are inspired by the classic recursive least squares (RLS) algorit…
Graph recursive least squares filter for topology inference in causal data processes
2017
In this paper, we introduce the concept of recursive least squares graph filters for online topology inference in data networks that are modelled as Causal Graph Processes (CGP). A Causal Graph Process (CGP) is an auto regressive process in the time series associated to different variables, and whose coefficients are the so-called graph filters, which are matrix polynomials with different orders of the graph adjacency matrix. Given the time series of data at different variables, the goal is to estimate these graph filters, hence the associated underlying adjacency matrix. Previously proposed algorithms have focused on a batch approach, assuming implicitly stationarity of the CGP. We propose…
Comment améliorer la prévision des ventes pour le marketing ? Les apports de la théorie du chaos
2013
National audience; La littérature en marketing constate un décalage entre les avancées réalisées par les chercheurs qui développent de nouvelles méthodes de prévision des ventes, et l'usage massif de méthodes traditionnelles reposant sur l'hypothèse de linéarité des processus analysés. Cette recherche expose la contribution poten¬tielle de la théorie du chaos à l'amélioration de la prévision des ventes. Une illustration de ces apports est proposée avec une application à la prévision des ventes de consoles de jeux vidéo au Japon. Les résultats mettent en évidence la capacité de la méthode proposée à détecter la présence de chaos dans la série et montrent la possibilité de préciser l'horizon …
A new method for forecasting energy output of a large-scale solar power plant based on long short-term memory networks a case study in Vietnam
2021
Abstract This paper proposes a new model for short-term forecasting power generation capacity of large-scale solar power plant (SPP) in Vietnam considering the fluctuations of weather factors when applying the Long Short-Term Memory networks (LSTM) algorithm. At first, a configuration of the model based on the LSTM algorithm is selected in accordance with the weather and operating conditions of SPP in Vietnam. Not only different structures of LSTM model but also other conventional forecasting methods for time series data are compared in terms of error accuracy of forecast on test data set to evaluate the effectiveness and select the most suitable LSTM configuration. The most suitable config…
Interaction between seismicity and deformation on different time scales in volcanic areas: Campi Flegrei and Stromboli
2019
We study oscillations recorded at Stromboli and Campi Flegrei by different sensors: seismometers, strainmeters and tiltmeters. We examine both the high-frequency (>0.5 Hz) portion of the spectrum and very long period signals up to tidal scales. In this context, seismicity and deformation are investigated on different time scales (from minutes to days/years) in order to identify the basic elements of their interaction, whose understanding should provide new insights on the predictive models. In this work, the strict relation of tides and volcanic processes is shown. At Stromboli, indeed the transition from the stationary phase to the non-stationary phase seems to have a tidal precu…
Mutual nonlinear prediction as a tool to evaluate coupling strength and directionality in bivariate time series: Comparison among different strategie…
2008
We compare the different existing strategies of mutual nonlinear prediction regarding their ability to assess the coupling strength and directionality of the interactions in bivariate time series. Under the common framework of $k$-nearest neighbor local linear prediction, we test three approaches based on cross prediction, mixed prediction, and predictability improvement. The measures of interdependence provided by these approaches are first evaluated on short realizations of bivariate time series generated by coupled Henon models, investigating also the effects of noise. The usefulness of the three mutual nonlinear prediction schemes is then assessed in a common physiological application d…
Quantifying the complexity of short-term heart period variability through K nearest neighbor local linear prediction
2008
The complexity of short-term heart period (HP) variability was quantified exploiting the paradigm that associates the degree of unpredictability of a time series to its dynamical complexity. Complexity was assessed through k-nearest neighbor local linear prediction. A proper selection of the parameter k allowed us to perform either linear or nonlinear prediction, and the comparison of the two approaches to infer the presence of nonlinear dynamics. The method was validated on simulations reproducing linear and nonlinear time series with varying levels of predictability. It was then applied to HP variability series measured from healthy subjects during head-up tilt test, showing that short-te…
On estimating contemporaneous quarterly regional GDP
2007
Subnational regional jurisdictions rarely have at their disposal a reasonable array of timely statistics to monitor their economic condition. In light of this, we develop a procedure that simultaneously estimates a quarterly time series for all regions of a country based upon quarterly national and annual regional data. While other such techniques exist, we suggest a temporal error structure that eliminates possible spurious jumps. Using our approach, regional analysts should now be able to distribute national growth among regions as soon as quarterly national figures are released. In a Spanish application, we detail some practicalities of the process and show that our proposal produces bet…
Fluctuation patterns in high-frequency financial asset returns
2008
We introduce a new method for quantifying pattern-based complex short-time correlations of a time series. Our correlation measure is 1 for a perfectly correlated and 0 for a random walk time series. When we apply this method to high-frequency time series data of the German DAX future, we find clear correlations on short time scales. In order to subtract trivial autocorrelation parts from the pattern conformity, we introduce a simple model for reproducing the antipersistent regime and use alternatively level 1 quotes. When we remove the pattern conformity of this stochastic process from the original data, remaining pattern-based correlations can be observed.