## PROVEDA: A scheme for Progressive Visualization and Exploratory Data Analysis of clusters. (1999)

Venue: | In Proceedings of 3rd Software Visualization Workshop |

Citations: | 1 - 0 self |

### BibTeX

@INPROCEEDINGS{Quigley99proveda:a,

author = {Aaron J Quigley and Peter Eades},

title = {PROVEDA: A scheme for Progressive Visualization and Exploratory Data Analysis of clusters.},

booktitle = {In Proceedings of 3rd Software Visualization Workshop},

year = {1999}

}

### OpenURL

### Abstract

This paper presents a scale-oriented scheme for data visualization. The aim is to explore and validate the hypothesis, that a high quality visual layout exhibits a good quality hierarchical data clustering.In this scheme, the information to be visualized and clustered is represented as a graph, where the nodes represent pieces of information and the edges represent relationships between those pieces. This scheme supports three related models of information, the underlying graph structure, the graph clustered according to some geometric attributes and, the graph represented according to some drawing mechanism. Also introduced is a method for reducing the computational complexity of a graph drawing algorithm called, force directed placement,fromO(n 2 ) to O(n log n). This method is adapted from an n-body hierarchical force calculation, that allows larger data sets to be draw and visualized on various levels of abstraction. Finally, this method provides the framework for the integration and testing of numerous concepts about how the quality of layout and clustering are related.

### Citations

1666 |
A k-means clustering algorithm
- Hartigan, Wong
- 1979
(Show Context)
Citation Context ...emetric to determine similarity for cluster generation. The K-means algorithm is based on a distance metric where the attributes are scaled so that the Euclidean distance between cases is appropriate =-=[5,9]. -=-Objects are classified into only one partition or cluster by K-means. 2.3 Graph Drawing Since the attributes of objects can be represented as vectors in ℜ k . One method used to help understand this... |

1233 |
The Design and Analysis of Spatial Data Structures
- Samet
- 1989
(Show Context)
Citation Context ...rarchical regular space decomposition, Quad tree. For example in image processing, a commonly used data structure is a quad-tree which uses a regular decomposition to describe a two dimensional image =-=[10]-=-, an example of which is shown in Figure 5 with an area indicated. A segment of the corresponding data structure, with the corresponding leaf, is shown in Figure 6. The data structure is labeled in re... |

216 | A hierarchical o(n log n) force-calculation algorithm - Barnes, Hut - 1986 |

210 | Voronoi diagrams and Delaunay triangulations - Fortune - 1992 |

137 |
Which aesthetic has the greatest effect on human understanding
- Purchase
- 1997
(Show Context)
Citation Context ...her, more abstract measures, such as how much information is conveyed, how fast the diagram presents a piece of information are more difficult to measure and are the subject of considerable HCI study =-=[3,20,21,29]-=-. 3. Quality Measures In abstract terms a quality measure or metric is a function that returns a value. This value can be used to compare the results of applying more than one method to a particular p... |

124 |
Graph Drawing by Force-Directed Placement. Software-Practice & Experience
- Fruchterman, Reingold
- 1991
(Show Context)
Citation Context ...location for the node and a route for the edge, this is the classical Graph Drawing problem. This problem has been researched since graphics workstations were introduced in the 1980s see; for example =-=[3,4,11,12,15,27]-=-. Different types of objects and their relationships can be expressed in graph terms. This lends itself as a very good model for the representation of information. A good visual representation of thes... |

90 |
Cluster Analysis Algorithms for Data Reduction and Classification of Objects (Ellis Horwood
- Späth
- 1980
(Show Context)
Citation Context ...can be evaluated in terms of the distance between the vector endpoints. Similarity or dissimilarity measures based on k-dimensional vector distance comparisons are referred to as geometric clustering =-=[1,5,8]. In g-=-eneral any distance metric which obeys the following conditions can be used in similarity measures for geometric clustering. • D(x,y) ≥ 0 distances cannot be negative • D(x,y) = 0 Ifandonlyifx=y... |

81 | Choosing effective colours for data visualization
- HEALEY
(Show Context)
Citation Context ...her, more abstract measures, such as how much information is conveyed, how fast the diagram presents a piece of information are more difficult to measure and are the subject of considerable HCI study =-=[3,20,21,29]-=-. 3. Quality Measures In abstract terms a quality measure or metric is a function that returns a value. This value can be used to compare the results of applying more than one method to a particular p... |

56 |
Computer Simulation Using Particles, McGraw-Hill
- Hockney
- 1989
(Show Context)
Citation Context ...s (factors such as processing speed and parallel simulations also impact the size of n). A method used in nbody simulation called PIC (particle in cell) uses a regular grid to divide the entire space =-=[16,17]-=-. All the particles in a single grid cell are combined to form an average source density. Instead of computing each particle to particle force, a particle to cell comparison can be used (with a result... |

48 | An experimental comparison of force-directed and randomized graph drawing algorithms - Brandenburg, Himsholt, et al. - 1995 |

35 | Graph layout adjustment strategies
- Storey, Müller
- 1995
(Show Context)
Citation Context ...rious levels of abstraction. Visualization systems that deal with large data sets employ various techniques such as, a large virtual canvas (scrolling & zooming), information mural [24], fisheye view =-=[14]-=- and logical frames with trails [11] to aid the user in viewing and exploring the information presented. However, the complexity and size of these data sets means any visualization tool is limited in ... |

25 |
On-line animated visualization of huge graphs using a modified spring algorithm
- ML, Eades, et al.
- 1998
(Show Context)
Citation Context ...zation systems that deal with large data sets employ various techniques such as, a large virtual canvas (scrolling & zooming), information mural [24], fisheye view [14] and logical frames with trails =-=[11]-=- to aid the user in viewing and exploring the information presented. However, the complexity and size of these data sets means any visualization tool is limited in its ability to present an overall pi... |

25 | Algorithms for drawing clustered graphs
- Feng
- 1997
(Show Context)
Citation Context ...location for the node and a route for the edge, this is the classical Graph Drawing problem. This problem has been researched since graphics workstations were introduced in the 1980s see; for example =-=[3,4,11,12,15,27]-=-. Different types of objects and their relationships can be expressed in graph terms. This lends itself as a very good model for the representation of information. A good visual representation of thes... |

15 | A spring modeling algorithm to position nodes of an undirected graph in three dimensions - Kumar |

6 | The Psychology of Visualization
- Csinger
- 1992
(Show Context)
Citation Context ...her, more abstract measures, such as how much information is conveyed, how fast the diagram presents a piece of information are more difficult to measure and are the subject of considerable HCI study =-=[3,20,21,29]-=-. 3. Quality Measures In abstract terms a quality measure or metric is a function that returns a value. This value can be used to compare the results of applying more than one method to a particular p... |

6 | et al. Layout Adjustment and the Mental Map - Misue - 1995 |

4 |
Eades et al, “Graph Drawing, algorithms for visualization of graphs
- Battista, Peter
- 1999
(Show Context)
Citation Context ...location for the node and a route for the edge, this is the classical Graph Drawing problem. This problem has been researched since graphics workstations were introduced in the 1980s see; for example =-=[3,4,11,12,15,27]-=-. Different types of objects and their relationships can be expressed in graph terms. This lends itself as a very good model for the representation of information. A good visual representation of thes... |

3 |
Cluster Analysis for Social Scientists. Jossey-Bass
- Lorr
- 1983
(Show Context)
Citation Context ...ring literature “single-linkage clusters” are formed by successively eliminating links, from a minimum spanning tree of the graph, starting with the longest link first until some threshold is reac=-=hed [5,1]-=-. Figure 1 shows clusters formed from a single linkage cluster analysis. A B C D E F G H I J K L M N A 0 1 1 0 0 0 0 0 0 0 0 0 0 0 B 1 0 1 1 0 0 0 0 0 0 0 0 0 0 C 1 1 0 1 0 0 0 0 0 0 0 0 0 0 D 0 1 1 0... |

3 |
A heuristic for graph drawing. Congress Numerantium
- Eades
- 1984
(Show Context)
Citation Context |

3 | Generating abstractions for visualization - Reiss, Sarkar - 1992 |

3 |
et al, Ciao: A graphical Navigator for Software and Document Repositories
- Chen
- 1995
(Show Context)
Citation Context ...ut on and across various levels of abstraction. Visualization systems that deal with large data sets employ various techniques such as, a large virtual canvas (scrolling & zooming), information mural =-=[24]-=-, fisheye view [14] and logical frames with trails [11] to aid the user in viewing and exploring the information presented. However, the complexity and size of these data sets means any visualization ... |

1 |
Advances in System Analysis Vol. 4, Graphs as Structural Models: The Application
- Erhard
- 1988
(Show Context)
Citation Context ... Measures. 1. Introduction Exploratory data analysis or EDA is a method used by researchers and statisticians from various fields of science to investigate and explore the structure of large data sets=-=[1]. Th-=-e size and complexity of these data sets has resulted in the development of techniques which can ‘average’ parts of the data set so a better insight into the overall structure of the data can be o... |

1 |
A Practical Comparison of N-Body Algorithms
- Blellcoh
(Show Context)
Citation Context ...tion time and quality (i.e. difference from direct particle-particle computation, which can be thought of as optimal) In empirical studies a choice of Θ between 0.1 – 1.0 proves a practical trade o=-=ff [17,7]-=-. Once each particle is dealt with, then the forces from the internal nodes are propagated down the tree and applied to the particles. The tree is then rebuilt and the process repeated. The reduction ... |

1 | et al., Introduction to Computer Graphic,Addison-Wesley - Foley - 1994 |