Visualization of high dimensional data and predictive precursors in time series data |
Posted on:2011-11-21 | Degree:Ph.D | Type:Dissertation |
University:North Dakota State University | Candidate:Jockheck, William H | Full Text:PDF |
GTID:1468390011971769 | Subject:Computer Science |
Abstract/Summary: | |
This disquisition is based on several published papers that focus on the visualization of n-dimensional data including multiple parallel time series data sets. It includes developments and contributions to visual data mining techniques which allow the user to visualize the relationships in the data to gain insight into the information represented by the data. The techniques developed aid in the identification and extraction of classes and isolation of class outliers. Key developments include the jewel diagram, the fireworks diagram and predictive precursor snapshots. The work includes the design and development of the jewel diagram and fireworks diagram as tools to project highly dimensional data into a two dimensional space. Both of these techniques preserve a single view that incorporates attribute distribution and the n-dimensional projection. The predictive precursor snapshot technique provides an encapsulated view of the relationships between multiple parallel time series data sets. |
Keywords/Search Tags: | Time series data, Predictive precursor, Dimensional data |
|
Related items |