Unit-IV: Visualization of Different Structures

Relationship of Information

⮚ While most of the visualization techniques focus on the display of data values and their
attributes, another important application of visualization is the conveying of relational
information, e.g., how data items or records are related to each other.

⮚ These interrelationships can take many forms:

• part/subpart, parent/child, or other hierarchical relation
• connectedness, such as cities connected by roads or computers connected by networks
• derived from, as in a sequence of steps or stages
• shared classification
• similarities in values
• similarities in attributes (e.g., spatial, temporal)

⮚ Relationships can be simple or complex: unidirectional or bi-directional, nonweighted or

weighted, certain or uncertain.

⮚ Indeed, the relationships may provide more and richer information than that contained in
the data records.

Groups
1. Venn Diagram:
1. Description: Overlapping circles represent relationships between groups, with
each circle representing a distinct category.
2. Application: Used to show the overlap or commonalities between different sets
of entities.
2. Clustered Bar Chart:
1. Description: Bars are grouped together based on categories, and each bar within
a group represents a subcategory.
2. Application: Effective for comparing values within different categories and
subcategories.
3. Pie Chart:
1. Description: A circular chart divided into slices, each representing a proportion
of the whole.
 2. Application: Useful for displaying the distribution of a whole into different
categories or groups.
4. Heatmap:
1. Description: A matrix where colors represent values, and rows or columns are
grouped based on categories.
2. Application: Commonly used to visualize the intensity of relationships or values
within a matrix.
5. Dendrogram:
1. Description: A tree diagram that represents hierarchical relationships, with
entities grouped based on similarity.
2. Application: Commonly used in clustering analysis to show relationships within
hierarchical structures.
6. Bubble Chart:
1. Description: Circles represent groups, and the size of each circle indicates a
certain value or dimension.
2. Application: Useful for comparing multiple dimensions within different groups.
7. Tree Map:
1. Description: Rectangles represent groups, with each rectangle divided into sub-
rectangles based on a specific dimension.
2. Application: Efficient for visualizing hierarchical data and the distribution of
values within groups.
8. Chord Diagram:
1. Description: Circles represent groups, and chords connect related groups,
showing relationships between them.
2. Application: Suitable for visualizing connections or interactions between
multiple groups.
9. Radar Chart:
1. Description: A chart with spokes radiating from the center, each representing a
different category or group.
2. Application: Useful for comparing values across multiple dimensions for
different groups.
Venn Diagram

⮚ A Venn diagram is a type of chart used to visualize the overlap between two or more
datasets. It typically uses circles or ellipsis to illustrate relationships between segments,
graphically highlighting how they are similar to or different from each other.

⮚ Venn diagrams are widely used in many disciplines, most notably in mathematics,
statistics, logic, teaching, linguistics, computer science and business.

⮚ They are particularly effective at organizing and visualizing datasets, making the
relationship easier to understand. Because of its powerful visual effect, the Venn diagram
is a popular choice for business reports and presentations

Displaying Hierarchial Structures

⮚ Trees or hierarchies (we’ll use the terms interchangeably) are one of the most common
structures to hold relational information.

⮚ For this reason, many visualization techniques have been developed for display of such
information.

⮚ We can divide these techniques into two classes of algorithms:

⮚ space-filling.

⮚ non–space-filling.
Space Filling Methods

⮚ As the name implies, space-filling techniques make maximal use of the display space.

⮚ This is accomplished by using justatpositioning to imply relations, as opposed to, for

example, conveying relations with edges joining data objects.

⮚ The two most common approaches to generating space-filling hierarchies are rectangular
and radial layouts.

Tree Maps- Rectangular Space Filling

⮚ Treemaps and their many variants are the most popular form of rectangular space-filling
layout.

⮚ In the basic treemap, a rectangle is recursively divided into slices, alternating horizontal
and vertical slicing, based on the populations of the subtrees at a given level.

⮚ As mentioned, many variants on treemaps have been proposed and de veloped since they
were introduced, including squarified treemaps (to reduce the occurrence of long, thin
rectangles) and nested treemaps (to emphasize the hierarchical structure).

Algorithm
Radial Space Filling

⮚ Radial space-filling hierarchy visualizations, sometimes referred to as sunburst displays,

have the root of the hierarchy in the center of the display and use nested rings to convey
the layers of the hierarchy.

⮚ Each ring is divided based on the number of nodes at that level.

⮚ These techniques follow a similar strategy to treemaps, in that the number of terminal
nodes in a subtree determines the amount of screen space that will be allocated for it.

⮚ However, unlike treemaps, which assign most screen space to conveying the terminal
nodes, radial techniques also show the intermediate nodes.
Algorithm
Non Space Filling Methods

⮚ The most common representation used to visualize tree or hierarchical re lationships is a

node-link diagram.

⮚ Organizational charts, family trees, and tournament pairings are just some of the common
applications for such diagrams.

⮚ The drawing of such trees is influenced the most by two factors: the fan-out degree (e.g.,
the number of siblings a parent node can have) and the depth (e.g., the furthest node from
the root).

⮚ Trees that are significantly constrained in one or both of these aspects, such as a binary
tree or a tree with only three or four levels, tend to be much easier to draw than those
with fewer constraints.
 ⮚ When designing an algorithm for drawing any node-link diagram(not just trees), one
must consider three categories of often-contradictory guidelines: drawing conventions,
constraints, and aesthetics.

⮚ Conventions may include restricting edges to be either a single straight line, a series of
rectilinear lines, polygonal lines, or curves.

⮚ Other conventions might be to place nodes on a fixed grid, or to have all sibling nodes
share the same vertical position.

⮚ Constraints may include requiring a particular node to be at the center of the display, or
that a group of nodes be located close to each other, or that certain links must either go
from top to bottom or left to right.

⮚ Each of the above guidelines can be used to drive the algorithm design.

⮚ Aesthetics, however, often have significant impact on the interpretability of a tree or

graph drawing, yet often result in conflicting guidelines.

⮚ Some typical aesthetic rules include:

• minimize line crossings
• maintain a pleasing aspect ratio
• minimize the total area of the drawing
• minimize the total length of the edges
• minimize the number of bends in the edges
• minimize the number of distinct angles or curvatures used
• strive for a symmetric structure

⮚ For trees, especially balanced ones, it is relatively easy to design algorithms that adhere
to many, if not most, of these guidelines.

⮚ For example, a simple tree drawing procedure is given below

1. Slice the drawing area into equal-height slabs, based on the depth of the tree.
2. For each level of the tree, determine how many nodes need to be drawn.
3. Divide each slice into equal-sized rectangles based on the number of nodes at that level.
4. Draw each node in the center of its corresponding rectangle.
5. Draw a link between the center-bottom of each node to the center-top of its child node(s).
 ⮚ Many enhancements can be made to this rather basic algorithm in order to improve space
utilization and move child nodes closer to their parents

⮚ Some of these include:

• Rather than using even spacing and centering, divide each level based on the number of
terminal nodes belonging to each subtree.
• Spread terminal nodes evenly across the drawing area and center parent nodes above them.
• Add some buffer space between adjacent non sibling nodes to emphasize relationships.
• If possible, reorder the subtrees of a node to achieve more symmetry and balance.
• Position the root node in the center of the display and lay out child nodes radially, rather than
vertically.
Cone Tree

⮚ For large trees, a popular approach is to use the third dimension, supplemented with tools
for rotation, translation, and zooming.

⮚ Perhaps the most well-known of such techniques is called a cone tree.

⮚ In this layout, the children of a node are arranged radially at evenly spaced angles and
then offset perpendicular to the plane.

⮚ The two parameters critical to this process are the radius and offset distance; varying
these influences the density of the display and the level of occlusion.

⮚ Minimally they should be set so that separate branches of the tree do not fall into the
same section of 3D space.

⮚ One method to ensure this is to have the radius inversely proportional to the depth of a
node in the tree.
Displaying Graphs and Networks

⮚ Technically speaking, a tree is a connected, unweighted, acyclic graph.

⮚ Clearly, there are many other possibilities, including graphs with weighted edges,
undirected graphs, graphs with cycles, disconnected graphs, and so on.

⮚ Types of graphs or networks include:

1. Node-link graphs
2. Planar Graphs
3. Matrix representation of graphs

Node-Link Graphs

⮚ Force-directed graph drawing methods use a spring analogy to represent the links, with
node positions iteratively refined until the overall energy or stress of the system is
minimized
⮚ For each pair of connected nodes, there are two forces: fij, the force caused by the spring
between them, and gij, an electrical repulsion force to keep nodes from getting too close.

⮚ A simple model is to use Hooke’s law to represent the spring force and an inverse square
law to represent the repulsion force.

⮚ If d(i,j) is the Euclidean distance between nodes i and j, si,j is the natural spring length (at
rest), and kij is the spring tension, the x-component of the spring force between two nodes
can be computed as

⮚ fij(x)=kij ∗ (d(i,j) − sij) ∗ (xi − xj)/d(i,j).

⮚ If rij is the strength of the repulsion between nodes i and j, the x component of the
repulsion force can be computed as gij(x)=(rij/d(i,j)2) ∗ (xi − xj)/d(i,j)

⮚ Thus, one step of the position refinement process would calculate the sum of all the
forces on each node (x-, y-, and z-components, as appropriate) and move its position
proportional to that force.

⮚ Clearly, once points have moved, all the forces need to be recalculated and another shift
of positions made.

⮚ To avoid oscillation, it is common to start with movements that are a significant

percentage of the force and then use smaller and smaller step sizes to converge on the
point where the forces are minimized.
 ⮚ Initial positions can be assigned randomly.

Planar Graphs

⮚ Planar graph drawing techniques start with the assumption that the underlying graph is
planar, e.g., it has no edge crossings.

⮚ These algorithms have gotten a lot of attention, for several reasons.

⮚ First, as the theory of planar graphs has a long history, there are many concepts that can
be exploited from the literature.

⮚ Second, as edge-crossings tend to make graphs difficult to read, it is a good strategy to

minimize or eliminate such crossings.

⮚ Finally, planar graphs tend to be sparse; Euler’s formula

indicates that a planar graph with n vertices has at most 3n − 6
edges.

⮚ Concentrating on planar graphs is not overly restrictive, as one can eliminate crossings by
inserting dummy nodes at the crossings, perform the layout using a planar graph
algorithm, and then remove the dummy nodes.

⮚ We will, in addition, assume that the graph is connected, e.g., there is a path from every
node to every other node.

⮚ Graphs that are not connected can be separated into subgraphs that can be drawn
separately.

⮚ A subgraph that is maximally connected (all nodes are connected) is a connected

component of the graph.
Definitions of Planar Graphs

⮚ A face is a partition of the plane isolated by a set of connected vertices.

⮚ A neighbor set is a counter-clockwise listing of the vertices incident to a particular

vertex.

⮚ A planar embedding is a class of planar graph drawings with the same neighbor sets for
each vertex. A planar graph can have an exponential number of such embeddings.

⮚ A cut vertex is any node that causes the graph to be disconnected if it is removed.

⮚ A biconnected graph is one without a cut vertex.

⮚ A block is a maximally biconnected subgraph of a graph.

 ⮚ A separating pair means two vertices whose removal causes a biconnected graph to
become disconnected.

⮚ A triconnected graph is one without a separating pair. A planar triconnected graph has a
unique embedding.
Algorithm to Represent a Planar Graph

⮚ A graph is planar only if all of its connected components are also planar.

⮚ Similarly, we can state that a connected graph is planar only if all its biconnected
components are planar.

⮚ Thus, we just need an algorithm that determines if a biconnected graph is planar or not.

⮚ We will perform a divide-and-conquer approach by noting that if our graph contains a

cycle such that no other cycle is present that doesn’t contain an edge of the original cycle
(e.g., there aren’t cycles left when the edges involved in the original cycle are removed),
what remain are paths that start and stop on one of the vertices of the cycle (called
attachments).

⮚ These pieces of the graph can be drawn either within the cycle or outside the cycle.

⮚ Two such pieces interlace if they both start and end on nodes of the cycle, and the two
ends of one piece are separated by one end of the other piece.

⮚ To be drawn in a planar fashion, one of these interlaced pieces would need to be drawn
inside the cycle, and the other on the outside.

⮚ If we now create a graph of all the pieces, with an edge between two pieces if they
interlace, as long as this graph is bipartite (separable into two sets of vertices such that no
edge exists between members of the same set), the original graph is planar.

⮚ If the graph contains more cycles after removing the edges of the original cycle, this
means that one or more of the pieces contains a cycle.

⮚ In this case, we create a subgraph containing this piece and a section of the original cycle
connecting the end points of the part, and recursively call the planarity test algorithm.
Pseudocode of Planar Graph Algorithm

⮚ Given a biconnected graph G and a separating cycle C:

⮚ 1. Compute all the pieces of G with respect to C.

⮚ 2. For each piece P that is not a simple path (e.g., that contains a cycle).

⮚ (a) Create graph G consisting of P plus C.

⮚ (b) Create cycle C consisting of a path through P plus the section of C joining the ends.

⮚ (c) Apply the algorithm to (G,C).

⮚ If the result is nonplanar, G is nonplanar.

⮚ 3. Compute the interlacement graph I of the pieces of G.

⮚ 4. If I is not bipartite, G is nonplanar; else G is planar

Converting Non-Planar Graph to Planar Graph

⮚ If a graph is nonplanar, we can make it planar using the following strategy:

⮚ 1. Determine the largest planar subgraph of the graph.

⮚ 2. For the remaining vertices, place each within a face that minimizes the number of edge
crossings.

⮚ 3. For each edge crossing, break the edges into two parts each, and connect the broken
ends to a new dummy vertex.
Drawing Planar Graphs

⮚ Once a graph has been either determined to be planar or has been aug mented to achieve
planarity, there are many possible strategies for generating a drawing.

⮚ One such technique, called the visibility approach, consists of a two-step process.

⮚ In the first step, called the visibility step, a visibility representation of the graph is
formed.

⮚ In such a representation, each vertex is depicted as a horizontal line segment, and each
edge is depicted as a vertical line connecting the corresponding vertex segments.

⮚ It should be clear that for a planar graph, it is always possible to draw such a
representation without crossing edges other than where they meet the vertex segments.

⮚ In the second step, called the replacement step, each vertex segment is collapsed to a
single point, and each vertical connector is replaced by a polyline that follows the
original edge as much as possible, with a segment at each end connecting the edge to its
corresponding vertex.

⮚ Many options exist for the replacement step, including the location of the nodes and the
strategy used to form the connections (e.g., straight versus curved lines, single segment
versus multiple segments).

Matrix Representation of Graphs

⮚ An alternate visual representation of a graph is via an adjacency matrix, which is an N by

N grid(where N is the number of nodes),where position (i,j) represents the existence (or
not) of a link between nodes i and j.
 ⮚ This may be a binary matrix, or the value might represent the strength or weight of the
link between the two nodes.

⮚ This method overcomes one of the biggest problems with node-link diagrams, namely
that of crossing edges.

⮚ Bertin was one of the first researchers to investigate the power of this representation,
using different reordering strategies to organize the rows and columns to reveal structures
within the graph.

⮚ There have been numerous algorithms proposed for reordering the rows and columns of
the matrix to expose the most structure.

⮚ Some are primarily user-driven, which would support ordering based on the values in one
of the rows or columns as a starting point.

⮚ Others are purely automatic, which rely on some metric for evaluating a particular
ordering and a strategy for generating orders to test.

⮚ we can use a simplistic order evaluation strategy, namely to count the number of
occurrences of matching elements in adjacent rows or columns.

⮚ This tends to group nodes that link or don’t link to a common node.

Issues with Visualizing Hierarchial Structures (Trees/Graphs)

Once a basic visualization of a tree or graph has been developed, there are a number of additional
considerations, primarily addressing the issue of interpretability. Two such important
considerations will be elaborated upon in this section: labeling and interaction
Labeling
Proper labeling of a visualization is crucial to allow a viewer to understand what is being shown.
A map would be of little value without some form of labeling; similarly, a color-coded plot would
be difficult to understand without some indication of the meaning associated with the colors. In
tree and graph drawing, the problem of labeling is compounded, not only because of the potential
for many nodes, but also because labels might also be needed for the links between nodes. If there
are only a small number of distinct labels, such as showing the type of link or a class associated
with a node, it is best to use nontextual labels, such as the color, size, or shape of a node or the
color, thickness, or line style of a link. This does not require much screen space and can usually
be interpreted unambiguously even in the presence of modest amount of line crossing and node
occlusion. However, if the number of distinct labels exceeds five or six, the likelihood of
misinterpretation can become large. A key for interpreting the graphical attribute mapping is
essential. For small graphs, a common strategy for node labeling is to put the labels within the
nodes, using rectangular or oval node shapes to accommodate the text. To avoid distorting the
perception of the nodes, the size of the nodes should be dictated by the length of the longest label.
For situations where the labels can be very long, one option is to use abbreviations or numeric
labels, along with a key for interpretation. Viewers will eventually learn the correspondences
between the shortened labels and their actual meaning. A similar strategy can be used for edge
labeling, placing the labels near the center of the edge. For edges that are predominantly vertical,
these should be to the left or right of the edge, while for predominantly horizontal edges, they
should be above or below. Using a consistent strategy will reduce the potential for erroneously
associating a label with the wrong edge. At the other extreme, if there are a large number of distinct
labels that need to be shown, or the labels themselves are quite long, it becomes readily apparent
that simultaneous display of all labels will be ineffective. Several strategies have been developed
to cope with this problem. A common solution is to only show labels in a small region of the graph,
for example, within a certain radius of the cursor position. If the density of the display is too high,
a distortion of the visualization may be required (see the next subsection) to provide more screen
space for that section of the graph. An alternate to distortion that sometimes works is to rotate the
graph to reduce the overlap between labels.
Another interesting solution is to only show a random subset of the labels for a short period of
time, and then switch to showing the labels for a different subset. The idea behind this approach is
that the viewer’s short-term memory will enable recall of a larger number of labels as compared
to a static display, especially if this memory is refreshed on a regular basis.
Interactions
Some types of interaction, such as panning and zooming, are common to all types of visualization,
and thus will only be briefly mentioned here for completeness. Others, such as focus+context,
while applicable to a wide range of visualizations, have been primarily developed in the area of
tree and graph visualization and will thus be described in more detail here.

Interactions with Virtual Camera

Interactions such as panning, zooming, and rotation can be viewed as simple changes to the virtual
camera being used to capture a segment of a scene. These allow the viewer to incrementally build
up a mental model of the objects of the scene and their inter relationships. Operations of this type
are often manually controlled, though automated techniques such as data-driven fly-throughs and
spinning of 3D objects can be automatically derived and presented.
Interactions with Graph Elements
Most interactions of this type start with a selection operation, where one or more of the components
of the graph are isolated for some action, such as highlighting, deleting, masking, moving, or
obtaining details. For example, to declutter a graph one might select some nodes and drag them to
a less-occupied section of the screen, while maintaining their links. Similarly, one might select and
move or change the shape of a link to eliminate a crossing or improve the aesthetics of a graph.
Selection may involve a single object, all objects within a specified region or distance, or a set of
objects that satisfy a user-specified set of constraints (e.g., all nodes directly connected to a given
node). One of the biggest problems with selecting elements in a graph occurs in dense regions of
the drawing, where elements are so close together that unambiguous selection is difficult or
impossible. This exposes the need for other types of interaction, such as zooming or the distortion
techniques described later.
Interactions with Graph Structures
There are two classes of interactions that are directed at the graph structure. The first class result
in changes to the structure itself. For example, reordering the branches of a tree may expose
relationships that were not apparent in the original ordering. Redrawing a graph with different
weights on the constraints can generate graphs that make certain tasks easier to perform.
Reordering the columns or rows in a matrix visualization can expose new features or relations
within the data. Techniques within this class are often very specific to the type of graph being
shown. A second class of interactions associated with the graph structure com prises the so-called
focus+context techniques, where a selected subset of the structure (focus) is presented in detail,
while the rest of the structure is shown in low detail to help the viewer maintain context. These
techniques are related to panning and zooming, without the loss of context. The most popular of
these distortion techniques are the many variants on a fisheye lens, where the parts of the
visualization falling within a focal region are enlarged using a nonlinear scaling, while the parts
outside the focal region are proportionally shrunk to maintain their presence in the display. This
dis tortion can be performed either in screen space (i.e., based on pixels) or in structure space (i.e.,
based on the components of the graph). It is the latter case that is more interesting in graph
visualization, as we might, for example, enlarge one branch of a tree while reducing the size of
other branches, or en large all links within three connections of a particular node in order to view
its neighborhood in more detail. An example of structure space distortion.
A technique that can be considered related to both of these classes is that of selective hiding or
removal of sections of the graph. For example, once a branch of a tree has been thoroughly
investigated, the user might want to remove it from the display to provide more space for the
unexplored regions. In a sense, this can be seen as changing the structure (deleting a component),
or as reducing the level of detail for the branch to its root. The terms roll-up and drill-down are
often used to describe the process of hiding and exposing details in a visualization.

Cluster Visualization
1. Definition

Cluster visualization refers to the graphical representation of grouped or clustered data, where
objects within a cluster share similar attributes, and different clusters exhibit distinct properties.
The aim is to enable users to:

● Visually detect natural groupings in data.

● Identify relationships and separations between clusters.
 ● Explore patterns, correlations, and outliers.

2. Principles of Cluster Visualization

Cluster visualization operates on the following principles:

1. Similarity Measurement:
Clusters are formed based on distance or similarity metrics (e.g., Euclidean distance,
cosine similarity).
2. Dimensional Reduction:
High-dimensional data is projected into 2D/3D space using methods such as PCA or
MDS for visual inspection.
3. Encoding Clusters Visually:
o Color: Differentiates clusters.
o Shape and Size: Represent cluster centroids and density.
o Transparency: Conveys overlapping clusters or uncertainty.

3. Visualization Techniques for Clusters

a. Scatterplot Matrices (SPLOMs)

● Show pairwise relationships between attributes.

● Clustered regions appear as densely populated patches in subplots.
● Effective for detecting separations but limited in overlapping clusters.
 b.
Parallel Coordinates

● Each variable is an axis; each record is a polyline crossing all axes.

● Clusters appear as bundles of nearly parallel lines.
● Enhancements like hierarchical parallel coordinates show summarized clusters via
opacity bands.
c. Glyph-Based Techniques

● Represent clusters using composite icons (e.g., Chernoff faces, star plots).
● Each glyph encodes multiple variables for cluster comparison.

d. Vectorized RadViz (VRV)

● Converts multivariate dimensions into binary sub-dimensions.

● Clusters are visualized as separated regions in a circular layout.
e. Aggregated Parallel Coordinates

● Used for large data sets where individual records are replaced with cluster centers and
extents.
● Improves readability while preserving distribution information.

4. Applications of Cluster Visualization

 Application Area Purpose / Use Visualization Techniques
To group genes or proteins by Scatterplot matrices, heatmaps,
Bioinformatics
expression similarity RadViz
Market To identify customer groups with
Parallel coordinates, cluster maps
Segmentation similar behavior
To visualize topics or themes in large Self-Organizing Maps (SOMs),
Document Analysis
text corpora Themecapes
Force-directed layouts with
Network Analysis To detect communities within graphs
cluster coloring
To summarize high-dimensional spaces XmdvTool, hierarchical cluster
Visual Analytics
for exploration visualization
To distinguish anomalies from normal Scatterplots, glyphs, and radial
Outlier Detection
cluster groups displays

5. Advantages of Cluster Visualization

● Enables intuitive detection of structure and anomalies.

● Facilitates multi-dimensional exploration.
● Supports interactive analysis via brushing, linking, and hierarchical roll-ups.
● Enhances pattern recognition in large or noisy data sets.

Metaphorical Visualization
1. Definition

Metaphorical visualization uses real-world analogies or metaphors to represent abstract data

visually.
Instead of directly mapping data values to geometric primitives (like bars, lines, or points), the
visualization borrows familiar physical or conceptual metaphors — such as landscapes,
planets, buildings, or trees — to help users intuitively interpret complex information.

2. Purpose and Motivation

● To enhance comprehension of abstract or multidimensional data.

● To leverage human familiarity with physical environments.
● To provide intuitive exploration in visual analytics.
● To engage users through storytelling and memorable design.

Humans are naturally skilled at interpreting spatial, hierarchical, and physical relationships —
metaphorical visualization builds upon these capabilities.
3. Characteristics of Metaphorical Visualization

Feature Description
Analogical Data attributes are mapped to features of a familiar metaphor (e.g.,
Mapping buildings → entities; height → magnitude).
Cognitive
Reduces learning curve by using intuitive metaphors.
Accessibility
Engagement Encourages exploration through immersive visuals.
Scalability Often limited — works best for moderate data sizes.
Perceptual
Relies on consistent, meaningful mapping (e.g., larger = more important).
Coherence

4. Types of Metaphorical Visualization

a. Landscape Metaphors

● Represent data as terrains or topographic surfaces.

● Peaks = high values, valleys = low values.
● Useful in representing data distributions and similarity clusters.
Example: ThemeScapes visualize document clusters as “mountain ranges” based on topic
density.

b. City Metaphors

● Represent information spaces as urban layouts:

o Buildings = entities (files, classes, webpages).
o Building height = importance, size, or frequency.
o Streets = relationships or hierarchies.
● Commonly used in software visualization and website analytics.

Example: CodeCity visualization of software systems — classes as buildings, packages as

districts.

c. Tree and Botanical Metaphors

● Use natural growth patterns to represent hierarchies or evolutions.

● Branches = subcategories; leaves = data elements.
● Example: “TreeMap” or “Information Garden” metaphors.

d. Solar System and Planetary Metaphors

 ● Used for relationship-centered visualization.
o Central node (sun) = key entity.
o Orbiting nodes (planets) = related items.
o Distance = degree of relationship strength.
● Common in web analytics, bibliometric visualization, and recommendation systems.

Example: “Planetary View of Document Similarity” — Core topic surrounded by related

subtopics.

e. Metaphors in Virtual Reality (VR)

● Immersive metaphors such as “data rooms”, “knowledge galaxies”, and “data

oceans.”
● Allow navigation through high-dimensional data via natural spatial movements.
● Used in visual analytics and large-scale simulations.

5. Advantages

● Simplifies understanding of complex data spaces.

● Engages users emotionally and cognitively.
● Facilitates memory retention and pattern recall.
● Promotes exploratory analysis through familiar spatial organization.

6. Applications
Domain Use Case Visualization Type

Software Engineering Visualizing program structure City and building metaphors

Document Analysis Topic similarity mapping Landscape metaphor (ThemeScapes)

Web Analytics Website hierarchy Solar system metaphor

Biological Data Genetic relationships Tree and branch metaphors

Financial Data Market dynamics Terrain or skyline metaphor

Education and Storytelling Conceptual mapping Virtual or 3D metaphorical spaces