| dcterms.abstract | The growth of digital data is tremendous. These data come from many aspects
of life and matter such as medicine, science, environment monitoring, business,
finance, social networks, etc. When the data is multivariate, or the dimensionality
of the data becomes high, it can be a challenge for analysts to understand
the intricate relations among the data. The data types not only consist of static
data, but also dynamic data, geospatial data, network data etc. The various types
make it even more difficult for the analysis. Visual analytics can offer powerful
mechanisms to assist humans in the exploration of these complex data, by mining
the relations from the raw data and sculpting them as visualizations to help
humans gain insight. In the thesis, we focus on relation discovery in multivariate
static, dynamic, geospatial, and network data via several new visual analytics approaches.
First, we analyze the relations among the static multivariate data and
propose the data context map which can illustrate the relations among data items
and attributes. Then we extend the mapping to the dynamic case, aiming to capture
and visualize the attribute relation behaviors in dynamic flows with our tool
StreamVisND. Next, we move to the geospatial data to recover the relations in the
geospatial data. To achieve this, we developed the ColorMapND framework to visualize
and colorize multi-field, multi-channel, multi-spectral data on the geospatial
or image domain. Finally, we consider the complex topology that shapes the multivariate data, such as network data and visualize the relations in this kind of
complex network topology. We first study the relations of common networks by
modified spectral embedding and then extend our work to multi-dimensional torus
networks with the proposed framework TorusTra f f icND. | |
