- Tetherless World Constellation

Report
RESULTS
Using d3.js For Visualization of
Corporate Board Membership
Alexei Bulazel1 ([email protected]), Bharath Santosh1
([email protected]), James Hendler1 ([email protected])
1Tetherless
World Constellation, Department of Computer Science,
Rensselaer Polytechnic Institute
INTRODUCTION
The OrgPedia Open Organizational Data
Project will use Linked Data technologies to
present a free, not-for-profit online repository
of data on companies, drawing upon
government databases, private open
databases, and public contributions. As an
early-stage prototype for this project we
created an interactive data visualization
displaying links between corporate board
members of notable companies,
demonstrating the inherent value of interlinked
corporate data that OrgPedia seeks to expose.
METHODS
The visualization shows the connections
between board members of major US
1. Given a list of 314 stock tickers of interest, we wrote a Python script to access an API from
Companies.
LittleSis.org to download data on the board of directors for each company.
OBJECTIVE
2. Data on 2800+ unique board member from these company boards was then downloaded from
LittleSis.org.
Our objective for this project was to leverage
the d3.js ("Data Driven Documents")
JavaScript visualization framework and free
data from LittleSis.org, a “free database
detailing the connections between powerful
people and organizations”, to create an
interactive web-based visualization of the
interconnected nature of corporate boards in
the United States.
3. A graph of board member was created, with members as nodes, and mutual board affiliation as
links.
Links:
D3.js visualization framework:
http://mbostock.github.com/d3/
LittleSis: http://littlesis.org/
Scan QR for online copy 
4. Graph algorithms were used to pair the graph down to a more manageable ~400 nodes of the
most highly connected members.
a. Companies not possessing a member connected to at least four companies on the graph
were removed, leaving only the most well connected boards, with about 200 members in
total.
b. To then make the graph more interesting, removed companies with boards containing
members sharing affiliation with a company already on the graph were added back in with a
12% probability (this percentage was chosen so as to arrive at a manageable final number of
board members in the graph).
5. Final formatting processing was applied, and the information in the graph was output.
6. The graph JSON (“JavaScript Object Notation”) file was input to the d3.js framework along with
styling information for visualization.
• Our work yielded an interactive,
informative, web-based visualization that
can be used to illustrate the interconnected
nature of US corporate leadership.
• The visualization allows users to easily
get a broad overview of the structure with
one quick glance, while also providing them
with the possibility for deeper exploration of
the data.
• The d3.js force directed graph framework
allows users to drag the nodes as they
please to pick out particular parts of the
graph.
• Board member nodes are colored by
affiliation for easy identification, and scaled
in size according to the number of boards
that they sit on. Members on two or more
boards have their names displayed next to
their node.
• For more information, nodes can be
hovered over to see name and corporate
affiliation, and clicked to view the
individual’s page on LittleSis.org
DISCUSSION
• The d3.js framework provides an easy
way to create interactive data visualizations
quickly and easily.
• We were limited in the size of our
visualization by computer speed and
browser efficiency – the d3.js visualization
is all web based and uses modern
technologies of HTML5, CSS3, and
JavaScript, but can be slow even in the
fastest browsers. Our pruning of ~2400
board members from our total data pull was
mainly done for performance reasons.
• Freely available information from
LittleSis.org, and in the future, from the
OrgPedia project allows for easy creation of
visualizations like this one, that can give a
good broad overview of issues to
individuals without extensive backgrounds
in the subject matter.

similar documents