{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/on-the-effectiveness-of-discretizing","title":"On the Effectiveness of Discretizing Quantitative Attributes in Linear Classifiers","arxiv_id":"1701.07114","date":"2017-01-24","proceeding":null,"authors":["Nayyar A. Zaidi","Yang Du","Geoffrey I. Webb"],"abstract":"Learning algorithms that learn linear models often have high representation\nbias on real-world problems. In this paper, we show that this representation\nbias can be greatly reduced by discretization. Discretization is a common\nprocedure in machine learning that is used to convert a quantitative attribute\ninto a qualitative one. It is often motivated by the limitation of some\nlearners to qualitative data. Discretization loses information, as fewer\ndistinctions between instances are possible using discretized data relative to\nundiscretized data. In consequence, where discretization is not essential, it\nmight appear desirable to avoid it. However, it has been shown that\ndiscretization often substantially reduces the error of the linear generative\nBayesian classifier naive Bayes. This motivates a systematic study of the\neffectiveness of discretizing quantitative attributes for other linear\nclassifiers. In this work, we study the effect of discretization on the\nperformance of linear classifiers optimizing three distinct discriminative\nobjective functions --- logistic regression (optimizing negative\nlog-likelihood), support vector classifiers (optimizing hinge loss) and a\nzero-hidden layer artificial neural network (optimizing mean-square-error). We\nshow that discretization can greatly increase the accuracy of these linear\ndiscriminative learners by reducing their representation bias, especially on\nbig datasets. We substantiate our claims with an empirical study on $42$\nbenchmark datasets.","url_abs":"http://arxiv.org/abs/1701.07114v1","url_pdf":"http://arxiv.org/pdf/1701.07114v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"on-the-effectiveness-of-discretizing","repo_url":"https://github.com/vedic-partap/Discretization","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"attribute","task_name":"Attribute"}],"methods":[{"method_slug":"logistic-regression","method_name":"Logistic Regression"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}