pith. sign in

arxiv: 2305.01429 · v1 · submitted 2023-05-02 · 💻 cs.LG · stat.ML

Unsupervised Feature Based Algorithms for Time Series Extrinsic Regression

David Guijo-Rubio , Matthew Middlehurst , Guilherme Arcencio , Diego Furtado Silva , Anthony Bagnall This is my paper

Pith reviewed 2026-05-24 08:23 UTC · model grok-4.3

classification 💻 cs.LG stat.ML
keywords time series extrinsic regressionFreshPRINCEDrCIFrotation forestInceptionTimebenchmark archivefeature extraction
0
0 comments X

The pith

FreshPRINCE and DrCIF are the only regressors that significantly outperform rotation forest on an expanded set of 63 TSER problems.

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

The paper expands the time series extrinsic regression archive from 19 to 63 problems and re-evaluates a wide range of regressors. It shows that a standard rotation forest regressor remains hard to beat, with most prior methods failing to improve on it. Two new algorithms adapted from time series classification work—FreshPRINCE, which extracts many summary features then applies rotation forest, and DrCIF, which builds features from summary statistics over random intervals—plus InceptionTime, perform significantly better than the remaining 18 regressors. Only FreshPRINCE and DrCIF significantly surpass rotation forest itself. A reader would care because the results supply stronger practical baselines and new methods for predicting a continuous target from time series that are not directly related to it.

Core claim

FreshPRINCE is a pipeline that transforms each time series into a wide range of summary features and passes them to a rotation forest regressor. DrCIF is a tree ensemble that creates features from summary statistics computed over random intervals. On the expanded archive of 63 problems these two algorithms, together with InceptionTime, significantly outperform the other 18 regressors tested; crucially, only FreshPRINCE and DrCIF also significantly outperform the standard rotation forest regressor.

What carries the argument

FreshPRINCE and DrCIF, which adapt unsupervised feature extraction and interval statistics from time series classification into regression pipelines built around rotation forest.

If this is right

  • Rotation forest remains a competitive baseline that new TSER methods must demonstrably beat.
  • Summary-feature pipelines and interval-based statistics transfer successfully from classification to extrinsic regression.
  • InceptionTime is strong but does not significantly surpass rotation forest on this benchmark.
  • The larger archive makes performance differences clearer than the original 19-problem set.

Where Pith is reading between the lines

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

  • Hybrid models that combine FreshPRINCE feature sets with DrCIF interval statistics could be tested next.
  • The same feature constructions might improve regression performance in related time-series tasks such as forecasting.
  • Releasing the 63-problem archive allows direct head-to-head testing of any future TSER proposal against these results.

Load-bearing premise

The 63 problems in the expanded TSER archive are representative enough of real-world extrinsic regression tasks that performance rankings will hold more generally.

What would settle it

A follow-up experiment on a fresh collection of real-world TSER problems in which neither FreshPRINCE nor DrCIF shows a statistically significant advantage over rotation forest.

Figures

Figures reproduced from arXiv: 2305.01429 by Anthony Bagnall, David Guijo-Rubio, Diego Furtado Silva, Guilherme Arcencio, Matthew Middlehurst.

Figure 1
Figure 1. Figure 1: Examples of soil spectrograms used to predict potassium concentration [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Diagrams visualising the DrCIF transformation (left) and DrCIF ensemble structure (right). [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Reproduction of the RMSE ranks on the original archive (19 datasets and 5 resamples). Left is the original image from [5]. Right is our recreation. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: RMSE ranks for 13 Regressors used in [5] on 63 TSER datasets. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: RMSE ranks for two feature based regressors, DrCIF and fresh [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Distribution of relative RMSE for eight regressors (lower values are [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Summary performance results for the best eight regressors. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 10
Figure 10. Figure 10: RMSE ranks for freshPRINCE, standard RotF and TSFresh (Fresh) [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Scatter plot of predicted vs actual for DrCIF on BarCrawl6min. [PITH_FULL_IMAGE:figures/full_fig_p009_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Scatter plot of predicted vs actual for InceptionE on BarCrawl6min. [PITH_FULL_IMAGE:figures/full_fig_p010_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Scatter plot of relative RMSE for DrCIF and InceptionE. [PITH_FULL_IMAGE:figures/full_fig_p010_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Two examples of yearly sound pollution in the AcousticContamina [PITH_FULL_IMAGE:figures/full_fig_p013_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Examples of yearly temperature profiles from two different latitudes [PITH_FULL_IMAGE:figures/full_fig_p014_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Two different years of measurements and the associated average [PITH_FULL_IMAGE:figures/full_fig_p014_16.png] view at source ↗
Figure 18
Figure 18. Figure 18: Temperature data and respective methane concentration in the [PITH_FULL_IMAGE:figures/full_fig_p015_18.png] view at source ↗
Figure 17
Figure 17. Figure 17: Example of currents, voltages, and ambient and coolant temperatures [PITH_FULL_IMAGE:figures/full_fig_p015_17.png] view at source ↗
Figure 20
Figure 20. Figure 20: Two different Covid-19 waves and their mortality rates in the [PITH_FULL_IMAGE:figures/full_fig_p016_20.png] view at source ↗
Figure 19
Figure 19. Figure 19: Two samples of accelerometer data and alcohol concentration in the [PITH_FULL_IMAGE:figures/full_fig_p016_19.png] view at source ↗
Figure 21
Figure 21. Figure 21: Examples of gas price time series from the NaturalGasPriceSentiment [PITH_FULL_IMAGE:figures/full_fig_p017_21.png] view at source ↗
read the original abstract

Time Series Extrinsic Regression (TSER) involves using a set of training time series to form a predictive model of a continuous response variable that is not directly related to the regressor series. The TSER archive for comparing algorithms was released in 2022 with 19 problems. We increase the size of this archive to 63 problems and reproduce the previous comparison of baseline algorithms. We then extend the comparison to include a wider range of standard regressors and the latest versions of TSER models used in the previous study. We show that none of the previously evaluated regressors can outperform a regression adaptation of a standard classifier, rotation forest. We introduce two new TSER algorithms developed from related work in time series classification. FreshPRINCE is a pipeline estimator consisting of a transform into a wide range of summary features followed by a rotation forest regressor. DrCIF is a tree ensemble that creates features from summary statistics over random intervals. Our study demonstrates that both algorithms, along with InceptionTime, exhibit significantly better performance compared to the other 18 regressors tested. More importantly, these two proposals (DrCIF and FreshPRINCE) models are the only ones that significantly outperform the standard rotation forest regressor.

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 / 1 minor

Summary. The manuscript claims to expand the TSER archive from 19 to 63 problems, reproduce prior baseline comparisons, evaluate a wider range of regressors, and introduce two new algorithms (FreshPRINCE: a summary-feature pipeline with rotation forest; DrCIF: an interval-based tree ensemble). It reports that DrCIF, FreshPRINCE, and InceptionTime significantly outperform the other 18 regressors tested, and that only DrCIF and FreshPRINCE significantly beat the rotation forest regressor.

Significance. If the empirical claims hold after clarification of protocols, the work would be useful for enlarging the TSER benchmark and highlighting competitive feature-based methods. The reproduction of prior results, identification of rotation forest as a strong baseline, and derivation of DrCIF/FreshPRINCE from classification techniques are positive contributions. The value is primarily in the expanded archive and the performance ranking, provided the archive is representative and the statistics are transparent.

major comments (2)
  1. [TSER Archive expansion section] TSER Archive expansion section: No description is provided of how the 44 new problems were sourced, selected, or validated for diversity in domains, lengths, or feature characteristics. This directly affects the load-bearing claim that the observed rankings (DrCIF and FreshPRINCE as the only significant outperformers of rotation forest) generalize.
  2. [Experiments and Results section] Experiments and Results section: The manuscript omits the full experimental protocol, including data splits, the exact statistical tests (e.g., name of test, p-value threshold, multiple-comparison correction), and any pre-specification details. Without these, the abstract's significance claims cannot be verified and post-hoc selection cannot be ruled out.
minor comments (1)
  1. [Abstract] Abstract: The phrasing 'these two proposals (DrCIF and FreshPRINCE) models are the only ones' is grammatically awkward; rephrase for clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight opportunities to improve transparency. We address each major comment below and will revise the manuscript to incorporate the requested details.

read point-by-point responses

  1. Referee: [TSER Archive expansion section] No description is provided of how the 44 new problems were sourced, selected, or validated for diversity in domains, lengths, or feature characteristics. This directly affects the load-bearing claim that the observed rankings (DrCIF and FreshPRINCE as the only significant outperformers of rotation forest) generalize.

    Authors: We acknowledge the omission and will add a dedicated subsection to the TSER Archive expansion section. This will describe the sourcing of the 44 problems (primarily from the UEA/UCR time series classification archive and other public repositories), the inclusion criteria (continuous extrinsic targets not directly derived from the series), and summary statistics demonstrating diversity across domains, lengths, and characteristics. These additions will support the generalizability of the performance rankings. revision: yes

  2. Referee: [Experiments and Results section] The manuscript omits the full experimental protocol, including data splits, the exact statistical tests (e.g., name of test, p-value threshold, multiple-comparison correction), and any pre-specification details. Without these, the abstract's significance claims cannot be verified and post-hoc selection cannot be ruled out.

    Authors: We agree that the protocol details should be stated explicitly. The revised Experiments and Results section will specify the data splitting procedure (using provided train/test splits where available, otherwise 10-fold cross-validation), the Wilcoxon signed-rank test with Holm correction at p < 0.05, and confirmation that the comparisons were pre-specified according to the study design. We will also provide a link to the full code repository and results to enable independent verification. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical benchmark study

full rationale

The paper expands the TSER archive from 19 to 63 problems and performs comparative evaluation of regressors, including two new proposals (FreshPRINCE and DrCIF) adapted from classification work. All claims concern observed performance rankings on the external benchmark collection. No derivations, equations, fitted parameters renamed as predictions, or self-citation chains that reduce the central result to its own inputs appear in the provided text. The evaluation is statistically independent of any internal construction that would trigger the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The performance claims rest on the representativeness of the chosen 63 problems and on standard assumptions that the statistical significance tests correctly identify superior regressors without multiple-comparison issues.

free parameters (1)
  • model hyperparameters
    Rotation forest, FreshPRINCE, and DrCIF contain tunable parameters whose values affect reported rankings but are not enumerated in the abstract.
axioms (2)
  • domain assumption The expanded archive of 63 problems is representative of TSER tasks.
    All comparative claims depend on this collection being a fair test bed.
  • domain assumption Standard cross-validation and significance testing procedures were applied without post-hoc data selection.
    The abstract asserts significant outperformance but does not detail the protocol.

pith-pipeline@v0.9.0 · 5760 in / 1327 out tokens · 25343 ms · 2026-05-24T08:23:48.617480+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

58 extracted references · 58 canonical work pages · 1 internal anchor

  1. [1]

    The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances,

    A. Bagnall, J. Lines, A. Bostrom, J. Large, and E. Keogh, “The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances,”Data Mining and Knowledge Discovery, vol. 31, no. 3, pp. 606–660, 2017. 31https://perma.cc/37GN-BMRL 32https://perma.cc/4M79-7QY4 33https://perma.cc/8AP5-5R7R 34https://tsml-eval.read...

    work page 2017

  2. [2]

    Bake off redux: a review and experimental evaluation of recent time series classification algorithms

    M. Middlehurst, P. Sch ¨afer, and A. Bagnall, “Bake off redux: a review and experimental evaluation of recent time series classification algo- rithms,” arXiv preprint arXiv:2304.13029 , 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  3. [3]

    The great multivariate time series classification bake off: a review and experimental evaluation of recent algorithmic advances,

    A. P. Ruiz, M. Flynn, J. Large, M. Middlehurst, and A. Bagnall, “The great multivariate time series classification bake off: a review and experimental evaluation of recent algorithmic advances,” Data Mining and Knowledge Discovery , vol. 35, no. 2, pp. 401–449, 2021

    work page 2021

  4. [4]

    Makridakis, S

    S. Makridakis, S. C. Wheelwright, and R. J. Hyndman, Forecasting methods and applications . John wiley & sons, 2008

    work page 2008

  5. [5]

    Time series extrinsic regression,

    C. W. Tan, C. Bergmeir, F. Petitjean, and G. I. Webb, “Time series extrinsic regression,” Data Mining and Knowledge Discovery , vol. 35, no. 3, pp. 1032–1060, 2021

    work page 2021

  6. [6]

    Rocket: exceptionally fast and accurate time series classification using random convolutional kernels,

    A. Dempster, F. Petitjean, and G. I. Webb, “Rocket: exceptionally fast and accurate time series classification using random convolutional kernels,” Data Mining and Knowledge Discovery , vol. 34, no. 5, pp. 1454–1495, 2020

    work page 2020

  7. [7]

    Inceptiontime: Finding alexnet for time series classification,

    H. Ismail Fawaz, B. Lucas, G. Forestier, C. Pelletier, D. F. Schmidt, J. Weber, G. I. Webb, L. Idoumghar, P.-A. Muller, and F. Petitjean, “Inceptiontime: Finding alexnet for time series classification,” Data Mining and Knowledge Discovery , vol. 34, no. 6, pp. 1936–1962, 2020

    work page 1936

  8. [8]

    Multirocket: 18 multiple pooling operators and transformations for fast and effective time series classification,

    C. W. Tan, A. Dempster, C. Bergmeir, and G. I. Webb, “Multirocket: 18 multiple pooling operators and transformations for fast and effective time series classification,” Data Mining and Knowledge Discovery , vol. 36, no. 5, pp. 1623–1646, 2022

    work page 2022

  9. [9]

    The freshprince: A simple transforma- tion based pipeline time series classifier,

    M. Middlehurst and A. Bagnall, “The freshprince: A simple transforma- tion based pipeline time series classifier,” in International Conference on Pattern Recognition and Artificial Intelligence . Springer, 2022, pp. 150–161

    work page 2022

  10. [11]

    Rotation forest: A new classifier ensemble method,

    J. J. Rodriguez, L. I. Kuncheva, and C. J. Alonso, “Rotation forest: A new classifier ensemble method,” IEEE transactions on pattern analysis and machine intelligence , vol. 28, no. 10, pp. 1619–1630, 2006

    work page 2006

  11. [12]

    A time series forest for classification and feature extraction,

    H. Deng, G. Runger, E. Tuv, and M. Vladimir, “A time series forest for classification and feature extraction,”Information Sciences, vol. 239, pp. 142–153, 2013

    work page 2013

  12. [13]

    The canonical interval forest (cif) classifier for time series classification,

    M. Middlehurst, J. Large, and A. Bagnall, “The canonical interval forest (cif) classifier for time series classification,” in2020 IEEE international conference on big data (big data) . IEEE, 2020, pp. 188–195

    work page 2020

  13. [14]

    Hive-cote 2.0: a new meta ensemble for time series classification,

    M. Middlehurst, J. Large, M. Flynn, J. Lines, A. Bostrom, and A. Bag- nall, “Hive-cote 2.0: a new meta ensemble for time series classification,” Machine Learning, vol. 110, no. 11, pp. 3211–3243, 2021

    work page 2021

  14. [15]

    catch22: Canonical time-series characteristics,

    C. H. Lubba, S. S. Sethi, P. Knaute, S. R. Schultz, B. D. Fulcher, and N. S. Jones, “catch22: Canonical time-series characteristics,” Data Mining and Knowledge Discovery , vol. 33, no. 6, pp. 1821–1852, 2019

    work page 2019

  15. [16]

    Estimator selection and combination in scalar-on-function regression,

    J. Goldsmith and F. Scheipl, “Estimator selection and combination in scalar-on-function regression,” Computational Statistics & Data Analy- sis, vol. 70, pp. 362–372, 2014

    work page 2014

  16. [17]

    Methods for scalar-on-function regression,

    P. T. Reiss, J. Goldsmith, H. L. Shang, and R. T. Ogden, “Methods for scalar-on-function regression,” International Statistical Review , vol. 85, no. 2, pp. 228–249, 2017

    work page 2017

  17. [18]

    Support vector regression machines,

    H. Drucker, C. J. Burges, L. Kaufman, A. Smola, and V . Vapnik, “Support vector regression machines,” Advances in neural information processing systems , vol. 9, 1996. [Online]. Available: https://proceedings.neurips.cc/paper/1996/file/ d38901788c533e8286cb6400b40b386d-Paper.pdf

    work page 1996

  18. [19]

    Random forests,

    L. Breiman, “Random forests,” Machine learning , vol. 45, no. 1, pp. 5–32, 2001

    work page 2001

  19. [20]

    Xgboost: A scalable tree boosting system,

    T. Chen and C. Guestrin, “Xgboost: A scalable tree boosting system,” in Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining , 2016, pp. 785–794

    work page 2016

  20. [21]

    Time series classification from scratch with deep neural networks: A strong baseline,

    Z. Wang, W. Yan, and T. Oates, “Time series classification from scratch with deep neural networks: A strong baseline,” in Proceedings of the IEEE international joint conference on neural networks, 2017, pp. 1578– 1585

    work page 2017

  21. [22]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

    work page 2016

  22. [23]

    Combining recurrent, convolutional, and continuous-time models with linear state space layers,

    A. Gu, I. Johnson, K. Goel, K. Saab, T. Dao, A. Rudra, and C. R ´e, “Combining recurrent, convolutional, and continuous-time models with linear state space layers,” in Advances in Neural Information Processing Systems, M. Ranzato, A. Beygelzimer, Y . Dauphin, P. Liang, and J. W. Vaughan, Eds., vol. 34. Curran Associates, Inc., 2021, pp. 572–585. [Online]....

    work page 2021

  23. [24]

    Liquid structural state-space models,

    R. Hasani, M. Lechner, T.-H. Wang, M. Chahine, A. Amini, and D. Rus, “Liquid structural state-space models,” 2022

    work page 2022

  24. [25]

    Graph neural networks for multivariate time series regression with application to seismic data,

    S. Bloemheuvel, J. van den Hoogen, D. Jozinovi ´c, A. Michelini, and M. Atzmueller, “Graph neural networks for multivariate time series regression with application to seismic data,” International Journal of Data Science and Analytics , 2022

    work page 2022

  25. [26]

    Septor: Seismic depth estimation using hierarchical neural networks,

    M. A. Siddiquee, V . M. A. Souza, G. E. Baker, and A. Mueen, “Septor: Seismic depth estimation using hierarchical neural networks,” in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2022, p. 3889–3897

    work page 2022

  26. [27]

    Essays on machine learning methods in economics,

    M. Bayani, “Essays on machine learning methods in economics,” Ph.D. dissertation, City University of New York, 2022. [Online]. Available: https://academicworks.cuny.edu/cgi/viewcontent.cgi?article= 6069&context=gc etds

    work page 2022

  27. [28]

    Generalizing DTW to the multi-dimensional case requires an adaptive approach,

    M. Shokoohi-Yekta, B. Hu, H. Jin, J. Wang, and E. Keogh, “Generalizing DTW to the multi-dimensional case requires an adaptive approach,”Data Mining and Knowledge Discovery , vol. 31, no. 1, pp. 1–31, 2017

    work page 2017

  28. [29]

    MultiRocket: multiple pooling operators and transformations for fast and effective time series classification,

    C. W. Tan, A. Dempster, C. Bergmeir, and G. Webb, “MultiRocket: multiple pooling operators and transformations for fast and effective time series classification,” Data Mining and Knowledge Discovery , vol. 36, p. 1623–1646, 2022

    work page 2022

  29. [30]

    Deep learning for time series classification: a review,

    H. Ismail Fawaz, G. Forestier, J. Weber, L. Idoumghar, and P.-A. Muller, “Deep learning for time series classification: a review,”Data Mining and Knowledge Discovery, vol. 33, no. 4, pp. 917–963, 2019

    work page 2019

  30. [31]

    Convolutional neural networks for time series classification,

    B. Zhao, H. Lu, S. Chen, J. Liu, and D. Wu, “Convolutional neural networks for time series classification,” Journal of Systems Engineering and Electronics, vol. 28, no. 1, pp. 162–169, 2017

    work page 2017

  31. [32]

    Binary shapelet transform for multiclass time series classification,

    A. Bostrom and A. Bagnall, “Binary shapelet transform for multiclass time series classification,”Transactions on Large-Scale Data and Knowl- edge Centered Systems , vol. 32, pp. 24–46, 2017

    work page 2017

  32. [33]

    Time series shapelets: a novel technique that allows accurate, interpretable and fast classification,

    L. Ye and E. Keogh, “Time series shapelets: a novel technique that allows accurate, interpretable and fast classification,” Data Mining and Knowledge Discovery, vol. 22, no. 1-2, pp. 149–182, 2011

    work page 2011

  33. [34]

    The BOSS is concerned with time series classification in the presence of noise,

    P. Sch ¨afer, “The BOSS is concerned with time series classification in the presence of noise,” Data Mining and Knowledge Discovery , vol. 29, no. 6, pp. 1505–1530, 2015

    work page 2015

  34. [35]

    Weasel 2.0–a random dilated dictionary transform for fast, accurate and memory constrained time series classi- fication,

    P. Sch ¨afer and U. Leser, “Weasel 2.0–a random dilated dictionary transform for fast, accurate and memory constrained time series classi- fication,”arXiv preprint arXiv:2301.10194 , 2023

    work page arXiv 2023

  35. [36]

    The temporal dictionary ensemble (TDE) classifier for time series classification,

    M. Middlehurst, J. Large, G. Cawley, and A. Bagnall, “The temporal dictionary ensemble (TDE) classifier for time series classification,” in Proceedings of European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases , ser. Lecture Notes in Computer Science, vol. 12457, 2020, pp. 660–676

    work page 2020

  36. [37]

    Time series feature extraction on basis of scalable hypothesis tests (tsfresh–a python package),

    M. Christ, N. Braun, J. Neuffer, and A. W. Kempa-Liehr, “Time series feature extraction on basis of scalable hypothesis tests (tsfresh–a python package),” Neurocomputing, vol. 307, pp. 72–77, 2018

    work page 2018

  37. [38]

    Is rotation forest the best classifier for problems with continuous features?

    A. Bagnall, M. Flynn, J. Large, J. Line, A. Bostrom, and G. Cawley, “Is rotation forest the best classifier for problems with continuous features?” arXiv preprint arXiv:1809.06705 , 2018

    work page arXiv 2018

  38. [39]

    Ro- tation forests for regression,

    C. Pardo, J. F. Diez-Pastor, C. Garc ´ıa-Osorio, and J. J. Rodr ´ıguez, “Ro- tation forests for regression,” Applied Mathematics and Computation , vol. 219, no. 19, pp. 9914–9924, 2013

    work page 2013

  39. [40]

    Classification and regression trees,

    L. Breiman, J. Friedman, R. Olshen, and C. Stone, “Classification and regression trees,” 1984

    work page 1984

  40. [41]

    Fast and accurate time series classification through supervised interval search,

    N. Cabello, E. Naghizade, J. Qi, and L. Kulik, “Fast and accurate time series classification through supervised interval search,” in 2020 IEEE International Conference on Data Mining (ICDM) . IEEE, 2020, pp. 948–953

    work page 2020

  41. [42]

    The contract random interval spectral ensemble (c-rise): the effect of contracting a classifier on accuracy,

    M. Flynn, J. Large, and A. Bagnall, “The contract random interval spectral ensemble (c-rise): the effect of contracting a classifier on accuracy,” in International Conference on Hybrid Artificial Intelligence Systems. Springer, 2019, pp. 381–392

    work page 2019

  42. [43]

    Derivative dynamic time warping,

    E. J. Keogh and M. J. Pazzani, “Derivative dynamic time warping,” in Proceedings of the 2001 SIAM international conference on data mining . SIAM, 2001, pp. 1–11

    work page 2001

  43. [44]

    hctsa: A computational framework for automated time-series phenotyping using massive feature extraction,

    B. D. Fulcher and N. S. Jones, “hctsa: A computational framework for automated time-series phenotyping using massive feature extraction,” Cell Systems, vol. 5, no. 5, pp. 527–531, 2017

    work page 2017

  44. [45]

    Accurate occupancy detection of an office room from light, temperature, humidity and CO 2 measurements using statistical learning models,

    L. M. Candanedo and V . Feldheim, “Accurate occupancy detection of an office room from light, temperature, humidity and CO 2 measurements using statistical learning models,” Energy and Buildings , vol. 112, pp. 28–39, 2016

    work page 2016

  45. [46]

    Comparison of machine learning al- gorithms for the power consumption prediction:-case study of tetouan city–,

    A. Salam and A. El Hibaoui, “Comparison of machine learning al- gorithms for the power consumption prediction:-case study of tetouan city–,” in 2018 6th International Renewable and Sustainable Energy Conference (IRSEC). IEEE, 2018, pp. 1–5

    work page 2018

  46. [47]

    Assessing beijing’s pm2. 5 pollution: severity, weather impact, apec and winter heating,

    X. Liang, T. Zou, B. Guo, S. Li, H. Zhang, S. Zhang, H. Huang, and S. X. Chen, “Assessing beijing’s pm2. 5 pollution: severity, weather impact, apec and winter heating,” Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences , vol. 471, no. 2182, p. 20150257, 2015

    work page 2015

  47. [48]

    Predicting car park occupancy rates in smart cities,

    D. H. Stolfi, E. Alba, and X. Yao, “Predicting car park occupancy rates in smart cities,” in Smart Cities: Second International Conference, Smart- CT 2017. Springer, 2017, pp. 107–117

    work page 2017

  48. [49]

    Snowpack, precipitation, and temper- ature measurements at the Central Sierra Snow Laboratory for water years 1971 to 2019,

    R. Osterhuber and A. Schwartz, “Snowpack, precipitation, and temper- ature measurements at the Central Sierra Snow Laboratory for water years 1971 to 2019,” 2021, dryad

    work page 1971

  49. [50]

    Online decorrelation of humidity and temperature in chemical sensors for continuous monitoring,

    R. Huerta, T. Mosqueiro, J. Fonollosa, N. F. Rulkov, and I. Rodriguez- Lujan, “Online decorrelation of humidity and temperature in chemical sensors for continuous monitoring,” Chemometrics and Intelligent Lab- oratory Systems, vol. 157, pp. 169–176, 2016

    work page 2016

  50. [51]

    Bioinspired early detection through gas flow modulation in chemo-sensory systems,

    A. Ziyatdinov, J. Fonollosa, L. Fern ´andez, A. Gutierrez-G ´alvez, S. Marco, and A. Perera, “Bioinspired early detection through gas flow modulation in chemo-sensory systems,” Sensors and Actuators B: Chemical, vol. 206, pp. 538–547, 2015

    work page 2015

  51. [52]

    Learning to 19 detect heavy drinking episodes using smartphone accelerometer data

    J. A. Killian, K. M. Passino, A. Nandi, D. R. Madden, J. D. Clapp, N. Wiratunga, F. Coenen, and S. Sani, “Learning to 19 detect heavy drinking episodes using smartphone accelerometer data.” in KHD@ IJCAI , 2019, pp. 35–42. [Online]. Available: https://ceur-ws.org/V ol-2429/paper6.pdf

    work page 2019

  52. [53]

    Covid-19 contagion forecast- ing framework based on curve decomposition and evolutionary artificial neural networks: A case study in andalusia, spain,

    M. D ´ıaz-Lozano, D. Guijo-Rubio, P. A. Guti ´errez, A. M. G ´omez- Orellana, I. T ´u˜nez, L. Ortigosa-Moreno, A. Romanos-Rodr ´ıguez, J. Padillo-Ruiz, and C. Herv ´as-Mart´ınez, “Covid-19 contagion forecast- ing framework based on curve decomposition and evolutionary artificial neural networks: A case study in andalusia, spain,” Expert Systems with Applic...

    work page 2022

  53. [54]

    Natural gas prices with Twitter sentiment scores,

    R. Ghosh, “Natural gas prices with Twitter sentiment scores,” 2022, kaggle

    work page 2022

  54. [55]

    Statistical comparisons of classifiers over multiple data sets,

    J. Dem ˇsar, “Statistical comparisons of classifiers over multiple data sets,” Journal of Machine Learning Research , vol. 7, pp. 1–30, 2006

    work page 2006

  55. [56]

    An extension on “statistical comparisons of classifiers over multiple data sets

    S. Garc ´ıa and F. Herrera, “An extension on “statistical comparisons of classifiers over multiple data sets” for all pairwise comparisons,” Journal of Machine Learning Research , vol. 9, pp. 2677–2694, 2008. [Online]. Available: https://www.jmlr.org/papers/v9/garcia08a.html

    work page 2008

  56. [57]

    Should we really use post-hoc tests based on mean-ranks?

    A. Benavoli, G. Corani, and F. Mangili, “Should we really use post-hoc tests based on mean-ranks?” Journal of Machine Learning Research , vol. 17, pp. 1–10, 2016. [Online]. Available: https://jmlr.org/papers/v17/benavoli16a.html

    work page 2016

  57. [58]

    A new family of power transformations to improve normality or symmetry,

    I. Yeo and R. A. Johnson, “A new family of power transformations to improve normality or symmetry,” Biometrika, vol. 87, no. 4, pp. 954– 959, 2000

    work page 2000

  58. [59]

    Estimating electric motor temperatures with deep residual machine learning,

    W. Kirchg ¨assner, O. Wallscheid, and J. B ¨ocker, “Estimating electric motor temperatures with deep residual machine learning,” IEEE Trans- actions on Power Electronics , vol. 36, no. 7, pp. 7480–7488, 2021. David Guijo-Rubio received the BSc degree in Computer Science in 2016 and the MSc degree in Artificial Intelligence in 2017. In 2021, he received the...

    work page 2021