Stationary Time Series Research Articles

Time Series Analysis and Algorithm for Fluctuation Pattern Recognition and Forecasting of Economic Indicators

This study examines the use of artificial neural network (ANN) algorithms in machine learning for recognizing fluctuation patterns and forecasting economic indicators in the field of political economy. In today's changing global economic scene, economic forecasting is critical for informing policy decisions, directing strategic planning, and managing risks. Traditional statistical approaches and time series analysis have long been used for this purpose, but the advent of ANN algorithms provides a new way to capture the intricate dynamics of economic systems. Using ANNs' intrinsic flexibility and nonlinear modelling capabilities, this study analyzes their ability to effectively forecast future swings in economic indices. This work intends to understand the strengths, limits, and practical consequences of ANN-based forecasting models by conducting a complete assessment of literature, methodological methods, and empirical studies. This study adds to the continuing discussion of economic forecasting approaches by diving into theoretical frameworks, methodological concerns, and empirical data. Finally, incorporating ANN algorithms into machine learning offers promising opportunities for improving the understanding of economic processes and supporting informed decision-making in the political economy.

Effects of major storms over a spit-tidal flat interaction system (Punta Rasa-Samborombón, Argentina)

Sandy coastal areas are very dynamic systems in which morphological changes occur over different time scales from hours to decades. However, it has been widely reported that major storms are the main responsible of the most significant changes in short to medium time scales. Major storms have been defined using a variety of environmental variables, but they are normally associated with high values of wave height, duration, return period and direction.Here, we aim to characterize types of major storms and to categorize associated morphological impacts over a complex coastal system. The study site, known as Punta Rasa, is located in the Samborombón bay in the outer part of the Río de La Plata estuary (Argentina) and corresponds to a zone of interaction between a large sandy spit and a backwash tidal flat system. Methods combine statistics of wave climate time-series, analysis of wave energy using nearshore numerical modelling (SWAN) and comparison of pre- and post-storm morphological changes by means of shoreline change detection and satellite images derived indexes (CoastSat Toolkit and NDWI index respectively).Results allowed to characterize four types of major storms impacting the study area: High-Energy Storms (HES), defined by an average storm wave below the 1 % exceedance (>2.6 m), Long-Lived Storms (LLS) represented by an exceedance of the 1 % of Du (>60 h), Storm Groups (SG) in which storm return period is <6 days and Northeastern moderate storms (NMS) defined by their eastern, onshore oriented direction. Under HES and NMS storms erosional areas are dominant over depositional, causing shoreline retreat, a growth of the end-spit and the increase on sand deposition on the back-barrier areas. Under LLS and SG storms, the morphological impact varies alongshore, with multiple erosional hotspots found along the shoreline accompanied by a general flattens of the end-spit system.

Asymptotic normality of spectral means of Hilbert space valued random processes

A variety of statistics for functional time series allows for a representation as weighted average of corresponding periodogram operators over the frequency domain. We study consistency and asymptotic normality of such spectral mean estimators under mild assumptions. We show that weak convergence of these estimators can be deduced from the (joint) weak convergence of the sample autocovariance operators. The latter is established for a large class of weakly dependent functional time series, which admit expansions as Bernoulli shifts and the weak dependence is quantified by the condition of L4-m-approximability.

Risk Modeling, Return Forecasting, and Optimal Portfolio Selection

The main objective of this research is to form an optimal investment portfolio consisting of a number of stocks selected according to specific criteria in Amman Stock Exchange and to test the ability of a various economic models to predict the performance of this portfolio in the foreseeable future. A time series for the return of the selected portfolio and for the return of a market index are formed. A set of tests were conducted to reach a stationary time series return and, then, to follow the Box-Jenkins methodology in order to build predictive models (ARMA) and to examine the residuals of models and to model them using ARCH and GARCH models to reach the best prediction of the performance of the portfolio and the market index in the forecasted periods. The data were tracked on a daily basis for the study sample and the market index simultaneously for a period of three years. Twenty-one companies were selected in the investment portfolio distributed among several sectors. The study concluded that the formed portfolio achieved a good diversification and gave a high return in relation to the lowest possible risk according to the Sharpe scale. Also, it is concluded that the model ARMA (1,1) is the most suitable for estimating market portfolio returns and forecasting risks for the market index return, and ARMA (2,1) and the model ARMA - GARCH are the most capable one of achieving good results that can be relied upon in tracking the performance of the studied investment portfolio.

Demand forecasting for platelet usage: From univariate time series to multivariable models.

Platelet products are both expensive and have very short shelf lives. As usage rates for platelets are highly variable, the effective management of platelet demand and supply is very important yet challenging. The primary goal of this paper is to present an efficient forecasting model for platelet demand at Canadian Blood Services (CBS). To accomplish this goal, five different demand forecasting methods, ARIMA (Auto Regressive Integrated Moving Average), Prophet, lasso regression (least absolute shrinkage and selection operator), random forest, and LSTM (Long Short-Term Memory) networks are utilized and evaluated via a rolling window method. We use a large clinical dataset for a centralized blood distribution centre for four hospitals in Hamilton, Ontario, spanning from 2010 to 2018 and consisting of daily platelet transfusions along with information such as the product specifications, the recipients' characteristics, and the recipients' laboratory test results. This study is the first to utilize different methods from statistical time series models to data-driven regression and machine learning techniques for platelet transfusion using clinical predictors and with different amounts of data. We find that the multivariable approaches have the highest accuracy in general, however, if sufficient data are available, a simpler time series approach appears to be sufficient. We also comment on the approach to choose predictors for the multivariable models.

Statistical data pre-processing and time series incorporation for high-efficacy calibration of low-cost NO2 sensor using machine learning

Air pollution stands as a significant modern-day challenge impacting life quality, the environment, and the economy. It comprises various pollutants like gases, particulate matter, biological molecules, and more, stemming from sources such as vehicle emissions, industrial operations, agriculture, and natural events. Nitrogen dioxide (NO2), among these harmful gases, is notably prevalent in densely populated urban regions. Given its adverse effects on health and the environment, accurate monitoring of NO2 levels becomes imperative for devising effective risk mitigation strategies. However, the precise measurement of NO2 poses challenges as it traditionally relies on costly and bulky equipment. This has prompted the development of more affordable alternatives, although their reliability is often questionable. The aim of this article is to introduce a groundbreaking method for precisely calibrating cost-effective NO2 sensors. This technique involves statistical preprocessing of low-cost sensor readings, aligning their distribution with reference data. Central to this calibration is an artificial neural network (ANN) surrogate designed to predict sensor correction coefficients. It utilizes environmental variables (temperature, humidity, atmospheric pressure), cross-references auxiliary NO2 sensors, and incorporates short time series of previous readings from the primary sensor. These methods are complemented by global data scaling. Demonstrated using a custom-designed cost-effective monitoring platform and high-precision public reference station data collected over 5 months, every component of our calibration framework proves crucial, contributing to its exceptional accuracy (with a correlation coefficient near 0.95 concerning the reference data and an RMSE below 2.4 µg/m3). This level of performance positions the calibrated sensor as a viable, cost-effective alternative to traditional monitoring approaches.

Multiplier subsample bootstrap for statistics of time series

Block-based bootstrap, block-based subsampling and multiplier bootstrap are three common nonparametric tools for statistical inference under dependent observations. Combining the ideas of those three, a novel resampling approach, the multiplier subsample bootstrap (MSB), is proposed. Instead of generating a resample from the observations, the MSB imitates the statistic by weighting the block-based subsample statistics with independent standard Gaussian random variables. Given the asymptotic normality of the statistic, the bootstrap validity is established under some mild moment conditions. Involving the idea of MSB, the other resampling approach, the hybrid multiplier subsampling periodogram bootstrap (HMP), is developed for mimicking frequency-domain spectral mean statistics in the paper. A simulation study demonstrates that both the MSB and HMP achieve good performance.

Impact of climate change on dengue incidence in Singapore: time-series seasonal analysis

ABSTRACT This study aimed to identify the meteorological factors that contribute to dengue epidemics. The monthly incidence of dengue was used as the outcome variable, while maximum temperature, humidity, precipitation, and sunshine hours were used as independent variables. The results showed a consistent increase in monthly dengue cases from 2013 to 2021, with seasonal patterns observed in stationary time-series data. The ARIMA (2, 1, 3) × seasonal (0, 1, 2)12 model was used based on its lowest Akaike Information Criterion (AIC) values. The analysis revealed that a 1-unit increase in rainfall was positively correlated with a small 0.062-unit increase in dengue cases, whereas a 1-unit increase in humidity was negatively associated, leading to a substantial reduction of approximately 16.34 cases. This study highlights the importance of incorporating weather data into national dengue prevention programs to enhance public awareness and to promote recommended safety measures.

Evidence that the woman's ovarian cycle is driven by an internal circamonthly timing system.

The ovarian cycle has a well-established circa-monthly rhythm, but the mechanisms involved in its regularity are unknown. Is the rhythmicity driven by an endogenous clock-like timer or by other internal or external processes? Here, using two large epidemiological datasets (26,912 cycles from 2303 European women and 4786 cycles from 721 North American women), analyzed with time series and circular statistics, we find evidence that the rhythmic characteristics of the menstrual cycle are more likely to be explained by an endogenous clock-like driving mechanism than by any other internal or external process. We also show that the menstrual cycle is weakly but significantly influenced by the 29.5-day lunar cycle and that the phase alignment between the two cycles differs between the European and the North American populations. Given the need to find efficient treatments of subfertility in women, our results should be confirmed in larger populations, and chronobiological approaches to optimize the ovulatory cycle should be evaluated.

New graphical models for sequential data and the improved state estimations by data-conditioned driving noises

A prevalent problem in statistical signal processing, applied statistics, and time series analysis arises from the attempt to identify the hidden state of Markov process based on a set of available noisy observations. In the context of sequential data, filtering refers to the probability distribution of the underlying Markovian system given the measurements made at or before the time of the estimated state. In addition to the filtering, the smoothing distribution is obtained from incorporating measurements made after the time of the estimated state into the filtered solution. This work proposes a number of new filters and smoothers that, in contrast to the traditional schemes, systematically make use of the process noises to give rise to enhanced performances in addressing the state estimation problem. In doing so, our approaches for the resolution are characterized by the application of the graphical models; the graph-based framework not only provides a unified perspective on the existing filters and smoothers but leads us to design new algorithms in a consistent and comprehensible manner. Moreover, the graph models facilitate the implementation of the suggested algorithms through message passing on the graph.

Forecasting EV Stock Trends Based on ARIMA Model Represented by Tesla and BYD

In recent years, an increasing number of people have turned their attention to renewable energy, and sales of electric vehicles have skyrocketed, making electric vehicles (EVs) a very popular topic. To better understand the development of the electric vehicle market and provide some insights for investors, this study forecasts EV stock trends of two leading companies in the Electric Vehicle industry, Tesla and BYD, by the Autoregressive Integrated Moving Average (ARIMA) model. The data in this research includes a daily time series of stock prices. ARIMA model is chosen in this research for its effectiveness in predicting stationary time series data, for its effectiveness in predicting stationary time series data, particularly suitable given the limited data available in the recently emerged EV industry. This research demonstrates that both Tesla and BYD’s stock closing prices are not stationary data. After differencing, for Tesla, the predicted result suggests a slightly increasing trend, while BYD's prices are predicted to remain at the same level and relatively stable in future 100 days. The study provides insights into the EV stock market, indicating a potential upward yet not pronounced trend in the future. But such a stable price might be a good choice for the investor low-risk takers.

Reducing intravenous antibiotics in neonates born ≥35 weeks' gestation: A quality improvement study.

To assess the impact of the Early Onset Sepsis (EOS) calculator, implemented as a quality improvement study, to reduce the rate of unnecessary antibiotics in neonates born ≥35 weeks' gestation. An audit of routinely collected hospital data from January 2008 to March 2014 (retrospective) and from January 2018 to September 2019 (prospective) determined baseline incidence of EOS intravenous antibiotic use in neonates born ≥35 weeks' gestation in a tertiary level perinatal centre. Plan-do-study-act (PDSA) cycles were applied to implement the EOS calculator. Statistical process control methodology and time series analysis assessments were used to assess the potential impact of the PDSA cycles on the rate of intravenous antibiotics, blood culture collection, EOS, length of stay and health care costs (not adjusted for potential confounders). In the study population, from January 2008 to March 2014, the baseline incidence of intravenous antibiotic use was 10.49% (2970/28290), whilst only 0.067% (19/28290) neonates had culture proven EOS. From January 2018 to October 2019, prior to implementation of the EOS calculator, 13.3% (1119/8411) neonates were treated with intravenous antibiotic and the use decreased to 8.3% (61/734) post-implementation. The rate of blood culture collection decreased from 14.4% (1211/8411) to 11.9% (87/734). There were no cases of missed EOS. Length of stay decreased from 2.68 to 2.39 days, with an estimated cost saving of $366 per patient per admission. Implementing the EOS calculator in a tertiary hospital setting reduced invasive investigations for EOS and intravenous antibiotic use among neonates ≥35 weeks' gestation. This can result in reduced length of neonatal hospital stays, and associated health care cost savings and may reduce separation of mother and baby.

On Sample Size Needed for Block Bootstrap Confidence Intervals to Have Desired Coverage Rates

Block bootstrap is widely used in constructing confidence intervals for parameters estimated from stationary time series. Theoretically, the method should provide valid confidence intervals as the length of the time series goes to infinity. In practice, however, it is necessary to know how large of a finite sample is required for block bootstrap confidence intervals to work well. This study aims to answer this question in a simple simulation setting where the data are generated from a first-order autoregressive process. The empirical coverage rates of several commonly used bootstrap confidence intervals for the mean, standard deviation, and the lag-1 autocorrelation coefficient are compared. A quite large sample is found necessary for the intervals to have the right coverage rates even when estimating a simple parameter like the mean. Some block bootstrap methods could fail when estimating the lag-1 autocorrelation. It is surprising that the coverage property even deteriorates as the sample size increases with some commonly used block bootstrap confidence intervals including the percentile intervals and bias-corrected intervals. KEYWORDS: Autocorrelation; Bias-Correction; Centering; Dependent Data; Percentile; Resampling; Simulation; Time Series

Harmonized European Union subnational crop statistics can reveal climate impacts and crop cultivation shifts

Abstract. The availability of coherent time series of crop statistics is essential for better analyzing the past, understanding the present, and predicting future trends in yield, area, and production. Importantly, such data also underpin assessments and subsequent policy actions that can shape future food systems that are more resilient in the face of climate change and that are sustainable in terms of resource use efficiency. In the European Union (EU), there currently is no legal obligation for EU countries to provide subnational crop statistics. However, such data could improve in-season crop forecasts, climate change impacts and adaption need assessments, and evaluation of agri-environmental schemes. The dataset described in this paper includes a harmonized collection of subnational crop statistics on area, production, and yield, collected for the EU from National Statistical Institutes (NSIs) and the EUROSTAT REGIO database – subnational crop statistics voluntarily contributed by EU countries. The crops considered are wheat (including soft and durum wheat), barley (including winter and spring barley), grain maize, sunflower, and sugar beet. All data are harmonized towards the hierarchical structure of the EUROSTAT legend and the regional classification of NUTS (Nomenclature of Territorial Units for Statistics) version 2016, to provide coherent time series. A total of 344 282 records is presented here (115 974 for wheat, 122 705 for barley, 35 274 for grain maize, 34 916 for sugar beet, and 35 413 for sunflower) for a total of 961 regions in 27 EU countries. Statistics are reported from 1975 to 2020, with a median time-span range of 21 years. A flagging system details, for each data record, information on data sources, processing steps, and quality-checking results. This includes consistency checks between reported values for area, yield and production, identification of null values, missing and calculated data, information on crop legend matching, and NUTS versioning. We illustrate the value of this dataset by analyzing impacts on crops and production zone shifts in Europe due to climatic and economic factors. Recommendations and future developments of collecting subnational statistics at the EU level are briefly discussed. The dataset is accessible with European Commission Authentication Service (ECAS) login at https://doi.org/10.2905/685949ff-56de-4646-a8df-844b5bb5f835 (Ronchetti et al., 2023b).

A review of predictive uncertainty estimation with machine learning

Predictions and forecasts of machine learning models should take the form of probability distributions, aiming to increase the quantity of information communicated to end users. Although applications of probabilistic prediction and forecasting with machine learning models in academia and industry are becoming more frequent, related concepts and methods have not been formalized and structured under a holistic view of the entire field. Here, we review the topic of predictive uncertainty estimation with machine learning algorithms, as well as the related metrics (consistent scoring functions and proper scoring rules) for assessing probabilistic predictions. The review covers a time period spanning from the introduction of early statistical (linear regression and time series models, based on Bayesian statistics or quantile regression) to recent machine learning algorithms (including generalized additive models for location, scale and shape, random forests, boosting and deep learning algorithms) that are more flexible by nature. The review of the progress in the field, expedites our understanding on how to develop new algorithms tailored to users’ needs, since the latest advancements are based on some fundamental concepts applied to more complex algorithms. We conclude by classifying the material and discussing challenges that are becoming a hot topic of research.

Overlapping Batch Confidence Intervals on Statistical Functionals Constructed from Time Series: Application to Quantiles, Optimization, and Estimation

We propose a general purpose confidence interval procedure (CIP) for statistical functionals constructed using data from a stationary time series. The procedures we propose are based on derived distribution-free analogues of the χ 2 and Student’s t random variables for the statistical functional context, and hence apply in a wide variety of settings including quantile estimation, gradient estimation, M-estimation, CVAR-estimation, and arrival process rate estimation, apart from more traditional statistical settings. Like the method of subsampling, we use overlapping batches of time series data to estimate the underlying variance parameter; unlike subsampling and the bootstrap, however, we assume that the implied point estimator of the statistical functional obeys a central limit theorem (CLT) to help identify the weak asymptotics (called OB-x limits, x=I,II,III) of batched Studentized statistics. The OB-x limits, certain functionals of the Wiener process parameterized by the size of the batches and the extent of their overlap, form the essential machinery for characterizing dependence, and consequently the correctness of the proposed CIPs. The message from extensive numerical experimentation is that in settings where a functional CLT on the point estimator is in effect, using large overlapping batches alongside OB-x critical values yields confidence intervals that are often of significantly higher quality than those obtained from more generic methods like subsampling or the bootstrap. We illustrate using examples from CVaR estimation, ARMA parameter estimation, and NHPP rate estimation; R and MATLAB code for OB-x critical values is available at web.ics.purdue.edu/ ∼ pasupath .

Divergent tracks: Korean Government Railways’ employment and training systems under Japanese colonial rule, 1910–45

Abstract This study employs annual reports, time-series statistics, and internal training records of the colonial-era Korean Government Railways (KGR) to conduct a quantitative analysis of its labour management practices. It addresses the colonial characteristics associated with Japanese techno-imperialism beyond ethnic discrimination, revealing a dual-pronged labour strategy that adopted a Japanese government employee system to manage middle- and upper-level personnel and directly recruited on-site workers for the lower echelons. This deviates from the low rates of local employment in Western colonies, particularly self-governing British territories or integrated French territories. In contrast, KGR's employment practices demonstrated economic and ethnic inequalities. It predominantly made Koreans on-site labourers, whereas Japanese not only held similar roles, but also occupied upper- and middle-management positions. Worker mobility, particularly among Japanese employees, grew following the outbreak of war between Japan and the USA, leading to the mass external recruitment of Koreans and the expansion of internal education to alleviate labour shortages. Nevertheless, preferential treatment toward Japanese individuals, which had been in relative decline for promotion, wages, and admission rates to training schools, ultimately persisted. Understanding KGR's employment structure shows how colonisation mediated Korea's modernisation and imposed technical limitations on the management of local labour. National technological decolonisation thus required Koreans to further introduce external technologies after the end of Japan's imperial reign.

Diel and tidal cycles in summer habitat use by St. Lawrence Estuary Belugas through a pluriannual passive acoustics monitoring network

A network of 10 passive acoustics monitoring stations is used to examine patterns of habitat use at diel and tidal timescales by St. Lawrence Estuary beluga during summer 2018 and 2019. An occurrence index of vocal activity within the preferred frequency band of communications for belugas is used as a proxy for presence at the stations. Diel and tidal patterns of activity and mean residency time are extracted from statistics of hourly occurrence timeseries at the 10 stations. Spatially, diel and tidal occurrence levels of beluga communication sounds show patterns of variation that differ among the stations, but tend to be locally stable from one summer to next. Mean residency time at the 9 Estuary stations vary between 4 to 15 hours and most have an occurrence maximum during early morning. The Saguenay fjord station shows a distinct profile, with a mean residency time of 30 hours and high level of activity at evening and night. This work underlines the ability of passive acoustics, through continuous monitoring at high spatio-temporal resolution, to reveal the complexity of the habitat use by this confined marine mammal population and understand its responses to diel and tidal forcings of the ecosystem.

ANALYSIS OF COINTEGRATION AND CAUSALITY BETWEEN INDICATORS OF ECONOMIC GROWTH AND ENERGY EFFICIENCY OF EUROPEAN COUNTRIES

In modern conditions, when European countries have set themselves an extremely ambitious goal of reducing greenhouse gas emissions by at least 55% by 2030 compared to 1990 levels, it is important to analyze the cause-and-effect relationships between key indicators of energy efficiency of national economies and economic growth, the nature of their influence on each other. The article analyzes cointegration and causal relationships between panel data that determine the economic development and energy efficiency of 38 European countries for the period from 1995 to 2021. Stationary time series were analyzed for causality using the Dumitrescu Hurlin test, which, compared to the classical Granger test, more accurately takes into account the structure of panel data, namely cross-sectional relationships. The annual GDP growth rate has driven the intensity of CO2 emissions. For pairs of time series with the first level of integration, in the case of cointegration between them, a Vector Error Correction Model (VECM) was used to determine the type of long-term behaviour of the variables with their short-term feedback. Long-term causality was found from GDP per capita to the level of primary energy intensity of European countries. Exports of goods and services have proven to be a long-term cause of domestic consumption of natural gas and solid fossil fuels. Bidirectional long-term causality was found only between primary energy consumption and exports. It should be noted that in all short-term and long-term cause-and-effect relationships obtained in the article, economic development indicators are the cause for energy efficiency indicators. This signals that the level of energy efficiency of the European economy is determined to a large extent by the economic development of Europe in previous periods. ARDL models can be used to analyze causal relationships between time series that have different levels of integration.

Enhanced Seamless Indoor–Outdoor Tracking Using Time Series of GNSS Positioning Errors

The seamless integration of indoor and outdoor positioning has gained considerable attention due to its practical implications in various fields. This paper presents an innovative approach aimed at detecting and delineating outdoor, indoor, and transition areas using a time series analysis of Global Navigation Satellite System (GNSS) error statistics. By leveraging this contextual understanding, the decision-making process between GNSS-based and Visual-Inertial Odometry (VIO) for trajectory estimation is refined, enabling a more robust and accurate positioning. The methodology involves three key steps: proposing the division of our context environment into a set of areas (indoor, outdoor, and transition), exploring two methodologies for the classification of space based on a time series of GNSS error statistics, and refining the trajectory estimation strategy based on contextual knowledge. Real data across diverse scenarios validate the approach, yielding trajectory estimations with accuracy consistently below 10 m.