Date of Award
Thesis - Restricted
Master of Science (MS)
Electrical and Computer Engineering
Bansal, Naveen K.
Analysis of Time-Evolving Systems is an important and challenging problem. Although it is a necessary task, it is very difficult to retrieve information for the analysis of those systems using existing methods. Recent methods, such as, Time Series Data Mining (TSDM) are used to analyze time-related patterns also known as temporal patterns. TSDM focuses on retrieving a single temporal pattern that can be used throughout the data set, ignoring other potential temporal patterns. This thesis presents a novel method, Multiple Temporal Pattern Recognition (MTPR), that retrieves multiple patterns in the Phase Space Embedding, a method used to represent data in other spaces other than the original space representation without loosing the underlying properties of the data set, and uses these patterns to form a predictability scheme. The statistical interpretation of each temporal pattern's predictability confirms the usefulness of those patterns. The new method retrieves potential temporal patterns in phase spaces. By clustering the points in the phase space using a modified Boundary Detection Algorithm (rnBDA), the candidate temporal patterns are evaluated using statistical logistic regression.
Senyana, Odilon K., "Multiple Temporal Pattern Recognition and Predictability Analysis of Complex Time-Evolving Systems" (2007). Master's Theses (1922-2009) Access restricted to Marquette Campus. 4128.