Forecasting in the semiconductor industry
| dc.contributor.advisor | Pauly, Markus | |
| dc.contributor.author | Steinmeister, Louis | |
| dc.contributor.referee | Ligges, Uwe | |
| dc.date.accepted | 2024-12-19 | |
| dc.date.accessioned | 2025-04-08T05:52:15Z | |
| dc.date.available | 2025-04-08T05:52:15Z | |
| dc.date.issued | 2024 | |
| dc.description.abstract | This dissertation investigates different statistical questions in the domain of semiconductor market forecasting and trend detection. The first investigation brings us into the realm of technological trend detection and technology life cycles. Common models to achieve these ends are technology diffusion models, such as the simple logistic growth model. Since this model represents a closed system view of a highly interconnected and complex phenomenon, we investigate the robustness of derived life cycle estimates with regard to an external shock, namely the COVID-19 pandemic, and its consequences on consumer demand and supply chains. While the logistic growth model takes a long-term perspective, we focus on the detection of shorter-term market dynamics with an investigation into forecasting the semiconductor market more broadly. A premier provider of semiconductor market data and industry forecasts is the World Semiconductor Trade Statistics (WSTS). As our second focus of attention, we investigate the precision of industry expert forecasts and explore if data-driven models can enhance such forecasts. Given that the WSTS product classification hierarchy has a degenerate structure, we proceed with an exploration of degenerate hierarchical forecast reconciliation to enable coherent forecasts. Lastly, we investigate the performance of a popular reconciliation algorithm, Trace Minimization (MinT), which was originally introduced by Wickramasuriya et al. (2019), in the context of very short time series through an elaborate simulation study and compare a proposed iterative alternative which requires far fewer parameters. | en |
| dc.identifier.uri | http://hdl.handle.net/2003/43597 | |
| dc.identifier.uri | http://dx.doi.org/10.17877/DE290R-25430 | |
| dc.language.iso | en | |
| dc.subject | Semiconductor | en |
| dc.subject | Time series | en |
| dc.subject | Forecasting | en |
| dc.subject | Innovation | en |
| dc.subject | Market | en |
| dc.subject | Hierarchical time series | en |
| dc.subject | Reconciliation | en |
| dc.subject.ddc | 310 | |
| dc.subject.rswk | Halbleiterindustrie | de |
| dc.subject.rswk | Marktprognose | de |
| dc.subject.rswk | Parameterschätzung | de |
| dc.subject.rswk | Prognosemodell | de |
| dc.subject.rswk | Maschinelles Lernen | de |
| dc.title | Forecasting in the semiconductor industry | en |
| dc.type | Text | |
| dc.type.publicationtype | PhDThesis | |
| dcterms.accessRights | open access | |
| eldorado.secondarypublication | false |