Dokument: Micro-Level Reserving for Claim Count Data
| Titel: | Micro-Level Reserving for Claim Count Data | |||||||
| URL für Lesezeichen: | https://docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=65522 | |||||||
| URN (NBN): | urn:nbn:de:hbz:061-20240422-112043-7 | |||||||
| Kollektion: | Dissertationen | |||||||
| Sprache: | Englisch | |||||||
| Dokumententyp: | Wissenschaftliche Abschlussarbeiten » Dissertation | |||||||
| Medientyp: | Text | |||||||
| Autor: | Rosenstock, Alexander [Autor] | |||||||
| Dateien: |
| |||||||
| Beitragende: | Prof. Dr. Bücher, Axel [Gutachter] Prof. Dr. Scherer, Matthias [Gutachter] | |||||||
| Dewey Dezimal-Klassifikation: | 500 Naturwissenschaften und Mathematik » 510 Mathematik | |||||||
| Beschreibung: | Determining the ultimate loss of all claims occurred in an accident year is of primal importance
in actuarial practice. Widely used methods to determine these losses work on so-called loss triangles, contracting the available information to a tally of all payments for a particular accident year and development year combination. These triangles, while easier to work with than individual claims data, contain only a fraction of the information available to an insurer at a specific point in time, resulting in subpar predictions. With the progress of computational power and the advent of advanced analytical methods, research interest in more granular claim reserving methods has picked up in recent years, going back to Norberg [12]. For the subtask of determining the ultimate claim counts, a micro-level model with individual claims and individual contracts as the smallest unit of observation is developed. The model is used to derive a set of micro-level predictors for claim counts which are analyzed in a large scale simulation study and on a real-world general insurance dataset, showing increased out-of- sample accuracy on real-world data and increased robustness to violations of basic model assumptions on synthetic data. When analyzing observable insurance data the random truncation inherent in the observations must be taken into account. Methods to perform parameter estimation and distribution learning for arbitrary distributions in this setting are developed. Accompanying software in the form of an R package available on CRAN enables swift application of the methods. | |||||||
| Lizenz: |  Dieses Werk ist lizenziert unter einer Creative Commons Namensnennung 4.0 International Lizenz | |||||||
| Fachbereich / Einrichtung: | Mathematisch- Naturwissenschaftliche Fakultät » WE Mathematik » Mathematische Statistik und Wahrscheinlichkeitstheorie | |||||||
| Dokument erstellt am: | 22.04.2024 | |||||||
| Dateien geändert am: | 22.04.2024 | |||||||
| Promotionsantrag am: | 15.11.2023 | |||||||
| Datum der Promotion: | 11.04.2024 |
