Home Home Faculty Home Faculty of Computer Science
Free University of Bozen - Bolzano

Database and Information Systems

ABOUT DIS
Staff
Guests
How to reach us
Open Positions

TEACHING
BSc: DB Stream
MSc: DB Stream

COURSES
Advanced Topics in Databases
Advanced Topics in Inf. Systems
Approximation: Theory and Algorithms
Data Management Systems
Data Structures and Algorithms
Data Warehousing/Mining
Distributed Databases
Mobile Services
Seminar in Databases
Temporal and Spatial Databases

RESEARCH
Publications
PhD Projects
Software
Scientific Services
Seminars

RESEARCH PROJECTS
eBZ – 2015
COSPA
3DVDM-DS
Eurescom P817
Chorochronos
XVDM

FACULTY
IT Services
Faculty Council
Regulations

LINKS
DBLP
OnlineLibraries@unibz
G

W
About
Requirements
Download
Installation
Running
New Module
How Tos
Credits
Links

Density Surface Module: the First Session

A density surface is a surface, which encloses the data points of a and higher density. An example of density surface for a three-dimensional normal distribution is shown in Figure 1 (the last two figures). In the figures the most dense point is in the center of the figure. The density level parameter is chosen so that the surface encloses almost all data points. The APDF-DS module computes, and visualizes density surfaces.

The APDF-DS module of the 3DVDM system can be run with the following command:

The executable opens a VR++ window and a GUI of the density surfaces module (cf. Figure 1).
gui
GUI of the APDF-DS module
vr++, ds
VR++ window
vr++, scatter plot
Scatter Plot
Figure 1

Computation of density surfaces is done in two steps: (i) the apdf tree is computed for the database, and (ii) a density surface is computed from the apdf tree. In order to compute the apdf tree one needs to select the database, the X, Y, Z attributes of the database, shape error, and the size initial partition. Once these parameters are selected the the apdf tree is computed with the press of the ``Compute PDF'' button. Density surface is visualized immediatelly after the APDF tree is computed.

data source
Data Source
estimation parameters
Attributes and Estimation
density level parameters
Density Level
Figure 2

The data source controls the data for which density surfaces are calculated. The estimation error controls the precision of the estimation of the probability density function (PDF) of the dataset. The density level parameter controls the level of the density surface. As the parameter changes from 0.0 to 1.0 the biggest surface (the surface, which envelops all the data points) to the smallest (the surface, which envelops only the most dense data points) is visualized.

A more elaborate discussion of the DS module and its parameters can be found in the manual of the DS module.

Screen Shots

This section presents a number of screen shots of the DS module for various datasets.

Figure 3 presents a scatter plot together with density surfaces for the click stream data for the http://www.sunsite.dk/ site. In the figures, the document dimension is mapped to the Z axe (blue axe), time dimension is mapped to the Y axe (green axe), and domain dimension is mapped to the X axe (red dimension). The red surface represents .com clicks, while the green surface represents the .dk clicks. Density surfaces show the dynamics of the clicks (increase and decrease of the number of clicks during certain time periods of the day), which is impossible to see in the scatter plot. Click on the image to see the animated version of the picture.

click-stream-scatter-plot
Rotation of Scatter Plot
click-stream-rotate-ds
Rotation of Density Surfaces
click-stream-animate-ds
Animation of Density Surfaces
as Density Level Increases
Figure 3