Dynamic time warping to model daily life stress reactivity in a clinical and non-clinical sample - An ecological momentary assessment study.

Quantifying uncertainty in stratigraphic alignment of geological signals using probabilistic dynamic time warping

Measuring the similarity of time series is a fundamental task in numerous information processing applications. Dynamic Time Warping (DTW) is a widely used method for time series similarity measurement, yet its reliance solely on linear Euclidean distance and neglect of directional information often limits its ability in scenarios where subtle variations and trajectory orientation carry semantic significance. To address these limitations, we propose Angle-distance Penalized Metric DTW (APMDTW), a novel similarity measure method that integrates a nonlinear spatial distance metric with an adaptive angle-distance penalty. Specifically, a piecewise logarithmic transformation is introduced to enhance sensitivity to fine-grained local differences, while a parameterized angle-distance penalty, adaptively modulated by spatial distance, incorporates directional consistency into the cost function. This joint modeling of spatial magnitude and geometric orientation yields a more discriminative and robust time series similarity measurement. Experiments on 128 UCR benchmark datasets show that APMDTW outperforms six baseline algorithms on 114 datasets, and improves similarity measurement accuracy by an average of 56.52% over six state-of-the-art DTW variants.

Cortisol and testosterone are key digital biomarkers reflecting neuroendocrine activity across the hypothalamic–pituitary–adrenal (HPA) and hypothalamic–pituitary–gonadal (HPG) axes, encoding stress adaptation and behavioral regulation. Continuous real-world monitoring remains challenging due to the sparsity of sensing and the complexity of multimodal data. This study introduces a synthetic sensor-driven computational framework that models hormone variability through data-driven simulation and predictive learning, eliminating the need for continuous biosensor input. A hybrid deep ensemble integrates biological, behavioral, and contextual data using bidirectional multitask learning with one-dimensional convolutional neural network (1D-CNN) and long short-term memory (LSTM) branches, meta-gated expert fusion, Bayesian variational layers with Monte Carlo Dropout, and adversarial debiasing. Synthetically derived longitudinal hormone profiles that were validated by Kolmogorov–Smirnov (KS), Wasserstein, maximum mean discrepancy (MMD), and dynamic time warping (DTW) metrics account for class imbalance and temporal sparsity. Our framework achieved up to 99.99% macro F1-score on augmented samples and more than 97% for unseen data with ECE below 0.001. Selective prediction further maximized the convergence of predictions for low-confidence cases, achieving 99.9992–99.9998% accuracy on 99.5% of samples, which were smaller than 5 MB in size so that they can be employed in real time when mounted on wearable devices. Explainability investigations revealed the most important features on both the physiological and behavioral levels, demonstrating framework capabilities for adaptive clinical or organizational stress monitoring.

This study investigates three types of emission families in Hidden Markov Models (HMMs) for reconstructing Bedoyo Majapahit dance motion captured using a markerless system. The recorded skeleton data, consisting of 3,341 frames and 33 joints per frame, were normalized and reduced into a 30-dimensional latent space using Principal Component Analysis (PCA). Three emission variants were evaluated: single-Gaussian HMM, Gaussian-mixture HMM (GMM-HMM), and Multinomial HMM. The evaluation employed a tri-metric scheme consisting of Mean Squared Error (MSE), Dynamic Time Warping (DTW), and Fréchet distance to measure reconstruction fidelity. The experimental results showed that GMM-HMM consistently outperformed the other two models, achieving the lowest reconstruction error and the closest alignment to the original temporal and geometric motion profiles. The Gaussian HMM demonstrated moderate performance but tended to underestimate motion amplitude, while the Multinomial HMM produced the weakest results due to the discretization of continuous pose data. These findings indicate that multimodal emission functions provide a more expressive representation for continuous dance motion. The study highlights the suitability of GMM-HMM for traditional dance preservation through computational modeling and contributes to the development of digital motion archiving for cultural heritage.

A spatial-temporal fusion based nonlinear causality analysis framework for root cause diagnosis of faults in nonstationary industrial processes with asymmetric distribution.

BackgroundAcute pancreatitis (AP) morbidity has been increasing in recent years. Patients with AP exhibit highly variable symptom patterns over time, posting challenges to traditional analytical methods. Dynamic Time Warping (DTW) effectively aligns temporal sequences of different rhythms, offering a novel approach to model these complex dynamics.ObjectiveThis study employs DTW technology to systematically analyze the individualized developmental trajectories of symptom clusters in patients with AP, delving into the heterogeneous characteristics during the process of time series changes.MethodsIn a longitudinal study of 155 patients with AP, 32 symptoms were assessed using the Memorial Symptom Assessment Scale at hospitalization and 1, 3, 6, 9, and 12 months post-discharge. DTW was used to analyze temporal dynamics, generating individual symptom distance matrices. At the group level, these matrices are integrated using Distatis analysis, followed by hierarchical clustering to identify symptom clusters and network analysis to determine central symptoms.ResultsEach patient had unique symptom manifestations and dynamic change patterns. Six major symptom clusters were identified: emotional disorder cluster, appetite disorder cluster, multi-system physical discomfort cluster, localized physiological perception abnormality cluster, functional decline cluster, and abdominal discomfort cluster. Centrality analysis revealed that the appetite domain exhibited high centrality, suggesting that its variation may influence multiple aspects of patient experience.ConclusionDynamic Time Warping provides a novel and effective approach for analyzing the temporal trajectories of symptoms both within and across individuals. The research results provide methodological support and empirical evidence for individualized symptom management, early intervention, and predictive model construction of AP progression.

BackgroundA high prevalence of dry eye disease (DED) has intensified public health concerns in Taiwan. With the growing reliance on online resources for health information, platforms such as Google Trends (GT) provide a valuable method for capturing public interest. This approach also allows for the exploration of potential associations between public interest in DED and environmental parameters, which may further elucidate underlying factors contributing to the disease’s rising prevalence.ObjectiveThis study aims to (1) analyze public interest in DED in Taiwan using GT data, (2) investigate correlations between search interest and environmental parameters, and (3) identify shifts in the focus of search over time.MethodsWe analyzed GT data from December 2018 to July 2024, focusing on relative search volume (RSV) for DED across Taiwan and its 6 special municipalities. Temporal trends in RSV were assessed using spline regression models, and monthly variations were assessed using the Kruskal-Wallis test. The Spearman correlation analysis was used to evaluate the association between RSV and environmental parameters, while dynamic time warping analysis clarified the temporal alignment of RSV with these parameters. Rising search queries were analyzed to identify shifts in public interest over time. Furthermore, top Google search results for DED-related keywords were assessed for topic coverage, quality, and readability.ResultsA significant rising trend in RSV for DED was observed over the study period in Taiwan (mean instantaneous derivative=0.445; P<.001) and across all 6 special municipalities. Environmental parameters such as methane (CH4), total hydrocarbons, and nonmethane hydrocarbons were identified as novel pollutants strongly correlated with RSV (P<.001), along with known pollutants such as nitric oxide (NO), nitrogen dioxide (NO2), sulfur dioxide (SO2), nitrogen oxides (NOx), and carbon monoxide (CO). Dynamic time warping analysis revealed the strongest temporal alignment was between RSV and hydrocarbons, including CH4 and total hydrocarbons, further emphasizing their potential role in influencing public interest. Assessment of web-based DED information of 80 websites revealed generally low quality (DISCERN score: mean 2.14, SD 0.40), and the average readability corresponded to a college reading level (grade: mean 21.1, SD 4.5). Rising search queries shifted from diagnostic and treatment methods before the COVID-19 pandemic to natural remedies during the COVID-19 lockdown and self-diagnosis and treatment options after the pandemic. Gaps were also identified between public interest and the availability of online information.ConclusionsPublic interest in DED has increased significantly in Taiwan from 2018 to 2024, with hydrocarbons identified as strongly associated environmental parameters. The shifts in related queries reflect changing public interest, accentuating the need for health care information that aligns with public interest and addresses gaps in available resources.

To enhance the objectivity and precision of quality evaluation in flight training, this study proposes an assessment method for the landing flare phase based on time-series flight parameter data from Secure Digital (SD) card. By analyzing landing flare data from flight instructors and trainees, a standard sequence model was established, and the Dynamic Time Warping (DTW) algorithm was employed to calculate the similarity between individual trainee sequences and the standard sequence. Using K-means clustering, the landing flare quality was categorized into four distinct levels: Excellent (22.5%), Good (25.5%), Qualified (23.5%), and Improvement Needed (28.5%). The results demonstrated significant consistency with instructor evaluations (Spearman correlation coefficient 0.71). Furthermore, through the identification of weak parameters, specific technical deficiencies in areas such as airspeed control and pitch attitude maintenance could be accurately pinpointed. This approach not only effectively validates and supplements instructor assessments but also provides data-driven support for developing personalized training programs, thereby improving the quality and efficiency of flight training.

Abstract Accurate estimation of the State of Health (SOH) for lithium-ion batteries (LIBs) is critical for ensuring system safe operation. However, the presence of capacity recovery during battery aging, coupled with experimental uncertainties, introduces fluctuations and redundancies in the features extracted from post-discharge relaxation voltage curves, which in turn adversely affect the accuracy of SOH prediction. For this reason, this paper proposes a SOH estimation framework that integrates relaxation voltage features, autoencoder-based feature enhancement, and a transformer-inspired dual-residual network. Firstly, five physically original features are extracted from the relaxation voltage curve, including open-circuit voltage (OCV), first-second relaxation voltage data (FSRVD), integral area (IA), Dynamic Time Warping (DTW) distance, and Wasserstein (WAS) distance. Next, the autoencoder is designed to compress and denoise these original features, producing three-dimensional enhanced features (AE1–AE3) that improve feature stability and representation. Then, a transformer-inspired neural network with dual residual connections is constructed to enhance model depth and training stability. Finally, SOH estimation was performed under different discharge rates and SOC conditions. Experimental results show that the proposed method achieves favorable prediction accuracy under different conditions. Compared to original features, the enhanced features significantly improve SOH estimation, yielding an average Root Mean Square Error (RMSE) reduction of 23.33%, which provides valuable insights into the development of advanced battery management systems.

Efficient eye-hand coordination is fundamental to surgical proficiency, particularly in minimally invasive surgeries. A comprehensive analysis of eye-hand coordination across varying surgical skill levels is essential for evaluating surgical task proficiency and establishing a basis for objective surgical skill assessment and training. Eighteen participants, including five experts (attending general surgeons), five intermediates (surgical residents), and eight novices without prior surgical experience, were recruited in this study. Metrics such as gaze motion entropy, gaze-tooltip relative latency, and gaze-tooltip dynamic time warping (DTW) were employed to assess the irregularity of gaze motion and coordination between gaze and surgical instrument motion. Experts had significantly lower gaze motion entropy than novices, reflecting more organized and predictable gaze behavior. Experts and intermediates also exhibited superior gaze and surgical instrument coordination than novices, as indicated by significantly lower gaze-tooltip DTW values for surgical instruments in both dominant and non-dominant hands. Furthermore, longer gaze-tooltip relative latency for the surgical instrument in the non-dominant hand among experts compared to novices suggests anticipatory gaze motion, i.e., eye-leading-hand behavior. This study developed a spatiotemporal framework for analyzing eye-hand coordination, demonstrating that higher surgical skill levels are associated with lower gaze motion entropy, greater spatial coordination between gaze and instrument motion, and longer gaze-tooltip relative latency. These findings can contribute to objective surgical skill assessment and offer valuable insights for surgical training, potentially accelerating the learning curve for trainees.

This mixed-methods study examines whether backpropagation-based deep learning (DL) visualizations can strengthen metacognition and learning outcomes in a university Linear Programming course. Sixty undergraduates (8-week blended format) completed pre/post cognitive tests and the Metacognitive Awareness Inventory (MAI), while their LMS activity traces (e.g., time-on-task, revision frequency, error types) trained a multilayer perceptron. The intervention exposed students to DL visual artifacts—loss curves, gradient/weight updates, and error heatmaps—as reflective scaffolds linking machine error correction to human self-regulation. Quantitatively, mean test scores increased from 61.23 to 80.57 (paired-t, p &lt; .001), and total MAI rose from 135.40 to 159.85 (paired-t, p &lt; .001). Gains concentrated in regulation of cognition (monitoring/evaluation). Metacognitive improvement correlated with achievement (Pearson r = .62, p &lt; .001). Computationally, model loss decreased from 0.25 to 0.03 over 200 epochs with 89.4% validation accuracy; Dynamic Time Warping = 0.81 (p &lt; .01) indicated strong temporal alignment between DL loss minimization and students’ learning curves. Qualitatively, thematic analysis of weekly reflections and interviews revealed a progression from error recognition to strategy adjustment and reflective transformation, recasting errors as actionable signals. Triangulating quantitative, computational, and qualitative strands supports the Cognitive Backpropagation Learning (CBL) framework: DL error feedback parallels human metacognitive feedback, and its visualization functions as a digital mirror that externalizes reflection. Findings recommend interpretable DL dashboards as practical, class-deployable scaffolds to cultivate reflective, adaptive mathematical thinkers.

Parkinson’s disease (PD) is a neurodegenerative disorder characterized by motor symptoms, including gait impairments, which significantly affect patient mobility and quality of life. An accurate assessment of the severity of PD is crucial for clinical management. This study investigates the utility of smartphone-derived gait data to objectively cluster PD severity using unsupervised machine learning, with the aim of improving precision in disease monitoring. We analyze gait data from the mPower dataset, comprising 8779 accelerometer recordings from 1957 participants (PD patients and healthy controls). Stride cycles were segmented using frequency analysis and peak detection, followed by sequence padding to standardize input lengths. K-means clustering with dynamic time warping (DTW) was applied to identify gait patterns, while autoencoder embeddings and t-SNE visualized high-dimensional data. The groups were correlated with the MDS-UPDRS scores (Parts I and II) to assess severity. Four distinct gait clusters were identified, correlating with the severity of PD. The most severe group (Cluster 1) exhibited significantly higher MDS-UPDRS scores for balance/walking problems (2.43times) and freezing episodes (8.41times) compared to the least severe group (Cluster 4). The visualization of t-SNE confirmed the clear separation of the clusters, with higher severity scores concentrated in cluster 1. Sequence padding showed no significant impact on clustering outcomes (p > 0.05), validating its use for handling variable-length data. This study demonstrates that smartphone-derived gait patterns, analyzed via unsupervised clustering and visualization techniques, effectively stratify PD severity. Gait features related to balance, freezing, and walking difficulties are critical biomarkers for disease progression. Key advantage of our technique is the use of unsupervised learning to identify latent patterns without preconceived group assumptions, allowing subgroups to emerge organically and providing an unbiased exploration of gait-pattern relationships with Parkinson’s severity. While limitations include the reliance on self-reported MDS-UPDRS data and k-means algorithm variability, these findings highlight the potential of wearable sensors and machine learning to develop objective, scalable tools for PD assessment.

Abstract The inherent time‐lag effects, nonlinear dependencies, and dynamic coupling mechanisms in desulphurization processes pose significant challenges to precise quality prediction and proactive control. This study presents a novel collaborative optimization framework integrating causal inference with temporal feature engineering to achieve dynamic early warning and intelligent control. Methodologically, we first develop a multi‐modal time‐lag estimation approach combining dynamic time warping, Granger causality tests, and time‐delayed mutual information, resolving temporal asynchrony between process variables through dynamic programming and information‐theoretic analysis. Building upon this, a dynamic autoregressive latent variable model (DALM) with Bayesian estimation is established to capture cross‐variable interaction dynamics, enhanced by a long short‐term memory (LSTM)‐based architecture with attention mechanisms for nonlinear temporal dependency modelling. The proposed early‐warning system synergizes anomaly detection (isolation forest/one‐class SVM/autoencoder triad) with NOTEARS‐optimized causal graphs, achieving 93.1% prediction accuracy (AUC = 0.98) through multi‐feature fusion of temporal patterns, causal drivers, and multivariate anomalies. For control optimization, formulate a hybrid MPC strategy incorporating warning‐adaptive penalty terms, demonstrating 89.2% warning probability alignment with quality deviations while the concentrations of SO 2 /H 2 S are kept within a fixed range through restricted reactor feed flow adjustment. Validations confirm the framework reduces unplanned shutdowns by 37% compared to conventional PID control, with R 2 = 0.9144 for SO 2 and 0.9114 for H 2 S concentration predictions. This work provides a systematic solution addressing temporal‐causal decoupling challenges in complex chemical processes, significantly advancing intelligent optimization in pollution control systems.

ABSTRACT Accurate real-time monitoring of the ash content in flotation clean coal is pivotal for intelligent optimization and closed-loop control of the flotation process, directly affecting product quality and the economic performance of coal preparation plants. To address the limitations of traditional approaches–namely response lag, insufficient accuracy, and inefficient fusion of multisource information–this study proposes an intelligent online sensing method based on multisource data fusion, with the prediction pipeline decoupled into three stages: alignment – representation – prediction. First, a multiscale, differentiable dynamic time-warping (MSSoftDTW) scheme is employed to precisely align asynchronous multisource time-series data, thereby enhancing cross-modal temporal consistency. Second, an interpretable Constructive algorithm with response-weight mechanism (ICA-RW) is introduced to enable feature learning and structural adaptation, suppressing redundancy and collinearity while improving feature robustness. Third, an ensemble regression model that combines a relevance vector machine with adaptive boosting (RVM-Adaboost) is developed to better accommodate nonlinear relationships and drifts in operating conditions. By fusing X-ray fluorescence (XRF) spectra, key process variables, and features extracted from tailings images, the method achieves high-accuracy, real-time prediction of clean-coal ash content. Validation on industrial-site data demonstrates significant gains in both accuracy and stability over conventional regression baselines, meeting the real-time requirements of online monitoring and control and providing deployable support for flotation process optimization and intelligent upgrading.

Digital emotional contagion shapes how emotions spread and align among participants in the online community. This phenomenon is especially evident in the real-time interactions of tourism live streaming. However, little is known about how emotional contagion unfolds during these real-time interactions. Given that, this study seeks to examine the generative mechanisms of emotional contagion in tourism live streaming. Study One used dynamic time warping and panel vector autoregression to analyze the contagion process across 105 tourism live streaming sessions. Study Two applied BERTopic modeling to identify emotionally contagious topics engaging streamers and viewers. The findings deepen the theoretical understanding of digital emotional contagion by highlighting its interactivity, multi-directionality, and synchronicity. It also reveals the complex interplay between positive and negative emotional contagion and contagious topics. These insights help digital marketers understand the direction and characteristics of emotional contagions, and effectively foster viewers’ emotional engagement for community building and loyalty.

Abstract Reckless driving behaviors, including abrupt braking, sudden acceleration, and sharp vehicular turns, significantly elevate the risk of road traffic incidents and fatalities. To address this, the present study introduces a machine learning-based framework for real-time detection, classification, and categorization of turn maneuvers using orientation data acquired from commodity smartphone sensors. The proposed system offers a low-cost, infrastructure-independent solution for analyzing driver behavior in urban mobility environments. Empirical data were collected from two riders traversing a predefined route within the urban landscape of Durgapur, India. The methodology focuses on the x -axis orientation sensor, which effectively captures lateral motion during vehicular turns. After preprocessing and feature engineering, a logistic regression model was trained to detect turning events, achieving a classification accuracy of 91%. For directional turn classification (left/right), a decision tree classifier was employed, yielding an accuracy of 75%. Furthermore, Dynamic Time Warping (DTW) was utilized to compute similarity metrics between observed and reference turn signatures, facilitating post-classification categorization into quality-of-turn tiers ranging from Very Good Ride to Very Bad Ride . The experimental outcomes validate the efficacy of integrating smartphone-based sensing with lightweight machine learning models for granular driver behavior assessment. The proposed system demonstrates potential for deployment in intelligent transportation systems (ITS) and road safety analytics, thereby contributing to proactive risk mitigation and behavioral intervention strategies.

Tomato prices in Kolar market exhibit high volatility alongside recurring seasonal patterns, but the consistency of these patterns across years remains unclear. This study analysed weekly tomato prices and arrivals from 2010–2024 to quantify inter-annual variability using descriptive statistics, seasonal indices, and Dynamic Time Warping (DTW). Descriptive analysis confirmed extreme fluctuations (CV = 77% for prices, 102% for arrivals) with positive skewness and heavy tails, indicating frequent extreme events. Seasonal indices revealed recurring intra-year cycles, but year-to-year alignment varied substantially. DTW analysis for 2021–2024 quantified pattern similarity, showing that 2022–2023 had the highest alignment (DTW distance: 23,258) despite extreme price spikes (max: ₹7,429), whereas 2021–2022 exhibited the weakest alignment (distance: 39,049), reflecting structural shifts in market dynamics. Path length metrics indicated minimal temporal warping in 2022–2023 (71 points) versus extensive alignment in 2021–2022 (83 points). These results demonstrate that while seasonal patterns recur, their temporal consistency is not fixed, highlighting the need for forecasting models that adapt to both magnitude volatility and temporal shifts. The study also illustrates the utility of DTW for agricultural price analysis and the limitations of relying solely on fixed seasonal patterns in volatile commodity markets.

ABSTRACT The Global Navigation Satellite System-Interferometric Reflectometry (GNSS-IR) technique has strong potential for soil moisture (SM) monitoring but often suffers from phase jumps and unstable retrievals. This study proposes a comprehensive approach to address these limitations. First, a PID-based search algorithm (PSA) is used to obtain robust initial values for nonlinear least squares (NLS) phase fitting, thereby suppressing fitting-induced phase jumps. Second, Dynamic Time Warping (DTW) satellite selection combined with amplitude weighted multi-band fusion is applied to identify and down-weight or exclude inconsistent satellite contributions and to stabilize the fused SM estimates. Multi-frequency observations from two Plate Boundary Observatory (PBO) stations, P041 and P565 demonstrate that this comprehensive approach markedly improves phase – SM correlation and reduces retrieval errors, providing a practical solution for GNSS-IR SM monitoring and related applications.

In response to the persistent failures of traditional election polling, this study introduces the Decentralized Prediction Market Voter Framework (DPMVF), a novel tool to empirically test and quantify the predictive capabilities of Decentralized Prediction Markets (DPMs). We apply the DPMVF to Polymarket, analysing over 11 million on-chain transactions from 1 September to 5 November 2024 against aggregated polling in the 2024 U.S. Presidential Election across seven key swing states. By employing Cross-Correlation Function (CCF) for linear analysis and Dynamic Time Warping (DTW) for non-linear pattern similarity, the framework provides a robust, multi-faceted measure of the lead-lag relationship between market sentiment and public opinion. Results reveal a striking divergence in predictive clarity across different electoral contexts. In highly contested states like Arizona, Nevada, and Pennsylvania, the DPMVF identified statistically significant early signals. Using a non-parametric Permutation Test to validate the observed alignments, we found that Polymarket’s price trends preceded polling shifts by up to 14 days, a finding confirmed as non-spurious with a high confidence (p &lt; 0.01) and with an exceptionally high correlation (up to 0.988) and shape similarity. At the same time, in states with low polling volatility like North Carolina, the framework correctly diagnosed a weak signal, identifying a “low-signal environment” where the market had no significant polling trend to predict. This study’s primary contribution is a validated, descriptive tool for contextualizing DPM signals. The DPMVF moves beyond a simple “pass/fail” verdict on prediction markets, offering a systematic approach to differentiate between genuine early signals and market noise. It provides a foundational tool for researchers, journalists, and campaigns to understand not only if DPMs are predictive but when and why, thereby offering a more nuanced and reliable path forward in the future of election analysis.