Pith. sign in

REVIEW 4 major objections 5 minor 67 references

Machine Learning-based detection of long COVID using Heart Rate Variability Analysis

T0 review · 4 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that machine-learning classifiers on 16 heart-rate-variability features can distinguish long COVID patients from recovered COVID controls, with Gradient Boost reaching 85.2% accuracy and AUC 0.907.

desk verdict A careful HRV data collection effort whose headline classifier numbers are not trustworthy because feature selection and hyperparameter tuning used the full dataset before cross-validation; the underlying association is plausible but not established by these results. read the letter →

arxiv 2607.13668 v1 pith:3GYIEZKC submitted 2026-07-15 stat.AP

Brais Iglesias-Otero , Xosá A. Vila Sobrino , María J. Lado , Leandro Rodriguez-Liñares , Baltasar García Pérez-Schofield , Pedro Cuesta Morales , Arturo J. Méndez , María Bustillo Casado
show 1 more author
Alexandre García-Caballero
This is my paper
classification stat.AP
keywords longCOVIDheartratevariabilitymachinelearningGradientBoostautonomicnervoussystemsix-minutewalktestcoldpressureclassification
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that long COVID leaves a measurable trace in heart rate variability, and that machine-learning models can pick it up. Using recordings from 129 subjects — 63 long COVID patients and 66 fully recovered COVID controls — under rest, physical effort, recovery, and cold-induced stress, the authors reduced 75 HRV-derived features to 16 and trained ten classifiers. The best model, Gradient Boost, reached 85.2% accuracy, an F1-score of 84.9%, and an AUC (a standard measure of class separation) of 0.907; every model reached at least 0.833 AUC. A sympathetic reader would take this as evidence that HRV changes under exertion and stress differ between long COVID patients and recovered controls, and that machine learning could support diagnosis. The study is observational, so the claim is one of association, not causation.

What carries the argument

The central object is a 16-dimensional feature vector built from four HRV indices — rMSSD (short-term variability), HRVi (triangular index), SD2 (overall variability), and mean heart rate — each measured in four situations: during a six-minute walk, the walk-minus-baseline difference, recovery-minus-walk difference, and a cold water hand-immersion test. The argument turns on combining these weak individual signals, since univariate comparisons showed almost no significant differences except for mean heart rate. Gradient Boost, an ensemble method that sequentially combines weak decision trees, is the classifier that extracts the strongest separation; a 5-fold cross-validation provides the rep

What would settle it

Run the same 16-feature Gradient Boost pipeline on the same dataset with feature selection and hyperparameter tuning performed inside each cross-validation fold — never looking at the held-out subjects until the end — and compare the AUC. If it falls to near 0.5, the reported 0.907 was an artifact of the validation protocol; if it stays above 0.833, the association holds.

Watch

Extended reading notes

Core claim

On its own terms, the paper demonstrates that a small set of HRV indices — rMSSD, HRVi, SD2, and mean heart rate, measured during a six-minute walk, the walk minus baseline, recovery minus walk, and a cold water hand-immersion test — can separate long COVID patients from controls who have fully recovered from COVID-19. The authors report that resting recordings alone showed little separation, while effort and stress conditions brought out differences. After reducing 75 features to 16 and tuning ten classifiers, the Gradient Boost model achieved 85.2% accuracy, 84.9% F1-score, and 0.907 AUC, and all ten models reached at least 0.833 AUC. They also report that mean heart rate features were the

Load-bearing premise

The load-bearing premise is that the feature selection and hyperparameter tuning, done before the single 5-fold cross-validation, leaked no information from the held-out folds into the models.

Editorial extensions

If this is right

  • If the association holds, a six-minute walk plus recovery recording with a chest strap could become a low-cost, non-invasive screening aid for long COVID.
  • Because resting HRV alone showed little separation, any clinical protocol should include effort or stress challenges to reveal the signal.
  • The most informative features — mean heart rate during effort and recovery, plus rMSSD — point toward autonomic dysregulation, giving a concrete physiological target for future studies.
  • All ten models reaching at least 0.833 AUC indicates the group difference is not an artifact of a single classifier.
  • Combining HRV with other physiological markers, as the authors suggest, is a direct route to higher accuracy.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The reported figures should be read as upper bounds: feature selection and hyperparameter tuning were done before the single 5-fold cross-validation, so a validation that keeps selection and tuning inside each training fold could yield lower performance.
  • Since the authors state the data are not publicly available and that their results are hard to compare with earlier work, an independent cohort using the same protocol is the clearest way to confirm the association.
  • A natural extension is to test whether the same effort/recovery HRV signature appears in other post-viral fatigue conditions, which would show whether the signal is specific to long COVID or a general autonomic after-effect of infection.
  • The prominence of mean heart rate features suggests a simpler wearable device might capture much of the signal, though that needs direct testing.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper reports an observational ML study at the University Hospital of Ourense in which 63 long-COVID patients and 66 recovered COVID controls (129 subjects after quality filtering) underwent five heart-rate recordings: rest, six-minute walk, post-walk recovery, cold pressor, and post-cold-pressor recovery. Fifteen HRV indices are extracted per recording (75 features per subject), reduced to 16 features combining rMSSD, HRVi, SD2, and mHR across four conditions. Ten classifiers are trained, and the best reported model, Gradient Boost, is claimed to reach 85.2% accuracy, 84.9% F1-score, and 0.907 AUC, with all models above 0.833 AUC. The authors conclude that this demonstrates an association between long COVID and HRV and supports the diagnosis of long COVID.

Significance. The clinical question is relevant, and the study design has genuine strengths: a clinically characterized cohort, standardized provocative tests (six-minute walk, cold pressor), HRV features from several physiological conditions, comparison of ten classifiers, and a fixed random seed. The existence of an autonomic/HRV long-COVID association is independently supported by prior studies cited by the authors (Shah et al. 2022; Mooren et al. 2023; da Silva et al. 2023; Acanfora et al. 2022), so the paper is not chasing a purely self-generated signal. However, the specific quantitative contribution — the high accuracy and AUC values — is not yet established because the reported evaluation leaks label information into feature selection and hyperparameter tuning. The study could become a useful contribution if the analysis is re-run with a leak-free validation protocol; in its current form, the headline numbers should not be used as evidence of diagnostic utility.

major comments (4)
  1. [§2.5–§2.6 and Table 4] The reported discrimination metrics are produced by a leaky validation protocol. In §2.5, the 16-feature set is chosen by an 'initial study' of discrimination ability between long COVID patients and controls using the full dataset of 129 subjects, with the statistical criterion not specified. In §2.6, hyperparameters are then tuned on the same data ('all parameters where tested') before a 5-fold CV is performed on the complete dataset. Therefore, the test folds are not independent: every subject's label contributed to feature selection and model configuration. This selection leakage biases accuracy/F1/AUC upward, and with n=129 and 75 candidate features the bias can be substantial. The authors' own Fig. 2 shows little univariate separation for rMSSD(6-B), HRVi(6-B), and SD2(6-B), so the multivariate AUCs are heavily dependent on the validity of the selection step. A nested CV with featur
  2. [§2.1] The two groups differ in baseline characteristics that are not adjusted for or discussed. The long-COVID group has BMI 27.90 ± 5.42 and 44 women / 22 men; the control group has BMI 25.18 ± 4.29 and 40 women / 29 men. Age is similar, but HRV and mean heart rate are strongly associated with BMI and sex. Without adjustment or a matched-subgroup analysis, the classifiers may be capturing these group differences rather than long-COVID-specific autonomic dysfunction. This is especially relevant because mHR features are reported as the most important in Fig. 4, and mHR is not an HRV-derived index.
  3. [Section 4 and Introduction (Shah et al. 2022)] The conclusion states: 'We are not aware of previous works that use HRV and ML techniques to compare long COVID patients against cured COVID subjects.' However, the Introduction describes Shah et al. (2022) as using HRV time-domain features with an artificial-intelligence model and reporting 89.8% AUC and 89% accuracy for classification in post-COVID syndrome. This directly contradicts the novelty claim unless the authors specify a different comparison or target. The novelty statement must be corrected or qualified.
  4. [Section 3 and Table 4] Performance is reported as point estimates only, with no confidence intervals, no per-fold dispersion, and no correction for selecting the best of ten models after seeing the CV results. Even if feature selection and tuning were leak-free, choosing the best model on the same folds used for evaluation overstates expected performance. The statement that 'all models ... demonstrate ... AUC values exceeding 83%' is also stronger than what Table 4 shows: SGD is listed at 0.833, not above it. The evaluation section should report fold-level variability and treat the best-model numbers as exploratory unless nested model selection is performed.
minor comments (5)
  1. [Abstract and Table 4] The abstract says 'all models exceed 0.833 AUC', but Table 4 lists SGD with AUC = 0.833. Change to 'at least 0.833' or report rounded values consistently.
  2. [§2.6] Typographical and wording issues: 'Stochatic Gradient Descent' should be 'Stochastic Gradient Descent'; 'ensemblemethod' is missing a space; 'the lost function' should be 'the loss function'.
  3. [Table 1] The SD1 definition contains a duplicated phrase: 'perpendicular to the the identity line'.
  4. [Abstract] The phrase 'supporting its diagnose' should be 'supporting its diagnosis'.
  5. [Title page / affiliations] The affiliation numbering appears to contain a duplicate (two affiliations are both labelled 2), and the listed affiliations of the first two authors seem inconsistent with the contact information. Please correct.

Circularity Check

1 steps flagged · score 4.0 of 10

Reported predictive performance is partly generated by a label-informed evaluation loop: features and hyperparameters were selected on the full dataset before the single 5-fold CV. The central long-COVID/HRV association retains independent external support.

  1. fitted input called prediction [§2.5–§2.6 (Feature selection; Training and validation)]
    "an initial study was performed to choose which indices provide better discrimination ability between long COVID patients and the control group. ... all parameters where tested in order to find the most appropriate values. ... Once all models were configured, a 5-fold cross-validation was performed on the complete dataset."

    The 16-feature input representation was chosen by an 'initial study' of class-discrimination ability on the same 129 subjects whose labels are later predicted, and hyperparameters were tuned by testing 'all parameters' on that complete dataset before the CV. The CV test folds are therefore not independent: their labels already influenced which features and model settings survived. The reported 85.2% accuracy and 0.907 AUC are not estimates of a fixed a-priori pipeline; they partly re-express the label-based feature-selection signal as predictive performance. This is selection leakage, so the headline discrimination is not independent evidence for the association.

full rationale

The paper's central claim—that long COVID is associated with HRV changes—does not reduce to a self-citation or definition; it is supported by independent cited studies (Shah et al. 2022; Mooren et al. 2023; da Silva et al. 2023; Acanfora et al. 2022) and by the data collection itself. However, the paper's novel quantitative evidence, the ML discrimination table, is weakened by a circular evaluation loop: §2.5 selects features by their 'discrimination ability between long COVID patients and the control group' on the full dataset, and §2.6 tunes hyperparameters on the same complete dataset before a single 5-fold CV. Thus the reported test-fold predictions are not fully out-of-sample; the feature set already encodes the outcome labels. This is a correctness/validity flaw rather than a mathematical equivalence—the CV still trains model coefficients on training folds and the magnitude of the bias is not quantified—so it does not make the whole paper circular. The literature-review inconsistency ('We are not aware of previous works that use HRV and ML techniques to compare long COVID patients against cured COVID subjects' despite citing Shah et al. 2022) is a separate consistency issue, not load-bearing circularity. Score 4 reflects one partially circular evaluation step while the central association has independent support.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper's contribution rests on the 16-feature subset it selects from 75 candidates on the full dataset, an unreported significance/correlation threshold, and hyperparameters chosen by exhaustive search — all before a single 5-fold CV is reported. No new entities are introduced. The HRV-index validity and group-label accuracy are domain assumptions inherited from the clinical literature.

free parameters (4)
  • Feature-subset selection (75 → 16 features) = rMSSD, HRVi, SD2, mHR across {6WT, 6-B, A-6, CPT}
    Chosen by exploratory discrimination analysis and correlation-based reduction on the full dataset (§2.5), then evaluated by 5-fold CV on the same data — the headline accuracy/AUC are conditioned on these data-driven choices.
  • Discrimination/significance thresholds in exploratory analysis = not reported
    Which features 'proved to be relevant' or 'yielded statistically significant differences' is asserted without p-values, effect sizes, or correlation cutoffs (§2.5); the LF/HF exclusion is a data-driven choice.
  • ML hyperparameters = Table 3 (e.g., GB: max_depth=20, n_estimators=11, min_samples_leaf=20; SVC: C=0.2)
    Tuned by exhaustive search on the dataset; regardless of the internal CV used for tuning, evaluating the tuned models with a single 5-fold CV on the same data selects the best of many configurations and overstates expected performance (§2.6).
  • Recording-segment windows = 2-min BLR central; 3-min 6WT central; 1-min 6WA; 50-s CPT; 100-s CPA
    Segment lengths and discards are hand-chosen (§2.4); the 50-s and 60-s segments are shorter than the ~5 minutes typically recommended for stable frequency-domain HRV, which affects feature reliability.
assumptions (4)
  • domain assumption The 15 HRV indices extracted from short, non-stationary segments (50 s to 3 min) are valid, stable measures of autonomic function.
    Frequency-domain HRV indices from sub-5-minute segments are noisy; the paper uses a 50-second CPT segment and 1-minute 6WA segment (§2.4) without validating against longer recordings.
  • domain assumption Group labels (long COVID vs fully recovered) are accurate given recruitment criteria and self-reported symptoms.
    Long COVID classification rests on symptom report >2 months after confirmed infection (§2.1); mislabeled subjects would bias the classifier and inflate or deflate separability.
  • domain assumption The 6-minute walk test and cold pressor test elicit the intended autonomic states comparably across all subjects.
    Effort and stress intensity are uncontrolled between subjects (e.g., walking distance in 6WT is not recorded or normalized), so 6WT and CPT features mix ANS response with physical performance (§2.2).
  • ad hoc to paper 5-fold cross-validation estimates generalization when feature selection is external to the folds.
    The paper's protocol assumes the CV performance numbers reflect generalizable discrimination even though the 16-feature subset was selected using the full dataset's labels (§2.5, §2.6); this is the load-bearing methodological premise.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine Learning-based detection of long COVID using Heart Rate Variability Analysis." pith.science (2026). https://pith.science/paper/3GYIEZKC

@misc{pith2026260713668,
  author       = {Pith},
  title        = {Pith review of: Machine Learning-based detection of long COVID using Heart Rate Variability Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3GYIEZKC}},
  note         = {Machine review of arXiv:2607.13668}
}
read the original abstract

After COVID epidemic has ravaged the world, around 20% of infected subjects continue to manifest symptoms several months after their cure. This disorder is called long COVID. This paper presents a study carried out at the University Hospital of Ourense with the aim of establishing a relationship among the disease and variations in Heart rate variability (HRV) parameters using machine learning (ML). Five heart rate recordings were obtained per subject, both at rest and under conditions of physical effort and stress. Each record was processed and 15 HRV indices were extracted, giving 75 features per patient. Of these features, 16 were selected to train 10 different ML models: Support Vector Classification, Linear Support Vector Classification, Logistic Regression, Linear Discriminant Analysis, Stochastic Gradient Descent, Multiple Layer Perceptron, Naive Bayes, Random Forest, and Gradient and ADA Boost Classifiers. Results show that the best model, Gradient Boost, achieves an accuracy of 85.2%, F1-score of 84.9%, and an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.907, and that all models exceed 0.833 AUC. This study demonstrates an association between long COVID and heart rate variability (HRV), highlighting the utility of machine learning models in identifying this relationship and supporting its diagnose.

Figures

Figures reproduced from arXiv: 2607.13668 by the authors.

Figure 1
Figure 1. Example of data obtained from a control subject [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Comparative plots of 6-B records for rMSSD, HRVi, SD2 and mHR parameters [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Training and validation procedure • Stochatic Gradient Descent (SGD): the base of this algorithm is to iteratively adjust some parameters to minimize a cost function. In practice, the function to minimize is E(ω, b) = 1 n Pn i=1 L(yi , f(xi)) + αR(ω), being L a loss function that measures the model misfit. The perceptron function L(yi , f(xi)) = max(0, −yif(xi)) was used. The parameter that controls the regularizati… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Top relative importances of the features given by the RF model (left) and features with highest log odds ratio [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

67 extracted references

  1. [1]

    Journal of Pain Research , pages=

    Pain symptoms in patients with coronavirus disease (COVID-19): A literature review , author=. Journal of Pain Research , pages=. 2021 , publisher=

    2021

  2. [2]

    Frontiers in physiology , volume=

    Alteration of autonomic nervous system is associated with severity and outcomes in patients with COVID-19 , author=. Frontiers in physiology , volume=. 2021 , publisher=

    2021

  3. [3]

    Physiotherapy , volume=

    The 6-minute walk test in outpatient cardiac rehabilitation: validity, reliability and responsiveness—a systematic review , author=. Physiotherapy , volume=. 2012 , publisher=

    2012

  4. [4]

    Clinical Autonomic Research , volume=

    The expanding role of the cold pressor test: a brief history , author=. Clinical Autonomic Research , volume=. 2021 , publisher=

    2021

  5. [5]

    Scientific Reports , volume=

    Autonomic dysregulation in long-term patients suffering from Post-COVID-19 Syndrome assessed by heart rate variability , author=. Scientific Reports , volume=. 2023 , publisher=

    2023

  6. [6]

    Scientific Reports , volume=

    Impact of long COVID on the heart rate variability at rest and during deep breathing maneuver , author=. Scientific Reports , volume=. 2023 , publisher=

    2023

  7. [7]

    Journal of Clinical Medicine , volume=

    Cardiac Autonomic Function in Long COVID-19 Using Heart Rate Variability: An Observational Cross-Sectional Study , author=. Journal of Clinical Medicine , volume=. 2022 , publisher=

    2022

  8. [8]

    Biology , volume=

    Imbalance of Peripheral Temperature, Sympathovagal, and Cytokine Profile in Long COVID , author=. Biology , volume=. 2023 , publisher=

    2023

Show all 67 references
  1. [9]

    2015 , publisher=

    What is machine learning? , author=. 2015 , publisher=

    2015

  2. [10]

    Scientific Reports , volume=

    Cancer classification using machine learning and HRV analysis: preliminary evidence from a pilot study , author=. Scientific Reports , volume=. 2021 , publisher=

    2021

  3. [11]

    IEEE Access , volume=

    Screening cardiovascular autonomic neuropathy in diabetic patients with microvascular complications using machine learning: a 24-hour heart rate variability study , author=. IEEE Access , volume=. 2021 , publisher=

    2021

  4. [12]

    Journal of Electrocardiology , volume=

    An automated detection of atrial fibrillation from single-lead ECG using HRV features and machine learning , author=. Journal of Electrocardiology , volume=. 2022 , publisher=

    2022

  5. [13]

    science , volume=

    Power spectrum analysis of heart rate fluctuation: a quantitative probe of beat-to-beat cardiovascular control , author=. science , volume=. 1981 , publisher=

    1981

  6. [14]

    Frontiers in public health , pages=

    An overview of heart rate variability metrics and norms , author=. Frontiers in public health , pages=. 2017 , publisher=

    2017

  7. [15]

    Diabetes , volume=

    Heart-rate variability and cardiac autonomic function in diabetes , author=. Diabetes , volume=. 1990 , publisher=

    1990

  8. [16]

    Clinical cardiology , volume=

    Heart rate variability , author=. Clinical cardiology , volume=. 1990 , publisher=

    1990

  9. [17]

    European heart journal , volume=

    Heart rate variability in relation to prognosis after myocardial infarction: selection of optimal processing techniques , author=. European heart journal , volume=. 1989 , publisher=

    1989

  10. [18]

    Psychiatry investigation , volume=

    Stress and heart rate variability: a meta-analysis and review of the literature , author=. Psychiatry investigation , volume=. 2018 , publisher=

    2018

  11. [19]

    Alternative therapies , volume=

    Effect of laughter yoga on mood and heart rate variability in patients awaiting organ transplantation: a pilot study , author=. Alternative therapies , volume=

  12. [20]

    PLoS One , volume=

    Heart-rate-variability (HRV), predicts outcomes in COVID-19 , author=. PLoS One , volume=. 2021 , publisher=

    2021

  13. [21]

    medRxiv , pages=

    Heart rate variability as a prospective predictor of early Covid-19 symptoms , author=. medRxiv , pages=. 2021 , publisher=

    2021

  14. [22]

    Scientific Reports , volume=

    Automatic COVID-19 severity assessment from HRV , author=. Scientific Reports , volume=. 2023 , publisher=

    2023

  15. [23]

    Journal of Basic and Clinical Physiology and Pharmacology , volume=

    Characterization of cardiac autonomic function in COVID-19 using heart rate variability: A hospital based preliminary observational study , author=. Journal of Basic and Clinical Physiology and Pharmacology , volume=. 2021 , publisher=

    2021

  16. [24]

    Frontiers in cardiovascular medicine , volume=

    Reduction of cardiac autonomic modulation and increased sympathetic activity by heart rate variability in patients with long COVID , author=. Frontiers in cardiovascular medicine , volume=. 2022 , publisher=

    2022

  17. [25]

    Journal of medical Internet research , volume=

    Use of physiological data from a wearable device to identify SARS-CoV-2 infection and symptoms and predict COVID-19 diagnosis: observational study , author=. Journal of medical Internet research , volume=. 2021 , publisher=

    2021

  18. [26]

    Scientific reports , volume=

    Detection of COVID-19 using multimodal data from a wearable device: results from the first TemPredict Study , author=. Scientific reports , volume=. 2022 , publisher=

    2022

  19. [27]

    Scientific reports , volume=

    Clinical characterization of dysautonomia in long COVID-19 patients , author=. Scientific reports , volume=. 2021 , publisher=

    2021

  20. [28]

    PloS one , volume=

    Analyzing changes in respiratory rate to predict the risk of COVID-19 infection , author=. PloS one , volume=. 2020 , publisher=

    2020

  21. [29]

    Sensors , volume=

    A machine learning approach as an aid for early covid-19 detection , author=. Sensors , volume=. 2021 , publisher=

    2021

  22. [30]

    Journal of basic and clinical physiology and pharmacology , volume=

    Characterization of cardiac autonomic function in COVID-19 using heart rate variability: A hospital based preliminary observational study , author=. Journal of basic and clinical physiology and pharmacology , volume=. 2021 , publisher=

    2021

  23. [31]

    Viruses , volume=

    Impaired vagal activity in long-COVID-19 patients , author=. Viruses , volume=. 2022 , publisher=

    2022

  24. [32]

    Indian pacing and electrophysiology journal , volume=

    Heart rate variability as a marker of cardiovascular dysautonomia in post-COVID-19 syndrome using artificial intelligence , author=. Indian pacing and electrophysiology journal , volume=. 2022 , publisher=

    2022

  25. [33]

    2020 Third International Conference on Smart Systems and Inventive Technology (ICSSIT) , pages=

    Study the influence of normalization/transformation process on the accuracy of supervised classification , author=. 2020 Third International Conference on Smart Systems and Inventive Technology (ICSSIT) , pages=. 2020 , organization=

    2020

  26. [34]

    Information , volume=

    Machine learning in python: Main developments and technology trends in data science, machine learning, and artificial intelligence , author=. Information , volume=. 2020 , publisher=

    2020

  27. [35]

    Encyclopedia of Bioinformatics and Computational Biology , publisher=

    Berrar, Daniel , year= 2019, chapter=. Encyclopedia of Bioinformatics and Computational Biology , publisher=

    2019

  28. [36]

    2009 , publisher=

    Logistic regression models , author=. 2009 , publisher=

    2009

  29. [37]

    Nature biotechnology , volume=

    What is a support vector machine? , author=. Nature biotechnology , volume=. 2006 , publisher=

    2006

  30. [38]

    Encyclopedia of Bioinformatics and Computational Biology: ABC of Bioinformatics , volume=

    Bayes’ theorem and naive Bayes classifier , author=. Encyclopedia of Bioinformatics and Computational Biology: ABC of Bioinformatics , volume=. 2018 , publisher=

    2018

  31. [39]

    Institute for Signal and information Processing , volume=

    Linear discriminant analysis-a brief tutorial , author=. Institute for Signal and information Processing , volume=. 1998 , publisher=

    1998

  32. [40]

    The Journal of Machine Learning Research , volume=

    Analysis of a random forests model , author=. The Journal of Machine Learning Research , volume=. 2012 , publisher=

    2012

  33. [41]

    2022 , publisher=

    Inside deep learning: Math, algorithms, models , author=. 2022 , publisher=

    2022

  34. [42]

    1994 , publisher=

    An introduction to the bootstrap , author=. 1994 , publisher=

    1994

  35. [43]

    2009 , publisher=

    The elements of statistical learning: data mining, inference, and prediction , author=. 2009 , publisher=

    2009

  36. [44]

    IEEE Transactions on neural networks , volume=

    Perceptron-based learning algorithms , author=. IEEE Transactions on neural networks , volume=

  37. [45]

    International Journal of Advanced Computer Science and Applications , volume=

    Classification model evaluation metrics , author=. International Journal of Advanced Computer Science and Applications , volume=

  38. [46]

    Respiratory care , volume=

    The six-minute walk test , author=. Respiratory care , volume=. 2003 , publisher=

    2003

  39. [47]

    Advances in Physiology Education , volume=

    Cold stress and the cold pressor test , author=. Advances in Physiology Education , volume=. 2013 , publisher=

    2013

  40. [48]

    2017 , publisher=

    Heart rate variability analysis with the R package RHRV , author=. 2017 , publisher=

    2017

  41. [49]

    Journal of Medical Systems , volume=

    Evidence based recommendations for designing heart rate variability studies , author=. Journal of Medical Systems , volume=. 2019 , publisher=

    2019

  42. [50]

    IEEE Engineering in Medicine and Biology Magazine , volume=

    Time-frequency analysis of heart-rate variability , author=. IEEE Engineering in Medicine and Biology Magazine , volume=. 1997 , publisher=

    1997

  43. [51]

    International Journal of Medical Informatics , volume=

    VARSE: Android app for real-time acquisition and analysis of heart rate signals , author=. International Journal of Medical Informatics , volume=. 2022 , publisher=

    2022

  44. [52]

    2013 , publisher=

    The R software , author=. 2013 , publisher=

    2013

  45. [53]

    2014 , publisher=

    A handbook of statistical analyses using R , author=. 2014 , publisher=

    2014

  46. [54]

    The Corsini encyclopedia of psychology , pages=

    Mann-Whitney U Test , author=. The Corsini encyclopedia of psychology , pages=. 2010 , publisher=

    2010

  47. [55]

    2019 , publisher=

    Statistical computing with R , author=. 2019 , publisher=

    2019

  48. [56]

    International Conference on Intelligent Computing and Smart Communication 2019: Proceedings of ICSC 2019 , pages=

    Study of adaboost and gradient boosting algorithms for predictive analytics , author=. International Conference on Intelligent Computing and Smart Communication 2019: Proceedings of ICSC 2019 , pages=. 2020 , organization=

    2019

  49. [57]

    2017 , publisher=

    Mastering Machine Learning with scikit-learn , author=. 2017 , publisher=

    2017

  50. [58]

    Hyperparameter optimization in machine learning: make your machine learning and deep learning models more efficient , pages=

    Hyperparameter optimization using scikit-learn , author=. Hyperparameter optimization in machine learning: make your machine learning and deep learning models more efficient , pages=. 2021 , publisher=

    2021

  51. [59]

    World Health Organization , howpublished = "

  52. [60]

    Reviews in the Neurosciences , volume=

    Clinical experimental stress studies: methods and assessment , author=. Reviews in the Neurosciences , volume=. 2015 , publisher=

    2015

  53. [61]

    Neurophysiologie Clinique , volume=

    Sudomotor dysfunction in patients recovered from COVID-19 , author=. Neurophysiologie Clinique , volume=. 2021 , publisher=

    2021

  54. [62]

    Long COVID

    Cardiovascular autonomic dysfunction in “Long COVID”: pathophysiology, heart rate variability, and inflammatory markers , author=. Frontiers in cardiovascular medicine , volume=. 2023 , publisher=

    2023

  55. [63]

    Journal of Electrocardiology , volume=

    Relationship between ultra-short heart rate variability and short-term mortality in hospitalized COVID-19 patients , author=. Journal of Electrocardiology , volume=. 2024 , publisher=

    2024

  56. [64]

    BMC Infectious Diseases , volume=

    The impact of long COVID on heart rate variability: a cross-sectional study , author=. BMC Infectious Diseases , volume=. 2025 , publisher=

    2025

  57. [65]

    Journal of Interventional Cardiac Electrophysiology , volume=

    Heart rate variability and cardiac autonomic functions in post-COVID period , author=. Journal of Interventional Cardiac Electrophysiology , volume=. 2022 , publisher=

    2022

  58. [66]

    Journal of sports science & medicine , volume=

    Acute effects of an incremental exercise test on psychophysiological variables and their Interaction , author=. Journal of sports science & medicine , volume=

  59. [67]

    Healthcare , volume=

    Long-term impact of COVID-19 on heart rate variability: a systematic review of observational studies , author=. Healthcare , volume=. 2023 , organization=

    2023

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.