Tree ensemble methods for ordinal prediction
| dc.contributor.advisor | Pauly, Markus | |
| dc.contributor.author | Buczak, Philip | |
| dc.contributor.referee | Doebler, Philipp | |
| dc.date.accepted | 2025-04-03 | |
| dc.date.accessioned | 2025-05-05T05:19:22Z | |
| dc.date.available | 2025-05-05T05:19:22Z | |
| dc.date.issued | 2025 | |
| dc.description.abstract | Research questions and applications in the social and life sciences often involve ordinal response data. Student performance is assessed through ordinal grades, patients may express the perceived severity of their symptoms in ordinal levels and respondents of questionnaires may voice their political views through rating given statements. As such, the prediction of ordinal responses is relevant for many fields and can help, e.g., identifying which students may benefit from educational support systems. Traditionally, ordinal responses have been modeled through parametric models such as the proportional odds model. In light of the increasing quantities of data in these fields as well as the continued proliferation of machine learning (ML) methods, recent years saw the establishment of a new methodological stream of ordinal prediction methods based on ML. These methods promise high predictive performance for settings in which traditional parametric models may face difficulties (e.g., highly non-linear effects, high-dimensional data). However, many of these ML methods were originally not specifically tailored towards ordinal responses. Therefore, several extensions and adaptations of ML methods (particularly for tree-based methods) have been proposed to take ordinality into account. A particularly promising approach based on Random Forest (RF) is Ordinal Forest (OF; Hornung, 2019) which assigns numeric scores to the ordinal response categories and uses the scores to train a regression RF. To determine suitable score choices, OF performs a prior optimization step in which scores are optimized w.r.t. their predictive performance. | en |
| dc.identifier.uri | http://hdl.handle.net/2003/43689 | |
| dc.identifier.uri | http://dx.doi.org/10.17877/DE290R-25462 | |
| dc.language.iso | en | |
| dc.subject | Ordinal prediction | en |
| dc.subject | Random forest | en |
| dc.subject | Machine learning | en |
| dc.subject | Hierarchical data | en |
| dc.subject.ddc | 310 | |
| dc.subject.rswk | Ordnungsstatistik | de |
| dc.subject.rswk | Maschinelles Lernen | de |
| dc.subject.rswk | Prognosemodell | de |
| dc.title | Tree ensemble methods for ordinal prediction | en |
| dc.type | Text | |
| dc.type.publicationtype | PhDThesis | |
| dcterms.accessRights | open access | |
| eldorado.secondarypublication | false |