Abstract
AbstractFeature selection is popular for obtaining small, interpretable, yet highly accurate prediction models. Conventional feature-selection methods typically yield one feature set only, which does not suffice in certain scenarios. For example, users might be interested in finding alternative feature sets with similar prediction quality, offering different explanations of the data. In this article, we introduce alternative feature selection and formalize it as an optimization problem. In particular, we define alternatives via constraints and enable users to control the number and dissimilarity of alternatives. Next, we analyze the complexity of this optimization problem and show $$\mathcal{N}\mathcal{P}$$
N
P
-hardness. Further, we discuss how to integrate conventional feature-selection methods as objectives. Finally, we evaluate alternative feature selection in comprehensive experiments with 30 datasets representing binary-classification problems. We observe that alternative feature sets may indeed have high prediction quality, and we analyze factors influencing this outcome.
Funder
Karlsruher Institut für Technologie (KIT)
Publisher
Springer Science and Business Media LLC
Reference84 articles.
1. Artelt, A., Hammer, B.: “Even if ...”—diverse semifactual explanations of reject (2022). arXiv:2207.01898 [cs.LG]
2. Bach, J., Zoller, K., Trittenbach, H., et al.: An empirical evaluation of constrained feature selection. SN Comput. Sci. 3(6) (2022). https://doi.org/10.1007/s42979-022-01338-z
3. Bach, J.: Finding optimal diverse feature sets with alternative feature selection (2023). arXiv:2307.11607v1 [cs.LG]
4. Bailey, J.: Alternative clustering analysis: a review. In: Data Clustering: Algorithms and Applications, 1st edn. CRC Press, chap 21, pp. 535–550 (2014) https://doi.org/10.1201/9781315373515
5. Bestuzheva, K., Besançon, M., Chen, W.K., et al.: The SCIP Optimization Suite 8.0. Tech. rep., Zuse Institute Berlin, Germany (2021) http://nbn-resolving.de/urn:nbn:de:0297-zib-85309
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
1. Correction to: Alternative feature selection with user control;International Journal of Data Science and Analytics;2024-05-14