Generation Of Synthetic Time Series Research Articles

Comparative assessment of synthetic time series generation approaches in healthcare: leveraging patient metadata for accurate data synthesis

BackgroundSynthetic data is an emerging approach for addressing legal and regulatory concerns in biomedical research that deals with personal and clinical data, whether as a single tool or through its combination with other privacy enhancing technologies. Generating uncompromised synthetic data could significantly benefit external researchers performing secondary analyses by providing unlimited access to information while fulfilling pertinent regulations. However, the original data to be synthesized (e.g., data acquired in Living Labs) may consist of subjects’ metadata (static) and a longitudinal component (set of time-dependent measurements), making it challenging to produce coherent synthetic counterparts.MethodsThree synthetic time series generation approaches were defined and compared in this work: only generating the metadata and coupling it with the real time series from the original data (A1), generating both metadata and time series separately to join them afterwards (A2), and jointly generating both metadata and time series (A3). The comparative assessment of the three approaches was carried out using two different synthetic data generation models: the Wasserstein GAN with Gradient Penalty (WGAN-GP) and the DöppelGANger (DGAN). The experiments were performed with three different healthcare-related longitudinal datasets: Treadmill Maximal Effort Test (TMET) measurements from the University of Malaga (1), a hypotension subset derived from the MIMIC-III v1.4 database (2), and a lifelogging dataset named PMData (3).ResultsThree pivotal dimensions were assessed on the generated synthetic data: resemblance to the original data (1), utility (2), and privacy level (3). The optimal approach fluctuates based on the assessed dimension and metric.ConclusionThe initial characteristics of the datasets to be synthesized play a crucial role in determining the best approach. Coupling synthetic metadata with real time series (A1), as well as jointly generating synthetic time series and metadata (A3), are both competitive methods, while separately generating time series and metadata (A2) appears to perform more poorly overall.

A Synthetic Time-Series Generation Using a Variational Recurrent Autoencoder with an Attention Mechanism in an Industrial Control System.

Data scarcity is a significant obstacle for modern data science and artificial intelligence research communities. The fact that abundant data are a key element of a powerful prediction model is well known through various past studies. However, industrial control systems (ICS) are operated in a closed environment due to security and privacy issues, so collected data are generally not disclosed. In this environment, synthetic data generation can be a good alternative. However, ICS datasets have time-series characteristics and include features with short- and long-term temporal dependencies. In this paper, we propose the attention-based variational recurrent autoencoder (AVRAE) for generating time-series ICS data. We first extend the evidence lower bound of the variational inference to time-series data. Then, a recurrent neural-network-based autoencoder is designed to take this as the objective. AVRAE employs the attention mechanism to effectively learn the long-term and short-term temporal dependencies ICS data implies. Finally, we present an algorithm for generating synthetic ICS time-series data using learned AVRAE. In a comprehensive evaluation using the ICS dataset HAI and various performance indicators, AVRAE successfully generated visually and statistically plausible synthetic ICS data.

TSGBench: Time Series Generation Benchmark

Synthetic Time Series Generation (TSG) is crucial in a range of applications, including data augmentation, anomaly detection, and privacy preservation. Although significant strides have been made in this field, existing methods exhibit three key limitations: (1) They often benchmark against similar model types, constraining a holistic view of performance capabilities. (2) The use of specialized synthetic and private datasets introduces biases and hampers generalizability. (3) Ambiguous evaluation measures, often tied to custom networks or downstream tasks, hinder consistent and fair comparison. To overcome these limitations, we introduce TSGBench, the inaugural Time Series Generation Benchmark, designed for a unified and comprehensive assessment of TSG methods. It comprises three modules: (1) a curated collection of publicly available, real-world datasets tailored for TSG, together with a standardized preprocessing pipeline; (2) a comprehensive evaluation measures suite including vanilla measures, new distance-based assessments, and visualization tools; (3) a pioneering generalization test rooted in Domain Adaptation (DA), compatible with all methods. We have conducted comprehensive experiments using TSGBench across a spectrum of ten real-world datasets from diverse domains, utilizing ten advanced TSG methods and twelve evaluation measures. The results highlight the reliability and efficacy of TSGBench in evaluating TSG methods. Crucially, TSGBench delivers a statistical analysis of the performance rankings of these methods, illuminating their varying performance across different datasets and measures and offering nuanced insights into the effectiveness of each method.

Data augmentation through multivariate scenario forecasting in Data Centers using Generative Adversarial Networks

The Cloud paradigm is at a critical point in which the existing energy-efficiency techniques are reaching a plateau, while the computing resources demand at Data Center facilities continues to increase exponentially. The main challenge in achieving a global energy efficiency strategy based on Artificial Intelligence is that we need massive amounts of data to feed the algorithms. This paper proposes a time-series data augmentation methodology based on synthetic scenario forecasting within the Data Center. For this purpose, we will implement a powerful generative algorithm: Generative Adversarial Networks (GANs). Specifically, our work combines the disciplines of GAN-based data augmentation and scenario forecasting, filling the gap in the generation of synthetic data in DCs. Furthermore, we propose a methodology to increase the variability and heterogeneity of the generated data by introducing on-demand anomalies without additional effort or expert knowledge. We also suggest the use of Kullback-Leibler Divergence and Mean Squared Error as new metrics in the validation of synthetic time series generation, as they provide a better overall comparison of multivariate data distributions. We validate our approach using real data collected in an operating Data Center, successfully generating synthetic data helpful for prediction and optimization models. Our research will help optimize the energy consumed in Data Centers, although the proposed methodology can be employed in any similar time-series-like problem.

Stochastic Modeling of Hydroclimatic Processes Using Vine Copulas

The generation of synthetic time series is important in contemporary water sciences for their wide applicability and ability to model environmental uncertainty. Hydroclimatic variables often exhibit highly skewed distributions, intermittency (that is, alternating dry and wet intervals), and spatial and temporal dependencies that pose a particular challenge to their study. Vine copula models offer an appealing approach to generate synthetic time series because of their ability to preserve any marginal distribution while modeling a variety of probabilistic dependence structures. In this work, we focus on the stochastic modeling of hydroclimatic processes using vine copula models. We provide an approach to model intermittency by coupling Markov chains with vine copula models. Our approach preserves first-order auto- and cross-dependencies (correlation). Moreover, we present a novel framework that is able to model multiple processes simultaneously. This method is based on the coupling of temporal and spatial dependence models through repetitive sampling. The result is a parsimonious and flexible method that can adequately account for temporal and spatial dependencies. Our method is illustrated within the context of a recent reliability assessment of a historical hydraulic structure in central Mexico. Our results show that by ignoring important characteristics of probabilistic dependence that are well captured by our approach, the reliability of the structure could be severely underestimated.

A Methodology for Validating Diversity in Synthetic Time Series Generation

In order for researchers to deliver robust evaluations of time series models, it often requires high volumes of data to ensure the appropriate level of rigor in testing. However, for many researchers, the lack of time series presents a barrier to a deeper evaluation. While researchers have developed and used synthetic datasets, the development of this data requires a methodological approach to testing the entire dataset against a set of metrics which capture the diversity of the dataset. Unless researchers are confident that their test datasets display a broad set of time series characteristics, it may favor one type of predictive model over another. This can have the effect of undermining the evaluation of new predictive methods. In this paper, we present a new approach to generating and evaluating a high number of time series data. The construction algorithm and validation framework are described in detail, together with an analysis of the level of diversity present in the synthetic dataset.

Multi-objective closed-form algebraic expressions discovery approach application to the synthetic time-series generation

Time-series modeling is a well-studied topic of classical analysis and machine learning. However, large datasets are required to obtain the model with a better prediction quality with the increasing model complexity. Therefore, some applications demand synthetic datasets that are preserving modeling-sensitive properties. Another application of synthetic data is data anonymization. The synthetic data generation algorithm may be split into two parts: the time-series modeling and the synthetic data generation parts. The model must be interpretable to obtain the synthetic data with good quality. The model parameter interpretation allows controlling generation by adding noise to different groups of parameters. In the paper, the evolutionary multi-objective closed-form algebraic expressions discovery approach that allows obtaining the model in the form that may be analyzed using the mathematics is proposed. The analysis allows the interpretation of the model parameters for the controllable generation of the synthetic data. The notion of synthetic data quality is discussed. The examples of the synthetic time-series generation based on two datasets with different properties are shown.

Synthetic Time Series Generation Model for Analysis of Power System Operation and Expansion with High Renewable Energy Penetration

The increasing integration of renewable energy sources into current power systems has posed the challenge of adequately representing the statistical properties associated with their variable power generation. In this paper, a novel procedure is proposed to select a proper synthetic time series generation model for renewable energy sources to analyze power system problems. The procedure takes advantage of the objective of the specific analysis to be performed and the statistical characteristics of the available time series. The aim is to determine the suitable model to be used for generating synthetic time series of renewable energy sources. A set of indicators is proposed to verify that the statistical properties of synthetic time series fit the statistical properties of the original data. The proposal can be integrated into systematic tools available for data analysis without compromising the representation of the statistical properties of the original time series. The procedure is tested using real data from the New Zealand power system in a midterm analysis on integrating wind power plants into the power system. The results show that the proposed procedure reduces the error obtained in analyzing power systems compared with reference models.

Implications of the OEN mixture model of the mean wind vector for the generation of synthetic timeseries and for the assessment of extremes

This paper extends the results of the “fuzzy” OEN mixture model to estimate the probability densities, autocorrelations and spectra of the individual wind mechanisms within the mixed climates of Rome: Fiumicino and Ciampino, and discusses the implications for synthesising long-term simulation of the surface wind vector and for the assessment of extreme mean wind speeds. The paper confirms, from observations, that the strong diurnal effects observed at these locations are confined to a low range of wind speeds and do not significantly affect the extreme wind speeds for annual, or longer, epochs. A procedure is proposed to simulate exceptionally long timeseries of the mean wind vector from the OEN mixture model to examine convergence of the upper tail of the extreme wind speed distribution to the presumed asymptote.

Simulation and Evaluation of a Hybrid Renewable Energy System for Supplying a Desalination Unit on the Island of Lipsi, Greece

Water scarcity is a serious problem for the Aegean Islands in Greece. Due to the fact that tourist development grows in a continuous way, the situation has deteriorated over the last years. Current water resources management practices involving the exploitation of the groundwater reservoirs have provoked the salty water intrusion into the aquifers and in many arid islands water is transported by sea, at a considerably high cost (reaches about 12 €/m³ in some cases). Desalination is foreseen as a solution to this problem and it has already been adopted in many islands, as it is a process that can provide fresh and potable water in the required quantities, at a much lower cost. The coupling of desalination with renewable energy sources (RES) constitutes an appealing and promising option. This paper presents an integrated case study regarding the design and operation of a water-energy system for meeting irrigation and potable water demand in Lipsi Island (Dodecanese complex, Greece). As the desalination unit operation depends on the wind power, a detailed description regarding the generation of synthetic time series of wind speed data is also presented. Finally, a Cost-Benefit Analysis is carried out to discuss each scenario we examine from a financial perspective.

Correlated synthetic time series generation for energy system simulations using Fourier and ARMA signal processing

Correlated synthetic time series generation for energy system simulations using Fourier and ARMA signal processing

A Comparison Between Reconstruction Methods for Generation of Synthetic Time Series Applied to Wind Speed Simulation

Wind energy is an attractive renewable sources and its prediction is highly essential for multiple applications. Over the literature, there are several studies have been focused on the related researches of synthetic wind speed data generation. In this research, two reconstruction methods are developed for synthetic wind speed time series generation. The modeling is constructed based on different processes including independent values generation from the known probability distribution function, rearrangement of random values and segmentation. They have been named as Rank-wise and Step-wise reconstruction methods. The proposed methods are explained with the help of a standard time series and the examination on wind speed time series collected from Galicia, the autonomous region in the northwest of Spain. Results evidenced the potential of the developed models over the state-of-the-art synthetic time series generation methods and demonstrated a successful validation using the means of mean and median wind speed values, autocorrelations, probability distribution parameters with their corresponding histograms and confusion matrix. Pros and cons of both methods are discussed comprehensively.

Simulation of Stochastic Processes Exhibiting Any‐Range Dependence and Arbitrary Marginal Distributions

AbstractHydrometeorological processes are typically characterized by temporal dependence, short‐ or long‐range (e.g., Hurst behavior), as well as by non‐Gaussian distributions (especially at fine time scales). The generation of long synthetic time series that resembles the marginal and joint properties of the observed ones is a prerequisite in many uncertainty‐related hydrological studies, since they can be used as inputs and hence allow the propagation of natural variability and uncertainty to the typically deterministic water‐system models. For this reason, it has been for years one of the main research topics in the field of stochastic hydrology. This work presents a novel model for synthetic time series generation, termed Symmetric Moving Average (neaRly) To Anything, that holds out the promise of simulating stationary univariate and multivariate processes with any‐range dependence and arbitrary marginal distributions, provided that the former is feasible and the latter have finite variance. This is accomplished by utilizing a mapping procedure in combination with the relationship that exists between the correlation coefficients of an auxiliary Gaussian process and a non‐Gaussian one, formalized through the Nataf's joint distribution model. The generality of Symmetric Moving Average (neaRly) To Anything is stressed through two hypothetical simulation studies (univariate and multivariate), characterized by different dependencies and distributions. Furthermore, we demonstrate the practical aspects of the proposed model through two real‐world cases, one that concerns the generation of annual non‐Gaussian streamflow time series at four stations and another that involves the synthesis of intermittent, non‐Gaussian, daily rainfall series at a single location.

Digital generation of non‐Gaussian random vibration signals in railway transportation and package response analysis

SummaryThis paper presents two methods for generation artificial acceleration time series focused on the simulation of non‐Gaussian spiky fluctuation features. The random acceleration time series were recorded in railway boxcar; the simulations were performed under the PSD and higher statistical moments (skewness and kurtosis) restrictions. The first method is based on discrete Fourier transform method with EARPG(1) model, which, in nature, is a Fourier representation of time series with special selection of random Fourier phases. The synthetic time series generation are presented for a given target PSD property, skewness, and kurtosis. In the second method, we characterized the stationary and non‐Gaussian process using polynomial chaos expansion and Karhunen‐Loeve expansion approach based on the target PSD, skewness, and kurtosis. The accuracy of the methods are verified by moments and PSD comparison. The package is simplified to be a SDOF system; the acceleration response characteristics of the system are analyzed; the results can be applied in packaging system behaviors evaluation and packaging design optimization.

A Cautionary Note on the Reproduction of Dependencies through Linear Stochastic Models with Non-Gaussian White Noise

Since the prime days of stochastic hydrology back in 1960s, autoregressive (AR) and moving average (MA) models (as well as their extensions) have been widely used to simulate hydrometeorological processes. Initially, AR(1) or Markovian models with Gaussian noise prevailed due to their conceptual and mathematical simplicity. However, the ubiquitous skewed behavior of most hydrometeorological processes, particularly at fine time scales, necessitated the generation of synthetic time series to also reproduce higher-order moments. In this respect, the former schemes were enhanced to preserve skewness through the use of non-Gaussian white noise— a modification attributed to Thomas and Fiering (TF). Although preserving higher-order moments to approximate a distribution is a limited and potentially risky solution, the TF approach has become a common choice in operational practice. In this study, almost half a century after its introduction, we reveal an important flaw that spans over all popular linear stochastic models that employ non-Gaussian white noise. Focusing on the Markovian case, we prove mathematically that this generating scheme provides bounded dependence patterns, which are both unrealistic and inconsistent with the observed data. This so-called “envelope behavior” is amplified as the skewness and correlation increases, as demonstrated on the basis of real-world and hypothetical simulation examples.

Stochastic Model for Generation of High-Resolution Irradiance Data and Estimation of Power Output of Photovoltaic Plants

This paper presents a method for the generation of synthetic time series of solar radiation with a one-minute time resolution using an automatic cloud classification procedure. Solar radiation data from ground-based measurements are used to determine the correlation between the irradiance and the prevailing cloud classes extracted from satellite images. A specific set of Markov chains for each cloud class is adjusted empirically and used for a stochastic simulation of clear sky index time series. The different models are then combined to convert the irradiance into power output time series of photovoltaic power plants of different sizes. Tests to determine the methodology's performance showed positive results in reproducing the statistical characteristics of observed time series data. Estimates of a selected set of metrics were obtained for several sites in Brazil and allowed the characterization of the solar resource for photovoltaic plants of various sizes and mounting methods considering the generation potential and short-term variability.

A multivariate stochastic model for the generation of synthetic time series at multiple time scales reproducing long-term persistence

A time series generator is presented, employing a robust three-level multivariate scheme for stochastic simulation of correlated processes. It preserves the essential statistical characteristics of historical data at three time scales (annual, monthly, daily), using a disaggregation approach. It also reproduces key properties of hydrometeorological and geophysical processes, namely the long-term persistence (Hurst–Kolmogorov behaviour), the periodicity and intermittency. Its efficiency is illustrated through two case studies in Greece. The first aims to generate monthly runoff and rainfall data at three reservoirs of the hydrosystem of Athens. The second involves the generation of daily rainfall for flood simulation at five rain gauges. In the first emphasis is given to long-term persistence – a dominant characteristic in the management of large-scale hydrosystems, comprising reservoirs with carry-over storage capacity. In the second we highlight to the consistent representation of intermittency and asymmetry of daily rainfall, and the distribution of annual daily maxima.

Stochastic analysis and simulation of hydrometeorological processes associated with wind and solar energy

The current model for energy production, based on the intense use of fossil fuels, is both unsustainable and environmentally harmful and consequently, a shift is needed in the direction of integrating the renewable energy sources into the energy balance. However, these energy sources are unpredictable and uncontrollable as they strongly depend on time varying and uncertain hydrometeorological variables such as wind speed, sunshine duration and solar radiation. To study the design and management of renewable energy systems we investigate both the properties of marginal distributions and the dependence properties of these natural processes, including possible long-term persistence by estimating and analyzing the Hurst coefficient. To this aim we use time series of wind speed and sunshine duration retrieved from European databases of daily records. We also study a stochastic simulation framework for both wind and solar systems using the software system Castalia, which performs multivariate and multi-time-scale stochastic simulation, in order to conduct simultaneous generation of synthetic time series of wind speed and sunshine duration, on yearly, monthly and daily scale.

MCMC for Wind Power Simulation

This paper contributes a Markov chain Monte Carlo (MCMC) method for the direct generation of synthetic time series of wind power output. It is shown that obtaining a stochastic model directly in the wind power domain leads to reduced number of states and to lower order of the Markov chain at equal power data resolution. The estimation quality of the stochastic model is positively influenced since in the power domain, a lower number of independent parameters is estimated from a given amount of recorded data. The simulation results prove that this method offers excellent fit for both the probability density function and the autocorrelation function of the generated wind power time series. The method is a first step toward simple stochastic black-box models for wind generation.

Choice of reliability, resilience and vulnerability estimators for risk assessments of water resources systems / Choix d’estimateurs de fiabilité, de résilience et de vulnérabilité pour les analyses de risque de systèmes de ressources en eau

Definitions and estimators of water resources system reliability (the probability that the system will remain in a non-failure state), resilience (the ability of the system to return to non-failure state after a failure has occurred) and vulnerability (the likely damage of a failure event) have been thoroughly investigated. A behaviour analysis addressing monotonic behaviour, overlap and correlation between the estimators was carried out by routing time series of monthly runoff through a reservoir with a specified storage volume that is operated according to a fixed operation policy. Estimation based on historical time series is shown to be problematic and a procedure encompassing generation of synthetic time series with a length of at least 1000 years is recommended in order to stabilize the estimates. Moreover, the strong correlation between resilience and vulnerability may suggest that resilience should not be explicitly accounted for.