Show simple item record

dc.contributor.authorQu, Junfeng
dc.date.accessioned2014-03-04T01:06:26Z
dc.date.available2014-03-04T01:06:26Z
dc.date.issued2006-05
dc.identifier.otherqu_junfeng_200605_phd
dc.identifier.urihttp://purl.galileo.usg.edu/uga_etd/qu_junfeng_200605_phd
dc.identifier.urihttp://hdl.handle.net/10724/23223
dc.description.abstractThis research emphasizes discovery of important changes in structure of time series data. These structural changes imply the loss of customary patterns and the appearance of a novel pattern that has not been experienced previously. Most previous work in this area has been concentrated on identifying previously known or predefined patterns. The major distinction of my research is that the framework offers the ability to discover internal structural changes of time series online dynamically without statistical assumptions about data. The internal structure within time series can be used to improve solutions and provide important insights into the problem domains. Analysis of structured time series data is widely used for many applications, such as economic forecasting, stock market analysis, and networks, etc. This dissertation introduces our research on high-precision modeling, prediction, similarity matching of time series data with consideration of internal structures of data. Our objectives include: (i) formulating a framework for online dynamic gray modeling of time series data streams, (ii) analyzing and characterize the structure of time series stream data using reference and test models, (iii) developing real-time prediction methods based on the online modeling results of corresponding internal structure, and (iv) developing a real-time online similarity matching method that considers the identified internal structures of the time series. We have developed an integrated online structural changes mining (SCM) framework to achieve these objectives. The framework is composed of (a) a dynamic gray model (DGM) that captures the internal structure of time series data online, (b) an algorithm that whitens the incoming data into structures based on a reference model and a test model according to the underlying DGM, (c) a set of analytical methods for data analysis, including online subsequence matching, which generates dynamic query subsequences, defines new subsequence similarity measures, and performs similarity matching with consideration of the internal structures of time series data. This framework is very useful for real-time systems where response time is critical. We have applied the framework to multiple problem domains, such as financial data analysis and network traffic analysis.
dc.languageeng
dc.publisheruga
dc.rightspublic
dc.subjecttime series
dc.subjectdata mining
dc.subjectdynamic system
dc.subjectgray model
dc.subjectsimilarity matching
dc.titleTime series data mining of structure changes using dynamic systems
dc.typeDissertation
dc.description.degreePhD
dc.description.departmentComputer Science
dc.description.majorComputer Science
dc.description.advisorHamid R. Arabnia
dc.description.committeeHamid R. Arabnia
dc.description.committeeKhaled Rasheed
dc.description.committeeJack Houston


Files in this item

FilesSizeFormatView

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record