Date of Award:
5-2026
Document Type:
Thesis
Degree Name:
Master of Science (MS)
Department:
Computer Science
Committee Chair(s)
Steve Petruzza
Committee
Steve Petruzza
Committee
John Edwards
Committee
Soukaina Filali Boubrahimi
Abstract
Understanding how physical systems change over time is important in areas such as weather prediction, fluid dynamics, and environmental science. However, accurately predicting future behavior is difficult because these systems are complex and constantly evolving.
This research develops a deep learning approach to predict how such systems evolve over time. The model learns patterns from past observations and uses them to generate future states step by step. This provides a faster alternative to traditional simulation methods while maintaining strong predictive performance.
The proposed method focuses on improving the consistency of predictions over time and is designed to work across different physical conditions. This allows a single model to handle a variety of scenarios without needing separate training for each case.
The approach is tested on multiple datasets representing different types of flow behavior. Results show that it can produce accurate and stable predictions over longer time periods compared to existing methods.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Islam, Md Robiul, "ST-FMFormer: An Autoregressive Generation Framework for Scientific Ensemble Data Predictions" (2026). All Graduate Theses and Dissertations, Fall 2023 to Present. 791.
https://digitalcommons.usu.edu/etd2023/791
Included in
Copyright for this work is retained by the student. If you have any questions regarding the inclusion of this work in the Digital Commons, please email us at .