Time Series Methods Research Articles

Spillover effects of climate policy uncertainty on green innovation.

Green innovation provides a viable alternative for mitigating climate change. However, it is vulnerable to uncertainty shocks. This paper investigates the influence of climate policy uncertainty (CPU) on green innovation in 10 economies from 1997 to 2022 using time series methods (GVAR, VECM, and cointegrating relationships) along with four additional measures of uncertainty: economic, energy, financial, and geopolitical. Shock identification employs the dominant unit model of the GVAR to construct a structural shock, promoting a chain reaction that affects green innovation across economies. The results indicate that CPU shocks decrease green innovation across all economies, elucidating spillover effects. Additionally, CPU shocks increase economic policy uncertainty (EPU) and geopolitical risk (GPR). Decomposition of green innovation reveals that GPR is the principal factor, followed by EPU and CPU. The analysis shows heterogeneities: EPU is the main variable affecting green innovation in Japan and the U.S., while energy-related uncertainty (EUI) is significant in South Africa. The disaggregated analysis further breaks down green patents into four categories, with estimates indicating that the most sensitive patents to CPU shocks are those related to climate change mitigation and environmental management. Additional checks, including time-varying bilateral trade and financial integration, reinforce these findings. Higher uncertainty undermines green innovation, compromising the goal of mitigating climate change.

P-1629. Reduction in Clostridioides difficile Over-diagnosis Following Suppression of C. difficile Target from Multiplex Molecular Stool Assay in a Pediatric Healthcare System

Abstract Background Clostridioides difficile testing by polymerase chain reaction (PCR) does not differentiate asymptomatic colonization from infection, and children are frequently colonized with toxigenic strains of C. difficile. Over-diagnosis of C. difficile infections can lead to adverse patient outcomes and poorer quality metrics for hospitals.Figure 1:C. difficile PCR orders following change in GI PCR Panel, Interrupted Time Series Methods In January 2023 our pediatric health system enacted multimodal changes to the electronic ordering of stool tests for C. difficile and other gastrointestinal pathogens. A multiplex PCR test for stool was updated to suppress testing for toxigenic C. difficile. The only remaining method to test for C. difficile by PCR in house was a C. difficile toxin gene specific PCR. Additionally, appropriate clinical criteria with a hard stop were added to the C. difficile PCR order (all of: no laxatives in the past 24 hours, at least 3 liquid stools in the past 24 hours, at least one risk factor for C. difficile infection, and patient at least one year old). If providers still sought C. difficile testing without meeting these criteria, infectious disease specialist approval was required. We analyzed the effect of this intervention using interrupted time series analysis and pre-post aggregate rates from January 2019 through February 2024.Figure 2:C. difficile positive rate following change in GI PCR Panel, Interrupted Time Series Results There were 4426 stool PCR tests performed for C. difficile in the pre-intervention period and 622 in the post-intervention period. The mean (± sd) age of those tested was 7.8 years (± 6.1 years). After the intervention, there was an immediate 61% decrease in C. difficile testing incidence (95% CI [53% - 67%], Figure 1) and 51% decrease in C. difficile diagnosis incidence (95% CI [26% - 67%], Figure 2). There was no significant immediate change or change in slope in the C. difficile antibiotic treatment incidence or HFO C. difficile incidence, but there was a trend towards lower rates of healthcare facility onset C. difficile after the change (aggregate rate 4.13 per 10,000 non-NICU patient-days prior to intervention vs. 3.39 after, P = .28). Conclusion Removing C. difficile from multiplex PCR assays and adding required clinical criteria to the electronic order reduces the number of C. difficile tests and diagnoses in a pediatric health system. Disclosures Laura Filkins, PhD, Avsana Labs: Board Member|Avsana Labs: Stocks/Bonds (Private Company)|Biofire Diagnostics: Grant/Research Support

Combining InSAR and Time-Series Clustering to Reveal Deformation Patterns of the Heifangtai Loess Terrace

The Heifangtai Loess terrace in northwest China is frequently affected by landslides due to hydrological factors, establishing it as a significant research area for loess landslides. Advanced time-series InSAR technology facilitates the retrieval of surface deformation information, thereby aiding in the monitoring of landslide deformation status. However, existing methods that analyze deformation patterns do not fully exploit the displacement time series derived from InSAR, which hampers the exploration of potentially coexisting deformation patterns within the area. This study integrates InSAR with time-series clustering methods to reveal the surface deformation patterns and their spatial distribution characteristics in Heifangtai. Initially, utilizing the Sentinel-1 ascending and descending SAR data stack from January 2020 to June 2023, we optimize the interferometric phase based on distributed scatterer characteristics to reduce noise levels and obtain higher spatial density of measurement points. Subsequently, by combining the differential interferometric datasets from both ascending and descending orbits, the multidimensional small baseline subsets technique is employed to calculate the two-dimensional deformation time series. Finally, time-series clustering methods are utilized to extract the deformation patterns present and their spatial distribution from all measurement point time series. The results indicate that the deformation of the Heifangtai is primarily distributed around the surrounding area of the platform, with subsidence deformation being more intense than horizontal deformation. The entire terrace exhibits five deformation patterns: eastward subsidence, westward subsidence, vertical subsidence, westward movement, and eastward movement. The spatial distribution of these patterns suggests that the areas beneath the platform, namely Yanguoxia Town and Dangchuan Village, may be more susceptible to landslide threats in the future. Furthermore, wavelet analysis reveals the response relationship between rainfall and various deformation patterns, further enhancing the interpretability of these patterns. These findings hold significant implications for subsequent landslide monitoring, early warning, and risk prevention.

Slicing-free supervised dimension reduction for multivariate time series

Sufficient dimension reduction (SDR) methods have been extensively studied for regression models with independent data, but options for time series are limited, focusing mainly on scalar responses with TSIR, TSAVE, and TSSH. Although valuable, these SDR methods rely on the slice approach. Extending them to multivariate responses via marginal slicing leads to numerous slices. Furthermore, the slice approach also poses two main questions: how many slices should be chosen and how to divide all samples into different slices. To overcome these, we introduce TMDDM, a slicing-free SDR method for time series, using approximate joint diagonalization of supervised lagged martingale difference divergence matrices (MDDM) to account for the data temporal characteristics. We also discuss lag selection strategies and the dimensionality of dimension reduction space. Simulations and real data analysis demonstrate its favourable finite-sample performance.

DFCNformer: A Transformer Framework for Non-Stationary Time-Series Forecasting Based on De-Stationary Fourier and Coefficient Network

Time-series data are widely applied in real-world scenarios, but the non-stationary nature of their statistical properties and joint distributions over time poses challenges for existing forecasting models. To tackle this challenge, this paper introduces a forecasting model called DFCNformer (De-stationary Fourier and Coefficient Network Transformer), designed to mitigate accuracy degradation caused by non-stationarity in time-series data. The model initially employs a stabilization strategy to unify the statistical characteristics of the input time series, restoring their original features at the output to enhance predictability. Then, a time-series decomposition method splits the data into seasonal and trend components. For the seasonal component, a Transformer-based encoder–decoder architecture with De-stationary Fourier Attention (DSF Attention) captures temporal features, using differentiable attention weights to restore non-stationary information. For the trend component, a multilayer perceptron (MLP) is used for prediction, enhanced by a Dual Coefficient Network (Dual-CONET) that mitigates distributional shifts through learnable distribution coefficients. Ultimately, the forecasts of the seasonal and trend components are combined to generate the overall prediction. Experimental findings reveal that when the proposed model is tested on six public datasets, in comparison with five classic models it reduces the MSE by an average of 9.67%, with a maximum improvement of 40.23%.

Parametrical Identity Mapping: An Evaluation Procedure for THW Signals and Other Time-Series Methods

Parametrical Identity Mapping: An Evaluation Procedure for THW Signals and Other Time-Series Methods

A Time Series Method with Physically Guided Selection of Surface Indicators for Passive Microwave Brightness Temperature Swath Gap-Filling

A Time Series Method with Physically Guided Selection of Surface Indicators for Passive Microwave Brightness Temperature Swath Gap-Filling

Time series method for forecasting model for amount of ginger plant production

Time series method for forecasting model for amount of ginger plant production

Forecasting solar power generation with machine learning techniques

This study underscores the economic and environmental advantages of integrating solar energy into power systems. The unpredictable nature of solar power poses challenges to system operation and planning. To ensure the economic sustainability of newly constructed systems, precise forecasting of Photovoltaic (PV) system effectiveness and energy output is crucial. Addressing variations in solar power consumption, this work presents an enhanced Machine Learning (ML) model. Utilizing Python, the study explores Linear Regressor, Random Forest (RF) Regressor, XGBoost Regression, K-Nearest Neighbor (KNN) Regressor, and AdaBoost Regressor approaches, all proving effective in predicting electricity production. Results highlight the superior performance of ML algorithms over traditional time series methods and two baseline models, emphasizing their efficacy in solar power forecasting.

Time-series InSAR measurement using ICOPS and estimation of along-track surface deformation using MAI during the 2021 eruption of Fagradalsfjall Volcano, Iceland

The eruption in Fagradalsfjall Volcano, located in Reykjanes Peninsula, Iceland, from several centuries’ dormant states, occurred for the first time on March 19, 2021. Observations of Fagradalsfjall Volcano were conducted in 2021, and the eruption period lasted for six months until 18 September 2021. Six days pair of interferograms were generated from ninety synthetic aperture radar (SAR) data. Thus, the SAR data will be acquired from the Sentinel-1 satellite from January until December 2021. Time-series measurements were conducted using a combination of persistent scatterer (PS) and distributed scatterer (DS) points to produce denser measurement points (MPs) in the study area. The improved combined scatterers interferometry with optimized point scatterers (ICOPS) algorithm is the time-series method that utilizes both PS and DS MPs and optimizes those combined MPs using a deep learning algorithm over different temporal intervals and using a statistical clustering approach to optimize the MPs spatially. Validation was conducted by comparing the ICOPS result with GPS measurement in Reykjavik. The comparison with the GPS measurement was performed to validate the line-of-sight (LOS) deformation from the ICOPS measurement, which resulted in an RMSE value of about 0.58 cm, which is considered a good correlation. Besides the time-series Interferometry SAR (InSAR) measurement, we used the integrated InSAR and multiple aperture interferometry (MAI) methods to estimate both LOS and along-track surface deformation, respectively, during the Fagradalsfjall, Iceland volcanic eruption. A pair of ALOS-2 data was used between 28 February 2021 and 23 May 2021. The result from the MAI method shows a deformation of approximately ± 2 mm in the azimuth direction around Fagradalsfjall Volcano. The deformation around Fagradalsfjall Volcano was suggested to be due to the activity of the magma reservoir beneath the Earth’s surface, which was formed by dike intrusion. The analysis of the seismicity in Fagradalsfjall was discussed by visualization of the distribution of earthquakes during the deformation occurrence. Further analysis can be conducted by applying multitrack analysis to find the 3D deformation pattern due to the eruption.

Effects of missing data imputation methods on univariate blood pressure time series data analysis and forecasting with ARIMA and LSTM

BackgroundMissing observations within the univariate time series are common in real-life and cause analytical problems in the flow of the analysis. Imputation of missing values is an inevitable step in every incomplete univariate time series. Most of the existing studies focus on comparing the distributions of imputed data. There is a gap of knowledge on how different imputation methods for univariate time series affect the forecasting performance of time series models. We evaluated the prediction performance of autoregressive integrated moving average (ARIMA) and long short-term memory (LSTM) network models on imputed time series data using ten different imputation techniques.MethodsMissing values were generated under missing completely at random (MCAR) mechanism at 10%, 15%, 25%, and 35% rates of missingness using complete data of 24-h ambulatory diastolic blood pressure readings. The performance of the mean, Kalman filtering, linear, spline, and Stineman interpolations, exponentially weighted moving average (EWMA), simple moving average (SMA), k-nearest neighborhood (KNN), and last-observation-carried-forward (LOCF) imputation techniques on the time series structure and the prediction performance of the LSTM and ARIMA models were compared on imputed and original data.ResultsAll imputation techniques either increased or decreased the data autocorrelation and with this affected the forecasting performance of the ARIMA and LSTM algorithms. The best imputation technique did not guarantee better predictions obtained on the imputed data. The mean imputation, LOCF, KNN, Stineman, and cubic spline interpolations methods performed better for a small rate of missingness. Interpolation with EWMA and Kalman filtering yielded consistent performances across all scenarios of missingness. Disregarding the imputation methods, the LSTM resulted with a slightly better predictive accuracy among the best performing ARIMA and LSTM models; otherwise, the results varied. In our small sample, ARIMA tended to perform better on data with higher autocorrelation.ConclusionsWe recommend to the researchers that they consider Kalman smoothing techniques, interpolation techniques (linear, spline, and Stineman), moving average techniques (SMA and EWMA) for imputing univariate time series data as they perform well on both data distribution and forecasting with ARIMA and LSTM models. The LSTM slightly outperforms ARIMA models, however, for small samples, ARIMA is simpler and faster to execute.

A methodology based on the fuzzy identification for a hybrid method in times series forecasting models

This work presents a hybrid forecasting method, with concept and improvement through the fuzzy set’s theory coupled with the classic time series method. This study also proposed an identification tool that has not been published in the scientific literature for parametric point identification and estimation within the classical time series methodology combined with fuzzy sets theory, which generated a hybrid identification technique in fuzzy time data. In comparison to the forecasting methods consolidated in the literature, the results presented by the proposed method have been considered fully satisfactory due to the minimal error observed.

Time-Series InSAR Technology for Monitoring and Analyzing Surface Deformations in Mining Areas Affected by Fault Disturbances

Faults, as unique geological structures, disrupt the mechanical connections between rock masses. During coal mining, faults in the overlying strata can disturb the original stress balance, leading to fault activation and altering the typical subsidence patterns. This can result in abnormal ground deformation and significant damage to surface structures, posing a serious geological hazard in mining areas. This study examines the influence of a known fault (F13 fault) on ground subsidence in the Wannian Mine of the Fengfeng Mining Area. We utilized 12 Sentinel-1A images and applied SBAS-InSAR, StaMPS-InSAR, and DS-InSAR time-series InSAR methods, alongside the D-InSAR method, to investigate surface deformations caused by the F13 fault. The monitoring accuracy of these methods was evaluated using leveling measurements from 28 surface movement observation stations. In addition, the density of effective monitoring points and the relative strengths and limitations of the three time-series methods were compared. The findings indicate that, in low deformation areas, DS-InSAR has a monitoring accuracy of 7.7 mm, StaMPS-InSAR has a monitoring accuracy of 16.4 mm, and SBAS-InSAR has an accuracy of 19.3 mm.

Comparison of Feature Extraction Methods in High Dimensional Time Series

Working with high-dimensional datasets increases the workload on machine learning models. Therefore, before making predictions, the most meaningful data points in the entire data set must be determined. It is highly important to improve model performance, especially in the field of machine learning. For this reason, five feature selection methods—Mutual Information, Principal Component Analysis, Chi-square, Information Gain, and Variance Thresholding—commonly used in the literature, were tested on the 14400 feature data set obtained with a system previously proposed to determine the sand, silt and clay ratios in the soil. The success of these five methods is presented comparatively using R-square (R²) and Mean Absolute Error (MAE) metrics. The best results were obtained with the Information Gain method for sand (R2 = 0.44), with Chi-square for silt (R2 = 0.17), and with Variance Thresholding for clay (R2 = 0.61).

Technological and Innovative Transformation in Marketing: Examination of Social Media Text Stream Visualization Concept with Bibliometric Analysis: Future Agenda

Changes in technology and innovation processes affect many areas. The impact of technology and innovative applications lead to radical changes in marketing strategies. Social media text stream visualization stands out as one of the most important tools of these changes. Social media text stream visualization allows the analysis of text data coming from social media platforms through graphics and infographics. In this way, social media text stream visualization allows the development of more effective and efficient strategies by combining technology and innovation in marketing. In this study, the Web of Science database is used to reveal in which areas the concept of social media text stream visualization has been used in the last 10 years through bibliometric analysis. After the inclusion and exclusion criteria are applied, the final sample consists of 12 refereed articles. Five separate analyses were conducted to test the sample. Performance analysis determined the publication years of the articles, their contributions by country, the outputs of relevant journals, and the sectors where this concept is used most intensively. The findings of the study provide a broad perspective on the research conducted to date and identify potential research gaps. Accordingly, it is seen that the studies within the scope of social media text stream visualization are mostly conducted in the fields of social media, advertising, ecommerce and market research. It was concluded that machine learning, big data, text analysis and time series methods were used in the studies. It is recommended that studies be conducted in the retail, tourism and accommodation and technology sectors for future research. This research aims to contribute to the field of marketing by revealing the gaps in the literature. The study also offers suggestions for future research to be conducted by academics and sectoral researchers.

Future forecast of global mean surface temperature using machine learning and conventional time series methods

Future forecast of global mean surface temperature using machine learning and conventional time series methods

GCINet: global convolution interaction network with a pre-trained reversible normalization method for long-term time series forecasting

GCINet: global convolution interaction network with a pre-trained reversible normalization method for long-term time series forecasting

Limits on groundwater-surface water transitions got from temperature time series: characterizing goal-based edges

ABSTRACT This study addresses the challenge of accurately estimating groundwater-surface water fluxes, essential for sustainable water resource management, by improving traditional temperature time series methods. Existing approaches often struggle with distinguishing natural temperature variations from significant water exchanges, especially across different spatial and temporal scales under varying conditions like recharge and discharge. To overcome these issues, advanced statistical tools were applied to temperature data, and COMSOL Multiphysics was used to simulate interactions by coupling fluid flow and heat transport. Multisite and multivariate calibration techniques refined parameters like hydraulic conductivity and porosity, while models accounted for external factors such as climate variability and land-use changes. The findings reveal that the combined method was effective in estimating groundwater-surface water fluxes under recharge conditions. The COMSOL model achieved an MAE of 0.71, MAP has 0.92, RMSE of 0.80, and RMSLE has 0.30, indicating consistent performance across evaluation metrics. This advancement improves groundwater-surface water model precision, especially in distinguishing natural temperature variations from significant water exchanges. It offers more accurate flux estimation, aiding better water resource management.

COVID-19 Trends in a Northeastern Brazilian State from the Start of the Pandemic: Exploring an Alternative Time Series Method

Pernambuco, a northeastern Brazilian state, suffered significantly during the COVID-19 pandemic. The incidence, mortality, and fatality rates were incredibly high; these outcomes were a direct failure of the public health system to manage the COVID-19 pandemic. We developed an efficient method to analyze the trends in incidence, mortality, and lethality indicators of COVID-19 and the dynamics of their main determinants in the state of Pernambuco from March 2020 to March 2022. This was a time series ecological study. We calculated the lethality, mortality, and incidence rates with official and public data from the Health Department of this state. The effective reproduction number (Rt) was estimated, and then the periods were delimited, thus creating the efficiency model. Differences were considered significant when p < 0.05. We found 585,551 cases and 18,233 deaths from COVID-19. January 2022 was the month with the highest number of cases (6312) and, in May 2020, we recorded the highest number of deaths (149). Our method was efficient in analyzing periods of COVID-19, noting a contrast between the early and most recent periods, with the latter showing a stable number of new cases and patient deaths.

Differential Analysis Based on Different Decomposition Methods of Time Series

A detailed analysis of four popular time series decomposition methods: Classical, X11, SEATS, and STL is provided in this study, with a focus on their applications, strengths, and limitations. Understanding seasonal patterns in time series data is an important aspect in several fields such as economics or environmental science. The R package seasonal was employed to conduct the decompositions on the official data of National Oceanic and Atmospheric Administration (NOAA) Climate Data from January 1980 till July 2024. It offers a detailed analysis of the performance of each method in isolating trends, seasonal components, and residuals. Findings reveal that although all methods can be used for time series data decomposition, their success is influenced by data nature and particular analytical tasks. For instance, SEATS is robust enough to handle irregularities while Classical decomposition wins when it comes to simple understanding. Similarly, X11 methods were known for lack of flexibility as well as recurring seasonal changes sensitivity, where STL is particularly efficient for varying seasonal patterns datasets with nonlinearity trends. From the findings, it can be concluded that it is important to choose the appropriate decomposition method to increase reliability and make appropriate adjustments in the forecasts incorporating time series data.