Determining optimal architecture for dynamic linear models in time series applications

UNCW Author/Contributor (non-UNCW co-authors, if there are any, appear on document)
Kathleen Mary Karlon (Creator)
The University of North Carolina Wilmington (UNCW )
Web Site:
Edward Boone

Abstract: This work is focused on assessing the performance of one particular time series forecasting paradigm: Dynamic Linear Models (DLM). This research extends the M3 forecasting competition, a large-scale project to assess the e±cacy of various forecasting methods and also that of the research done in [14]. This work provides insight into the performance of the DLM against the model architecture. Symmetric Mean Absolute Percentage Error and Linear Mixed Models are used to analyze the competition results, which showed that paradigm performance is dependent upon the class of time series. Furthermore, in some cases, the chosen DLM models from this work outperform optimal models from [14]. This work explores di®erent DLM models and compares the results with previously chosen models to determine if the models from this work outperform other models.

Additional Information

A Thesis Submitted to the University of North Carolina at Wilmington in Partial Fulfillment of the Requirement for the Degree of Masters of Science
Language: English
Date: 2009
Linear models (Statistics), Time-series analysis
Time-series analysis
Linear models (Statistics)

Email this document to