pith. sign in

arxiv: 2305.09288 · v1 · submitted 2023-05-16 · 💻 cs.LG

A Dictionary-based approach to Time Series Ordinal Classification

Rafael Ayll\'on-Gavil\'an , David Guijo-Rubio , Pedro Antonio Guti\'errez , C\'esar Herv\'as-Martinez This is my paper

Pith reviewed 2026-05-24 08:50 UTC · model grok-4.3

classification 💻 cs.LG
keywords time series classificationordinal classificationdictionary-based methodsTemporal Dictionary EnsembleTSOC
0
0 comments X

The pith

An ordinal version of the Temporal Dictionary Ensemble improves accuracy on time series problems whose labels have a natural order.

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

The paper takes the current leading dictionary-based method for ordinary time series classification and modifies it to respect the ordering among class labels. It runs the new ordinal version against four standard nominal dictionary methods on eighteen problems that carry ordered labels. The experiments indicate that the ordering can be turned into a performance gain. Readers would care because many practical time series tasks involve graded outcomes such as severity levels or stages, yet most existing dictionary techniques treat those grades as unrelated categories.

Core claim

The ordinal TDE (O-TDE) is created by adapting the Temporal Dictionary Ensemble so that its internal decisions respect the natural ordering present in the labels; when tested on eighteen time series ordinal classification problems, O-TDE produces higher accuracy than four existing nominal dictionary-based techniques.

What carries the argument

The ordinal adaptation of the Temporal Dictionary Ensemble (O-TDE), which alters the ensemble's handling of class labels to exploit their ordering during dictionary construction and classification.

If this is right

  • Dictionary-based time series classifiers become usable on problems whose classes form an ordered scale.
  • The TSOC setting, previously unexplored with dictionary methods, now has at least one competitive approach.
  • Performance gains appear without requiring changes to the underlying dictionary extraction or ensemble steps beyond the ordinal adjustment.
  • The same adaptation principle could be applied to other dictionary ensembles that currently treat labels as nominal.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar ordinal adjustments might lift accuracy in non-dictionary time series methods that already handle ordered outputs.
  • Domains such as medical staging or quality grading could adopt the method once the eighteen-problem result is replicated on larger collections.
  • If the gain holds, future dictionary methods may default to an ordinal mode whenever labels are known to be ordered.

Load-bearing premise

That the natural ordering among labels can be used by a dictionary method to raise accuracy without introducing new biases or needing problem-specific adjustments.

What would settle it

A re-run of the eighteen problems in which O-TDE shows no accuracy improvement over the four nominal dictionary baselines under the same cross-validation protocol.

Figures

Figures reproduced from arXiv: 2305.09288 by C\'esar Herv\'as-Martinez, David Guijo-Rubio, Pedro Antonio Guti\'errez, Rafael Ayll\'on-Gavil\'an.

Figure 1
Figure 1. Figure 1: CDDs in terms of MAE (a), QWK (b), CCR (c) and 1-OFF (d). The [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
read the original abstract

Time Series Classification (TSC) is an extensively researched field from which a broad range of real-world problems can be addressed obtaining excellent results. One sort of the approaches performing well are the so-called dictionary-based techniques. The Temporal Dictionary Ensemble (TDE) is the current state-of-the-art dictionary-based TSC approach. In many TSC problems we find a natural ordering in the labels associated with the time series. This characteristic is referred to as ordinality, and can be exploited to improve the methods performance. The area dealing with ordinal time series is the Time Series Ordinal Classification (TSOC) field, which is yet unexplored. In this work, we present an ordinal adaptation of the TDE algorithm, known as ordinal TDE (O-TDE). For this, a comprehensive comparison using a set of 18 TSOC problems is performed. Experiments conducted show the improvement achieved by the ordinal dictionary-based approach in comparison to four other existing nominal dictionary-based techniques.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes an ordinal adaptation of the Temporal Dictionary Ensemble (TDE) called O-TDE for Time Series Ordinal Classification (TSOC). It reports a comparison on 18 TSOC problems and claims that O-TDE improves upon four existing nominal dictionary-based techniques by exploiting the natural ordering in the labels.

Significance. If the empirical gains are robust and the adaptation is clearly specified, the work would have moderate significance: TSOC is described as unexplored, dictionary methods are SOTA for TSC, and an ordinal variant could provide a useful baseline. The use of 18 problems is a reasonable scale for an initial study, though the current text supplies no quantitative results or method details with which to evaluate the contribution.

major comments (2)
  1. [Abstract] Abstract: the central claim that 'Experiments conducted show the improvement achieved by the ordinal dictionary-based approach' is unsupported by any description of the ordinal adaptation to TDE, any performance numbers, any statistical tests, or any account of how ordinality is encoded in the dictionary or ensemble steps. This absence is load-bearing for the empirical claim.
  2. [Abstract] Abstract: the assumption that label ordering can be exploited to produce measurable gains 'without introducing new biases or requiring problem-specific tuning' is stated but receives no supporting analysis, ablation, or discussion of the ordinal encoding mechanism.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments on the abstract. We agree that it would benefit from additional detail on the ordinal adaptation, results, and supporting claims to better substantiate the contribution. We will revise the abstract in the next version. Below we respond point by point.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that 'Experiments conducted show the improvement achieved by the ordinal dictionary-based approach' is unsupported by any description of the ordinal adaptation to TDE, any performance numbers, any statistical tests, or any account of how ordinality is encoded in the dictionary or ensemble steps. This absence is load-bearing for the empirical claim.

    Authors: The abstract is necessarily concise, but the full manuscript (Section 3) specifies the O-TDE adaptation: ordinality is encoded by replacing Euclidean distance with an ordinal-aware distance on word histograms and by using ordered aggregation in the ensemble rather than majority vote. Section 4 reports results on the 18 TSOC datasets, including mean accuracy, rank, and statistical tests (e.g., Wilcoxon signed-rank against the four nominal baselines). We will expand the abstract to include a one-sentence description of the adaptation and a summary of the performance gains with statistical support. revision: yes

  2. Referee: [Abstract] Abstract: the assumption that label ordering can be exploited to produce measurable gains 'without introducing new biases or requiring problem-specific tuning' is stated but receives no supporting analysis, ablation, or discussion of the ordinal encoding mechanism.

    Authors: The method section explains that the ordinal encoding reuses the existing TDE pipeline with only the distance and voting steps modified via standard ordinal techniques; no new hyperparameters or dataset-specific tuning are introduced. We will add a brief clause to the revised abstract and ensure the manuscript includes a short discussion or reference confirming the absence of additional bias or tuning requirements. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper presents an algorithmic adaptation (O-TDE) of an existing method (TDE) followed by an empirical comparison on 18 TSOC datasets against four nominal baselines. No equations, derivations, fitted parameters, or predictions appear in the provided text. The central claim is an observed performance improvement in experiments, which does not reduce to any self-definitional, fitted-input, or self-citation reduction. The work is self-contained as an empirical study with no load-bearing mathematical steps that could exhibit circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on the domain assumption that ordinal label structure can be leveraged inside existing dictionary ensembles without additional data or external validation.

axioms (1)
  • domain assumption Ordinality present in the labels of the chosen 18 problems can be exploited to improve dictionary-based classification performance.
    Stated in the abstract as the motivation for the O-TDE adaptation.

pith-pipeline@v0.9.0 · 5714 in / 1156 out tokens · 19237 ms · 2026-05-24T08:50:06.821293+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages

  1. [1]

    DATA MINING AND KNOWLEDGE DISCOVERY31(3), 606–660 (MAY 2017)

    Bagnall, A., Lines, J., Bostrom, A., Large, J., Keogh, E.: The great time series classification bake off: a review and experimental evaluation of recent algorith- mic advances. DATA MINING AND KNOWLEDGE DISCOVERY31(3), 606–660 (MAY 2017)

    work page 2017

  2. [2]

    Benavoli, A., Corani, G., Mangili, F.: Should we really use post-hoc tests based on mean-ranks? The Journal of Machine Learning Research 17(1), 152–161 (2016)

    work page 2016

  3. [3]

    In: Artificial Intelligence and Soft Computing: 14th International Conference, ICAISC 2015, Zakopane, Poland, June 14-18, 2015, Pro- ceedings, Part II 14

    Buza, K., Koller, J., Marussy, K.: Process: projection-based classification of elec- troencephalograph signals. In: Artificial Intelligence and Soft Computing: 14th International Conference, ICAISC 2015, Zakopane, Poland, June 14-18, 2015, Pro- ceedings, Part II 14. pp. 91–100. Springer (2015)

    work page 2015

  4. [4]

    In: Proceedings of the 22nd international conference on Machine learning

    Chu, W., Keerthi, S.S.: New approaches to support vector ordinal regression. In: Proceedings of the 22nd international conference on Machine learning. pp. 145–152 (2005)

    work page 2005

  5. [5]

    Data Mining and Knowledge Discovery 34, 1454–1495 (2020)

    Dempster, A., Petitjean, F., Webb, G.I.: Rocket: exceptionally fast and accurate time series classification using random convolutional kernels. Data Mining and Knowledge Discovery 34, 1454–1495 (2020)

    work page 2020

  6. [6]

    The Jour- nal of Machine learning research 7, 1–30 (2006)

    Demˇ sar, J.: Statistical comparisons of classifiers over multiple data sets. The Jour- nal of Machine learning research 7, 1–30 (2006)

    work page 2006

  7. [7]

    IEEE Transactions on Cybernetics 43(6), 2228–2240 (DEC 2013)

    Fernandez-Navarro, F., Campoy-Munoz, P., de la Paz-Marin, M., Hervas-Martinez, C., Yao, X.: Addressing the eu sovereign ratings using an ordinal regression ap- proach. IEEE Transactions on Cybernetics 43(6), 2228–2240 (DEC 2013)

    work page 2013

  8. [8]

    The Annals of Statistics 29(5), 1189–1232 (2001)

    Friedman, J.H.: Greedy function approximation: A gradient boosting machine. The Annals of Statistics 29(5), 1189–1232 (2001)

    work page 2001

  9. [9]

    Atmospheric Research 236, 104798 (2020)

    Guijo-Rubio, D., Casanova-Mateo, C., Sanz-Justo, J., Gutierrez, P., Cornejo- Bueno, S., Herv´ as, C., Salcedo-Sanz, S.: Ordinal regression algorithms for the anal- ysis of convective situations over madrid-barajas airport. Atmospheric Research 236, 104798 (2020)

    work page 2020

  10. [10]

    In: Advanced Analytics and Learning on Tempo- ral Data: 5th ECML PKDD Workshop, AALTD 2020, Ghent, Belgium, September 18, 2020, Revised Selected Papers

    Guijo-Rubio, D., Guti´ errez, P.A., Bagnall, A., Herv´ as-Mart´ ınez, C.: Ordinal versus nominal time series classification. In: Advanced Analytics and Learning on Tempo- ral Data: 5th ECML PKDD Workshop, AALTD 2020, Ghent, Belgium, September 18, 2020, Revised Selected Papers. pp. 19–29. Springer (2020)

    work page 2020

  11. [11]

    In: 2020 International Joint Conference on Neural Networks (IJCNN)

    Guijo-Rubio, D., Guti´ errez, P.A., Bagnall, A., Herv´ as-Mart´ ınez, C.: Time series or- dinal classification via shapelets. In: 2020 International Joint Conference on Neural Networks (IJCNN). pp. 1–8 (2020)

    work page 2020

  12. [12]

    Proceedings of the IEEE 66(1), 51–83 (1978)

    Harris, F.J.: On the use of windows for harmonic analysis with the discrete fourier transform. Proceedings of the IEEE 66(1), 51–83 (1978)

    work page 1978

  13. [13]

    Data mining and knowledge discovery 28(4), 851–881 (2014)

    Hills, J., Lines, J., Baranauskas, E., Mapp, J., Bagnall, A.: Classification of time series by shapelet transformation. Data mining and knowledge discovery 28(4), 851–881 (2014)

    work page 2014

  14. [14]

    In: proceedings of the Fifth Balkan Conference in Informatics

    Kurbalija, V., von Bernstorff, C., Burkhard, H.D., Nachtwei, J., Ivanovi´ c, M., Fodor, L.: Time-series mining in a psychological domain. In: proceedings of the Fifth Balkan Conference in Informatics. pp. 58–63 (2012)

    work page 2012

  15. [15]

    Intelligent Data Analysis23(5), 1073–1089 (2019)

    Large, J., Bagnall, A., Malinowski, S., Tavenard, R.: On time series classification with dictionary-based classifiers. Intelligent Data Analysis23(5), 1073–1089 (2019)

    work page 2019

  16. [16]

    In: Advances in Knowledge Discovery and Data Mining: 22nd Pacific-Asia Conference, PAKDD 2018, Melbourne, VIC, Australia, June 3-6, 2018, Proceedings, Part I 22

    Large, J., Kemsley, E.K., Wellner, N., Goodall, I., Bagnall, A.: Detecting forged alcohol non-invasively through vibrational spectroscopy and machine learning. In: Advances in Knowledge Discovery and Data Mining: 22nd Pacific-Asia Conference, PAKDD 2018, Melbourne, VIC, Australia, June 3-6, 2018, Proceedings, Part I 22. pp. 298–309. Springer (2018) 12 R. A...

    work page 2018

  17. [17]

    Data Mining and knowledge discovery 15, 107–144 (2007)

    Lin, J., Keogh, E., Wei, L., Lonardi, S.: Experiencing sax: a novel symbolic repre- sentation of time series. Data Mining and knowledge discovery 15, 107–144 (2007)

    work page 2007

  18. [18]

    Journal of Intelligent Information Systems 39, 287–315 (2012)

    Lin, J., Khade, R., Li, Y.: Rotation-invariant similarity in time series using bag- of-patterns representation. Journal of Intelligent Information Systems 39, 287–315 (2012)

    work page 2012

  19. [19]

    ACM Transactions on Knowledge Discovery from Data 12(5) (2018)

    Lines, J., Taylor, S., Bagnall, A.: Time series classification with hive-cote: The hierarchical vote collective of transformation-based ensembles. ACM Transactions on Knowledge Discovery from Data 12(5) (2018)

    work page 2018

  20. [20]

    Image and Vision Computing 106 (FEB 2021)

    Liu, Y., Wang, Y., Kong, A.W.K.: Pixel-wise ordinal classification for salient object grading. Image and Vision Computing 106 (FEB 2021)

    work page 2021

  21. [21]

    In: ESANN

    Malhotra, P., Vig, L., Shroff, G., Agarwal, P., et al.: Long short term memory networks for anomaly detection in time series. In: ESANN. vol. 2015, p. 89 (2015)

    work page 2015

  22. [22]

    Journal of the Royal Statistical Society: Series B (Methodological) 42(2), 109–127 (1980)

    McCullagh, P.: Regression models for ordinal data. Journal of the Royal Statistical Society: Series B (Methodological) 42(2), 109–127 (1980)

    work page 1980

  23. [23]

    In: Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2020, Ghent, Belgium, September 14–18, 2020, Proceedings, Part I

    Middlehurst, M., Large, J., Cawley, G., Bagnall, A.: The temporal dictionary ensemble (tde) classifier for time series classification. In: Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2020, Ghent, Belgium, September 14–18, 2020, Proceedings, Part I. pp. 660–676. Springer (2021)

    work page 2020

  24. [24]

    MACHINE LEARNING 110(11-12), 3211–3243 (DEC 2021)

    Middlehurst, M., Large, J., Flynn, M., Lines, J., Bostrom, A., Bagnall, A.: Hive- cote 2.0: a new meta ensemble for time series classification. MACHINE LEARNING 110(11-12), 3211–3243 (DEC 2021)

    work page 2021

  25. [25]

    In: Yin, H., Camacho, D., Tino, P., Tall´ on-Ballesteros, A.J., Menezes, R., Allmendinger, R

    Middlehurst, M., Vickers, W., Bagnall, A.: Scalable dictionary classifiers for time series classification. In: Yin, H., Camacho, D., Tino, P., Tall´ on-Ballesteros, A.J., Menezes, R., Allmendinger, R. (eds.) Intelligent Data Engineering and Automated Learning – IDEAL 2019. pp. 11–19. Springer International Publishing, Cham (2019)

    work page 2019

  26. [26]

    Data Mining and Knowledge Discovery 29, 1505–1530 (2015)

    Sch¨ afer, P.: The boss is concerned with time series classification in the presence of noise. Data Mining and Knowledge Discovery 29, 1505–1530 (2015)

    work page 2015

  27. [27]

    In: Proceedings of the 2017 ACM on Conference on Information and Knowledge Management

    Sch¨ afer, P., Leser, U.: Fast and accurate time series classification with weasel. In: Proceedings of the 2017 ACM on Conference on Information and Knowledge Management. pp. 637–646 (2017)

    work page 2017

  28. [28]

    Journal of Mathematical Psychol- ogy 85, 1–16 (2018)

    Schulz, E., Speekenbrink, M., Krause, A.: A tutorial on gaussian process regression: Modelling, exploring, and exploiting functions. Journal of Mathematical Psychol- ogy 85, 1–16 (2018)

    work page 2018

  29. [29]

    Computers in Industry 144, 103786 (2023)

    Vargas, V.M., Guti´ errez, P.A., Rosati, R., Romeo, L., Frontoni, E., Herv´ as- Mart´ ınez, C.: Deep learning based hierarchical classifier for weapon stock aesthetic quality control assessment. Computers in Industry 144, 103786 (2023)

    work page 2023

  30. [30]

    In: 2017 International joint conference on neural networks (IJCNN)

    Wang, Z., Yan, W., Oates, T.: Time series classification from scratch with deep neural networks: A strong baseline. In: 2017 International joint conference on neural networks (IJCNN). pp. 1578–1585. IEEE (2017)

    work page 2017

  31. [31]

    IEEE Transactions on Instrumentation and Measurement 71, 1–12 (2022)

    Zhou, Z., Huang, B., Zhang, R., Yin, M., Liu, C., Liu, Y., Yi, Z., Wu, X.: Methods to recognize depth of hard inclusions in soft tissue using ordinal classification for robotic palpation. IEEE Transactions on Instrumentation and Measurement 71, 1–12 (2022)

    work page 2022