pith. sign in

arxiv: 2606.07068 · v1 · pith:CQLSAIU4new · submitted 2026-06-05 · 💻 cs.LG

Bias in Filter Feature Selection Evaluation: A Meta-Analysis of Datasets, Baselines, and Experimental Design Choices

Malick Ebiele , Malika Bendechache , Rob Brennan This is my paper

Pith reviewed 2026-06-27 22:57 UTC · model grok-4.3

classification 💻 cs.LG
keywords filter feature selectionmeta-analysisevaluation biaswin ratelinear regressionexperimental designmachine learningfeature selection
0
0 comments X

The pith

The number of datasets, baselines, and new methods explains 33% of variance in new filter feature selection method win rates.

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

This paper performs a meta-analysis on 28 prominent filter feature selection studies spanning 1994 to 2025. It applies multivariate linear regression to link experimental design factors to the win rate of proposed methods over baselines. The analysis reveals that the counts of datasets, baselines, and new methods together account for 33 percent of the variance in those win rates. Readers should care because this points to systematic influences on reported performance that could reflect evaluation bias rather than true methodological advances. The work concludes with recommendations for improved evaluation practices in the field.

Core claim

The central claim is that multivariate linear regression using the number of datasets, the number of baselines, and the number of new methods as independent variables yields an R² value of 0.33 when predicting the win rate of new methods against the chosen baselines in filter feature selection studies. This means these three factors explain one-third of the observed performance differences, suggesting that evaluation outcomes are partly driven by the scale and composition of the experimental setup.

What carries the argument

Multivariate linear regression model applied to win rate as the dependent variable, with predictors being the counts of datasets, baselines, and new methods from the meta-analyzed studies.

Load-bearing premise

The sample of 28 high profile FFS studies is representative of the broader literature on filter feature selection, and the win rate metric is a valid proxy for detecting evaluation bias independent of other confounding variables.

What would settle it

Conducting the same regression analysis on a different sample of FFS papers and finding an R² value close to zero or substantially higher than 0.33.

read the original abstract

Background: Since 1990 many feature selection methods have been proposed across heterogeneous applications. To validate the usefulness of a new method, it needs to be compared against at least one baseline method from the existing literature on a feature selection task using at least one dataset. Recent developments in tabular Deep Learning (DL) and data valuation in Machine Learning (ML) suggest that the evaluation of new methods, algorithms, and models may be consciously or unconsciously biased. We hypothesise that a similar trend exists in feature selection (FS), particularly in filter feature selection (FFS). The aim of this study is therefore to examine FFS studies to identify factors that influence the evaluation and that might consist entry point for biases in order to recommend stronger principles for FFS evaluation. Methods: We analyse a sample of 28 high profile FFS studies published between 1994 and 2025. The analysis provides reflections on how to examine FFS studies, highlights lessons learned throughout the process, and gives five evidence-based recommendations for future FFS evaluation. Results: Multivariate Linear Regression analysis achieved a score of $R^2=0.33$. It means that 33% of the variance in the performance of new methods against chosen baselines (win rate) is explained by the number of datasets (#Datasets), the number of baselines (#Baselines), and the number of new methods (#NewMethods). Discussion: $R^2=0.33$ is considered medium explanation; which is promising given that this is the first such study. The medium explanation result is due to the fact that win rate is influenced by additional factors such as the maturity of the feature selection domain, the type of datasets and baselines, and the simplicity of the regression model used to explain the relationship.

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

4 major / 2 minor

Summary. The paper conducts a meta-analysis of 28 high-profile filter feature selection (FFS) studies published between 1994 and 2025. It defines a 'win rate' metric capturing how often new methods outperform chosen baselines, then fits a multivariate linear regression showing that the counts of datasets (#Datasets), baselines (#Baselines), and new methods (#NewMethods) together explain 33% of the variance in win rates (R²=0.33). The authors interpret this as evidence of evaluation bias and distill five evidence-based recommendations for future FFS studies.

Significance. If the regression result and its interpretation hold after methodological clarification, the work would be the first quantitative meta-analysis linking experimental design counts to reported performance advantages in FFS. An R² of 0.33 constitutes a medium effect size for a first study and could motivate the community to adopt more controlled evaluation protocols, reducing the risk that new-method claims are inflated by the number of datasets or baselines chosen. The explicit recommendations add immediate practical utility.

major comments (4)
  1. [Methods] Methods: the selection criteria and search strategy used to identify the 28 high-profile FFS studies are not described, so it is impossible to judge whether the sample is representative or whether the regression result generalizes beyond the chosen high-profile subset.
  2. [Results] Results: the precise definition and computation of the win-rate dependent variable (how a 'win' is scored when a new method is compared to multiple baselines across multiple datasets, how ties or statistical significance are handled) are not provided, yet this variable is the sole outcome in the regression whose R²=0.33 underpins the central claim.
  3. [Results] Results: the multivariate regression reports only R²=0.33 with no accompanying coefficient table, standard errors, p-values, multicollinearity diagnostics, or residual checks, leaving the claimed explanatory power of #Datasets, #Baselines, and #NewMethods unsupported by the usual statistical evidence.
  4. [Discussion] Discussion: the potential circularity that win rates are derived directly from the same 28 studies whose design choices are the predictors is not examined, so it remains unclear whether the regression captures an independent bias signal or merely reflects the studies' own experimental decisions.
minor comments (2)
  1. [Abstract] Abstract: the date range '1994 and 2025' should be clarified (does it include in-press papers or is 2025 a typographical error for 2024?).
  2. [Discussion] Discussion: the statement that R²=0.33 represents a 'medium explanation' would benefit from an explicit reference to conventional effect-size benchmarks (e.g., Cohen, 1988) for context.

Simulated Author's Rebuttal

4 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight important areas for clarification and strengthening. We address each major comment below, indicating planned revisions where appropriate. The core finding of R²=0.33 remains supported by the data, but we will enhance transparency and statistical reporting.

read point-by-point responses

  1. Referee: [Methods] Methods: the selection criteria and search strategy used to identify the 28 high-profile FFS studies are not described, so it is impossible to judge whether the sample is representative or whether the regression result generalizes beyond the chosen high-profile subset.

    Authors: We agree that explicit documentation of the search strategy and selection criteria is necessary for evaluating representativeness. The manuscript refers to a 'sample of 28 high profile FFS studies' but does not detail the process. In the revision we will add a Methods subsection describing the literature search (databases, keywords, time range 1994-2025), inclusion criteria (high-profile status defined by citation count and venue), and any exclusion rules applied. This will allow readers to assess generalizability. revision: yes

  2. Referee: [Results] Results: the precise definition and computation of the win-rate dependent variable (how a 'win' is scored when a new method is compared to multiple baselines across multiple datasets, how ties or statistical significance are handled) are not provided, yet this variable is the sole outcome in the regression whose R²=0.33 underpins the central claim.

    Authors: We acknowledge that the win-rate definition requires precise specification. A 'win' is scored per pairwise comparison (new method vs. baseline on a dataset) when the new method reports strictly higher performance; ties are counted as non-wins; statistical significance is not required because most source papers do not report it. The overall win rate for a study is the proportion of such wins across all reported comparisons. We will add this formal definition, including the aggregation formula, to the Results section in the revision. revision: yes

  3. Referee: [Results] Results: the multivariate regression reports only R²=0.33 with no accompanying coefficient table, standard errors, p-values, multicollinearity diagnostics, or residual checks, leaving the claimed explanatory power of #Datasets, #Baselines, and #NewMethods unsupported by the usual statistical evidence.

    Authors: The manuscript reports only the aggregate R² because the primary claim concerns the proportion of variance explained. However, we agree that full regression diagnostics strengthen the result. In the revision we will include a coefficient table with estimates, standard errors, p-values, variance inflation factors (VIF) for multicollinearity, and a brief note on residual normality. These will be added without altering the R² value or interpretation. revision: yes

  4. Referee: [Discussion] Discussion: the potential circularity that win rates are derived directly from the same 28 studies whose design choices are the predictors is not examined, so it remains unclear whether the regression captures an independent bias signal or merely reflects the studies' own experimental decisions.

    Authors: We will expand the Discussion to address this concern explicitly. The regression is not circular in the statistical sense: the predictors (#Datasets, #Baselines, #NewMethods) are factual counts extracted from each paper, while the outcome (win rate) is a derived performance metric computed from the same papers' reported results. The model quantifies the association between design choices and reported success rates. We will clarify that this measures a form of selection or reporting bias rather than an independent external signal, and note that future work could test the relationship on held-out studies. revision: yes

Circularity Check

0 steps flagged

No significant circularity; meta-regression is self-contained

full rationale

The paper's central result is an R²=0.33 from multivariate linear regression of win rate (computed per study from reported comparisons of new methods vs. baselines) on three study-level count variables (#Datasets, #Baselines, #NewMethods) across the same 28 studies. This is a standard meta-regression that tests association between design choices and observed performance ratios; the win rate metric is defined directly from each study's reported outcomes and is not constructed from the regression coefficients or any fitted parameter. No equations reduce a claimed prediction to its inputs by definition, no self-citation chain supports a uniqueness claim, and no ansatz or renaming is invoked. The analysis therefore remains independent of its own fitted values.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The result depends on the fitted regression parameters and the assumption that the sampled studies allow generalization to FFS evaluation practices.

free parameters (1)
  • beta_coefficients_for_predictors
    Coefficients in the multivariate linear regression model fitted to explain win rate using #Datasets, #Baselines, and #NewMethods.
axioms (2)
  • domain assumption The relationship between the number of datasets, baselines, new methods and win rate is linear
    Invoked by the use of multivariate linear regression.
  • domain assumption The 28 studies provide independent observations for the regression
    Required for standard regression inference.

pith-pipeline@v0.9.1-grok · 5866 in / 1253 out tokens · 28551 ms · 2026-06-27T22:57:24.219375+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

45 extracted references · 30 canonical work pages

  1. [1]

    Knowledge and Information Systems 66(3), 1575–1637 (2024) https://doi.org/10.1007/s10115-023-02010-5 16 Under-review

    Theng, D., Bhoyar, K.K.: Feature selection techniques for machine learning: a survey of more than two decades of research. Knowledge and Information Systems 66(3), 1575–1637 (2024) https://doi.org/10.1007/s10115-023-02010-5 16 Under-review

    work page doi:10.1007/s10115-023-02010-5 2024

  2. [2]

    Journal of machine learning research3(Mar), 1157–1182 (2003)

    Guyon, I., Elisseeff, A.: An introduction to variable and feature selection. Journal of machine learning research3(Mar), 1157–1182 (2003)

    2003

  3. [3]

    IEEE Transactions on Evolutionary Computation, 1–1 (2023) https: //doi.org/10.1109/TEVC.2023.3292527

    Jiao, R., Nguyen, B.H., Xue, B., Zhang, M.: A Survey on Evolutionary Mul- tiobjective Feature Selection in Classification: Approaches, Applications, and Challenges. IEEE Transactions on Evolutionary Computation, 1–1 (2023) https: //doi.org/10.1109/TEVC.2023.3292527

    work page doi:10.1109/tevc.2023.3292527 2023

  4. [4]

    Engineering Applications of Artificial Intelligence126, 107136 (2023) https://doi.org/10.1016/j.engappai.2023.107136

    Moslemi, A.: A tutorial-based survey on feature selection: Recent advancements on feature selection. Engineering Applications of Artificial Intelligence126, 107136 (2023) https://doi.org/10.1016/j.engappai.2023.107136

    work page doi:10.1016/j.engappai.2023.107136 2023

  5. [5]

    Briefings in Bioinformatics23(1), 354 (2022) https://doi.org/10.1093/bib/ bbab354

    Bommert, A., Welchowski, T., Schmid, M., Rahnenf¨ uhrer, J.: Benchmark of fil- ter methods for feature selection in high-dimensional gene expression survival data. Briefings in Bioinformatics23(1), 354 (2022) https://doi.org/10.1093/bib/ bbab354

    work page doi:10.1093/bib/ 2022

  6. [6]

    In: Proceedings of the Genetic and Evolutionary Computation Conference Companion

    Mostert, W., Malan, K., Engelbrecht, A.: Filter versus wrapper feature selec- tion based on problem landscape features. In: Proceedings of the Genetic and Evolutionary Computation Conference Companion. GECCO ’18, pp. 1489–1496. Association for Computing Machinery, New York, NY, USA (2018). https://doi. org/10.1145/3205651.3208305 .https://doi.org/10.1145/...

    work page doi:10.1145/3205651.3208305 2018

  7. [7]

    Computers in Biology and Medicine119, 103667 (2020) https://doi.org/10.1016/j.compbiomed.2020

    Li, C., Luo, X., Qi, Y., Gao, Z., Lin, X.: A new feature selection algorithm based on relevance, redundancy and complementarity. Computers in Biology and Medicine119, 103667 (2020) https://doi.org/10.1016/j.compbiomed.2020. 103667

    work page doi:10.1016/j.compbiomed.2020 2020

  8. [8]

    In: Proceedings of the 18th International Conference on Neural Information Processing Systems

    Guyon, I., Gunn, S., Hur, A.B., Dror, G.: Result analysis of the NIPS 2003 fea- ture selection challenge. In: Proceedings of the 18th International Conference on Neural Information Processing Systems. NIPS’04, pp. 545–552. MIT Press, Cambridge, MA, USA (2004)

    2003

  9. [9]

    NIPS ’22, pp

    Grinsztajn, L., Oyallon, E., Varoquaux, G.: Why do tree-based models still out- perform deep learning on typical tabular data? In: Proceedings of the 36th International Conference on Neural Information Processing Systems. NIPS ’22, pp. 507–520. Curran Associates Inc., Red Hook, NY, USA (2022)

    2022

  10. [10]

    Shwartz-Ziv, R., Armon, A.: Tabular data: Deep learning is not all you need. Inf. Fusion81(C), 84–90 (2022) https://doi.org/10.1016/j.inffus.2021.11.011

    work page doi:10.1016/j.inffus.2021.11.011 2022

  11. [11]

    Advances in Neural Information Processing Systems 36(2024)

    Cherepanova, V., Levin, R., Somepalli, G., Geiping, J., Bruss, C.B., Wilson, A.G., Goldstein, T., Goldblum, M.: A performance-driven benchmark for feature selec- tion in tabular deep learning. Advances in Neural Information Processing Systems 36(2024)

    2024

  12. [12]

    IEEE Transactions on Neural Net- works and Learning Systems35(6), 7499–7519 (2024) https://doi.org/10.1109/ TNNLS.2022.3229161

    Borisov, V., Leemann, T., Seßler, K., Haug, J., Pawelczyk, M., Kasneci, G.: Deep 17 Under-review Neural Networks and Tabular Data: A Survey. IEEE Transactions on Neural Net- works and Learning Systems35(6), 7499–7519 (2024) https://doi.org/10.1109/ TNNLS.2022.3229161

    arXiv 2024

  13. [13]

    In: Proceedings of the 35th International Conference on Neural Information Processing Systems

    Gorishniy, Y., Rubachev, I., Khrulkov, V., Babenko, A.: Revisiting Deep Learning Models for Tabular Data. In: Proceedings of the 35th International Conference on Neural Information Processing Systems. NIPS’21, pp. 18932–18943. ACM, ??? (2021)

    2021

  14. [14]

    Advances in Neural Information Processing Systems36, 28624– 28647 (2023)

    Jiang, K., Liang, W., Zou, J.Y., Kwon, Y.: OpenDataVal: a Unified Benchmark for Data Valuation. Advances in Neural Information Processing Systems36, 28624– 28647 (2023)

    2023

  15. [15]

    Neural Networks, IEEE Transactions on5, 537–550 (1994) https: //doi.org/10.1109/72.298224

    Battiti, R.: Using Mutual Information for Selecting Features in Supervised Neural Net Learning. Neural Networks, IEEE Transactions on5, 537–550 (1994) https: //doi.org/10.1109/72.298224

    work page doi:10.1109/72.298224 1994

  16. [16]

    IEEE Transactions on Neural Networks13(1), 143–159 (2002) https://doi.org/10.1109/ 72.977291

    Kwak, N., Choi, C.-H.: Input feature selection for classification problems. IEEE Transactions on Neural Networks13(1), 143–159 (2002) https://doi.org/10.1109/ 72.977291

    2002

  17. [17]

    Zhang, Y., Wu, J., Cai, J.: Compact Representation for Image Classification: To Choose or to Compress? In: 2014 IEEE Conference on Computer Vision and Pattern Recognition, pp. 907–914. IEEE, Columbus, OH, USA (2014). https:// doi.org/10.1109/CVPR.2014.121 .https://ieeexplore.ieee.org/document/6909516

    work page doi:10.1109/cvpr.2014.121 2014

  18. [18]

    IEEE Access8, 62011–62031 (2020) https://doi.org/10.1109/ACCESS.2020

    Huang, M., Sun, L., Xu, J., Zhang, S.: Multilabel Feature Selection Using Relief and Minimum Redundancy Maximum Relevance Based on Neighborhood Rough Sets. IEEE Access8, 62011–62031 (2020) https://doi.org/10.1109/ACCESS.2020. 2982536

    work page doi:10.1109/access.2020 2020

  19. [19]

    Neurocomputing552, 126564 (2023) https://doi.org/10.1016/ j.neucom.2023.126564

    Yuan, G., Zhai, Y., Tang, J., Zhou, X.: CSCIM fs: Cosine similarity coefficient and information measurement criterion-based feature selection method for high- dimensional data. Neurocomputing552, 126564 (2023) https://doi.org/10.1016/ j.neucom.2023.126564

    arXiv 2023

  20. [20]

    Information Sciences 699, 121801 (2025) https://doi.org/10.1016/j.ins.2024.121801

    Yuan, G., Li, X., Qiu, P., Zhou, X.: Feature selection method based on wavelet similarity combined with maximum information coefficient. Information Sciences 699, 121801 (2025) https://doi.org/10.1016/j.ins.2024.121801

    work page doi:10.1016/j.ins.2024.121801 2025

  21. [21]

    IEEE Transactions on Image Processing33, 809–824 (2024) https://doi.org/10.1109/TIP.2023.3348992

    Yuan, L., Mei, C., Wang, W., Lu, T.: Feature Selection Based on Intrusive Outliers Rather Than All Instances. IEEE Transactions on Image Processing33, 809–824 (2024) https://doi.org/10.1109/TIP.2023.3348992

    work page doi:10.1109/tip.2023.3348992 2024

  22. [22]

    Expert Systems with Applications239, 122490 (2024) https://doi

    Ihianle, I.K., Machado, P., Owa, K., Adama, D.A., Otuka, R., Lotfi, A.: Min- imising redundancy, maximising relevance: HRV feature selection for stress classification. Expert Systems with Applications239, 122490 (2024) https://doi. 18 Under-review org/10.1016/j.eswa.2023.122490

    work page doi:10.1016/j.eswa.2023.122490 2024

  23. [23]

    In: 2019 IEEE International Conference on Data Science and Advanced Ana- lytics (DSAA), pp

    Zhao, Z., Anand, R., Wang, M.: Maximum Relevance and Minimum Redun- dancy Feature Selection Methods for a Marketing Machine Learning Platform. In: 2019 IEEE International Conference on Data Science and Advanced Ana- lytics (DSAA), pp. 442–452 (2019). https://doi.org/10.1109/DSAA.2019.00059 . https://ieeexplore.ieee.org/abstract/document/8964172

    work page doi:10.1109/dsaa.2019.00059 2019

  24. [24]

    Computers8(2), 42 (2019) https://doi.org/10.3390/ computers8020042

    Jo, I., Lee, S., Oh, S.: Improved Measures of Redundancy and Relevance for mRMR Feature Selection. Computers8(2), 42 (2019) https://doi.org/10.3390/ computers8020042 . Publisher: Multidisciplinary Digital Publishing Institute

    2019

  25. [25]

    IEEE Trans- actions on Pattern Analysis and Machine Intelligence27(8), 1226–1238 (2005) https://doi.org/10.1109/TPAMI.2005.159

    Peng, H., Long, F., Ding, C.: Feature selection based on mutual information criteria of max-dependency, max-relevance, and min-redundancy. IEEE Trans- actions on Pattern Analysis and Machine Intelligence27(8), 1226–1238 (2005) https://doi.org/10.1109/TPAMI.2005.159

    work page doi:10.1109/tpami.2005.159 2005

  26. [26]

    Science China Information Sciences63(1), 112103 (2019) https://doi.org/10.1007/s11432-019-2633-y

    Bugata, P., Drotar, P.: On some aspects of minimum redundancy maximum rele- vance feature selection. Science China Information Sciences63(1), 112103 (2019) https://doi.org/10.1007/s11432-019-2633-y

    work page doi:10.1007/s11432-019-2633-y 2019

  27. [27]

    Applied Soft Computing95, 106537 (2020) https://doi.org/10

    Zhang, P., Gao, W.: Feature selection considering Uncertainty Change Ratio of the class label. Applied Soft Computing95, 106537 (2020) https://doi.org/10. 1016/j.asoc.2020.106537

    arXiv 2020

  28. [28]

    Expert Systems with Applications42(22), 8520–8532 (2015) https://doi.org/10.1016/j.eswa.2015.07.007

    Bennasar, M., Hicks, Y., Setchi, R.: Feature selection using Joint Mutual Informa- tion Maximisation. Expert Systems with Applications42(22), 8520–8532 (2015) https://doi.org/10.1016/j.eswa.2015.07.007

    work page doi:10.1016/j.eswa.2015.07.007 2015

  29. [29]

    ACM Comput

    Li, J., Cheng, K., Wang, S., Morstatter, F., Trevino, R.P., Tang, J., Liu, H.: Feature Selection: A Data Perspective. ACM Comput. Surv.50(6), 94–19445 (2017) https://doi.org/10.1145/3136625

    work page doi:10.1145/3136625 2017

  30. [30]

    Journal of Bioinformat- ics and Computational Biology03(02), 185–205 (2005) https://doi.org/10.1142/ S0219720005001004

    Ding, C., Peng, H.: MINIMUM REDUNDANCY FEATURE SELECTION FROM MICROARRAY GENE EXPRESSION DATA. Journal of Bioinformat- ics and Computational Biology03(02), 185–205 (2005) https://doi.org/10.1142/ S0219720005001004

    2005

  31. [31]

    In: Leonardis, A., Bischof, H., Pinz, A

    Lin, D., Tang, X.: Conditional Infomax Learning: An Integrated Framework for Feature Extraction and Fusion. In: Leonardis, A., Bischof, H., Pinz, A. (eds.) Computer Vision – ECCV 2006, pp. 68–82. Springer, Berlin, Heidelberg (2006). https://doi.org/10.1007/11744023 6

    work page doi:10.1007/11744023 2006

  32. [32]

    Procedia Technology10, 20–27 (2013) https://doi.org/10.1016/j.protcy.2013.12.332 19 Under-review

    Mandal, M., Mukhopadhyay, A.: An Improved Minimum Redundancy Maximum Relevance Approach for Feature Selection in Gene Expression Data. Procedia Technology10, 20–27 (2013) https://doi.org/10.1016/j.protcy.2013.12.332 19 Under-review

    work page doi:10.1016/j.protcy.2013.12.332 2013

  33. [33]

    In: 2013 IEEE Symposium on Computational Intelligence in Bioinformatics and Compu- tational Biology (CIBCB), pp

    Yang, J., Zhu, Z., He, S., Ji, Z.: Minimal-redundancy-maximal-relevance feature selection using different relevance measures for omics data classification. In: 2013 IEEE Symposium on Computational Intelligence in Bioinformatics and Compu- tational Biology (CIBCB), pp. 246–251 (2013). https://doi.org/10.1109/CIBCB. 2013.6595417 .https://ieeexplore.ieee.org...

    work page doi:10.1109/cibcb 2013

  34. [34]

    Journal of Statistical Compu- tation and Simulation (2016)

    Berrendero, J.R., Cuevas, A., Torrecilla, J.L.: The mRMR variable selection method: a comparative study for functional data. Journal of Statistical Compu- tation and Simulation (2016). Publisher: Taylor & Francis

    2016

  35. [35]

    Dubey, V.K., Saxena, A.K.: A Cosine-Similarity Mutual-Information Approach for Feature Selection on High Dimensional Datasets. J. Inf. Technol. Res.10(1), 15–28 (2017) https://doi.org/10.4018/JITR.2017010102

    work page doi:10.4018/jitr.2017010102 2017

  36. [36]

    Pattern Recognition67, 47–61 (2017) https://doi.org/10.1016/j.patcog.2017.01.026

    Senawi, A., Wei, H.-L., Billings, S.A.: A new maximum relevance-minimum multicollinearity (MRmMC) method for feature selection and ranking. Pattern Recognition67, 47–61 (2017) https://doi.org/10.1016/j.patcog.2017.01.026

    work page doi:10.1016/j.patcog.2017.01.026 2017

  37. [37]

    International Journal of Intelligent Systems32(2), 134–152 (2017) https://doi.org/10.1002/int

    Ram´ ırez-Gallego, S., Lastra, I., Mart´ ınez-Rego, D., Bol´ on-Canedo, V., Ben´ ıtez, J.M., Herrera, F., Alonso-Betanzos, A.: Fast-mRMR: Fast Minimum Redundancy Maximum Relevance Algorithm for High-Dimensional Big Data. International Journal of Intelligent Systems32(2), 134–152 (2017) https://doi.org/10.1002/int. 21833 . eprint: https://onlinelibrary.wil...

    work page doi:10.1002/int 2017

  38. [38]

    Information Sciences409-410, 68–86 (2017) https://doi.org/10.1016/j.ins.2017.05.013

    Che, J., Yang, Y., Li, L., Bai, X., Zhang, S., Deng, C.: Maximum relevance minimum common redundancy feature selection for nonlinear data. Information Sciences409-410, 68–86 (2017) https://doi.org/10.1016/j.ins.2017.05.013

    work page doi:10.1016/j.ins.2017.05.013 2017

  39. [39]

    BMC Bioinformatics18(1), 9 (2017) https://doi.org/10.1186/s12859-016-1423-9

    Radovic, M., Ghalwash, M., Filipovic, N., Obradovic, Z.: Minimum redundancy maximum relevance feature selection approach for temporal gene expression data. BMC Bioinformatics18(1), 9 (2017) https://doi.org/10.1186/s12859-016-1423-9

    work page doi:10.1186/s12859-016-1423-9 2017

  40. [40]

    Engineering Applications of Artificial Intelligence106, 104481 (2021) https://doi.org/10.1016/j.engappai.2021.104481

    Zhang, P., Gao, W., Hu, J., Li, Y.: A conditional-weight joint relevance metric for feature relevancy term. Engineering Applications of Artificial Intelligence106, 104481 (2021) https://doi.org/10.1016/j.engappai.2021.104481

    work page doi:10.1016/j.engappai.2021.104481 2021

  41. [41]

    IEEE Transactions on Cybernetics53(2), 707–717 (2023) https://doi.org/10.1109/TCYB.2021.3139898

    Wang, Y., Li, X., Ruiz, R.: Feature Selection With Maximal Relevance and Min- imal Supervised Redundancy. IEEE Transactions on Cybernetics53(2), 707–717 (2023) https://doi.org/10.1109/TCYB.2021.3139898

    work page doi:10.1109/tcyb.2021.3139898 2023

  42. [42]

    Strobl, A.-L

    Strobl, C., Boulesteix, A.-L., Zeileis, A., Hothorn, T.: Bias in random for- est variable importance measures: Illustrations, sources and a solution. BMC Bioinformatics8(1), 25 (2007) https://doi.org/10.1186/1471-2105-8-25

    work page doi:10.1186/1471-2105-8-25 2007

  43. [43]

    In: Holzinger, A., Kieseberg, P., Tjoa, A.M., Weippl, E

    Loecher, M.: Debiasing MDI Feature Importance and SHAP Values in Tree Ensembles. In: Holzinger, A., Kieseberg, P., Tjoa, A.M., Weippl, E. (eds.) Machine Learning and Knowledge Extraction. Lecture Notes in Computer Science, pp. 20 Under-review 114–129. Springer, Cham (2022). https://doi.org/10.1007/978-3-031-14463-9 8

    work page doi:10.1007/978-3-031-14463-9 2022

  44. [44]

    Communications in Computer and Infor- mation Science, pp

    Baudeu, R., Wright, M.N., Loecher, M.: Are SHAP Values Biased Towards High-Entropy Features? In: Machine Learning and Principles and Practice of Knowledge Discovery In Databases. Communications in Computer and Infor- mation Science, pp. 418–433. Springer, Cham (2023). https://doi.org/10.1007/ 978-3-031-23618-1 28

    2023

  45. [45]

    In: Fortino, G., Kumar, A., Swaroop, A., Shukla, P

    Ebiele, M., Bendechache, M., Ward, M., Geary, U., Byrne, D., Creagh, D., Brennan, R.: The Impact of Data Valuation on Feature Importance in Clas- sification Models. In: Fortino, G., Kumar, A., Swaroop, A., Shukla, P. (eds.) Proceedings of Third International Conference on Computing and Communica- tion Networks, pp. 601–617. Springer, Singapore (2024). htt...

    2024