Dokument: Modelling Human Uncertainty in Predictive Data Mining: Development of a Neuro-Stochastic Model to Describe Unreliable User Feedback and its Impact on User-Adaptive Information Systems
| Titel: | Modelling Human Uncertainty in Predictive Data Mining: Development of a Neuro-Stochastic Model to Describe Unreliable User Feedback and its Impact on User-Adaptive Information Systems | |||||||
| URL für Lesezeichen: | https://docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=55688 | |||||||
| URN (NBN): | urn:nbn:de:hbz:061-20210322-131212-3 | |||||||
| Kollektion: | Dissertationen | |||||||
| Sprache: | Englisch | |||||||
| Dokumententyp: | Wissenschaftliche Abschlussarbeiten » Dissertation | |||||||
| Medientyp: | Text | |||||||
| Autor: | Jasberg, Kevin [Autor] | |||||||
| Dateien: |
| |||||||
| Beitragende: | Dr. Dr. Sizov, Sergej [Gutachter] Jun.-Prof. Petersen, Wiebke [Gutachter] | |||||||
| Stichwörter: | Human Uncertainty, Data Mining, Machine Learning, Recommender Systems, Probabilistic Population Codes, Netflix Prize, Empathy for Human Nature | |||||||
| Dewey Dezimal-Klassifikation: | 000 Informatik, Informationswissenschaft, allgemeine Werke » 020 Bibliotheks- und Informationswissenschaften | |||||||
| Beschreibung: | The present dissertation was developed within the scope of ``Systems with Empathy for the Human Nature'', i.e.\ systems that explain individual human behaviour along with its peculiarities rather than reflecting aggregated group data. This thesis, in particular, focuses on the phenomenon of unreliable user feedback (human uncertainty) in the context of recommendation and personalisation.
At the beginning, the concept of measurement uncertainty is used to render human uncertainty measurable and to analyse its impact on the comparative assessment of predictive systems. The findings reveal a difficulty to distinguish between systems regarding a given accuracy metric. Furthermore, human uncertainty is shown to induce an offset on such metrics, which limits the detection of improvements. This furnishes the need for a mathematical model of unreliable decision-making, feasible test procedures for significant system distinction, and a user model to plausibly explain the present phenomenon. To this end, concepts of statistics will be combined with those of metrology and theoretical neuroscience. Using human uncertainty as an example, it is illustrated how systems with empathy for the human nature can be designed. The knowledge gained in this dissertation comprises a technical and an epistemological component. On the one hand, a specific characteristic of human decision-making is investigated and its origin is discussed against the background of a possible model of cognition. On the other hand, a mathematical framework is developed to analyse and implement this phenomenon for future systems of predictive data mining. This possibly paves the way for a new perspective within a currently prominent research direction. | |||||||
| Lizenz: |  Urheberrechtsschutz | |||||||
| Fachbereich / Einrichtung: | Philosophische Fakultät » Institut für Sprache und Information » Informationswissenschaft | |||||||
| Dokument erstellt am: | 22.03.2021 | |||||||
| Dateien geändert am: | 22.03.2021 | |||||||
| Promotionsantrag am: | 07.07.2020 | |||||||
| Datum der Promotion: | 02.03.2021 |
