FINANCIAL VISUALIZATION SUPPORT FOR BREAKDOWN ANALYSIS

1
FINANCIAL VISUALIZATIONSUPPORT FOR BREAKDOWN
ANALYSIS

• A method to visualize multiple foci polyarchies
  by successive breaking down of visualization
  based on its causalities

Sandeep Prabhakar Muthukumar Thirunavukkarasu Anus
ha Dandapani Ganesh Panchanathan
2
Domain information

• DuPont analysis
• DuPont analysis is a method of analyzing the
  performance of a company by means of various
  Ratios
• E.g. Return on Equity, Net Income, Total Equity
• The data
• The data consisted of the annual statements of
  100 companies for the last 20 years. Each company
  had 180 data items.
• These data items are used in calculating DuPont
  Ratios

3
What is a polyarchy?

• Polyarchy multiple hierarchies

4
Different aspects to breakdown

• Breakdown based on attributes
• Decomposition based on predefined formulae
• E.g. Frequency wavelength / c
• Frequency R(wavelength, c)
• Breakdown based on values
• Decomposition and grouping based on distinct
  values
• State R(Virginia, California, .)

5
What are the problems?

• Hierarchy Visualization at each node?
• Multiple Foci
• I want to see B and C siblings side by side
• B and C are in different hierarchies
• I want to see A (ancestor) and J (a deep
  descendant) side by side
• ltA,Jgt not in each others context
• Top Down breakdown of data sets
• I want to see which sub division affects the
  superset causality

6
Our solution Breakdown Visualization
7
Navigating Hierarchies
8
Interactions Supported

• Fix and move technique
• Brushing and linking
• Details on demand
• Drill down
• Comparison of two nodes not in each others
  context
• Visualization type can be selected that best
  suits the data
• Implicit representation of hierarchy

9
System Diagram
MS SQL database queried with JDBC
10
Future Work

• Allow a knowledgeable user to dynamically modify
  and store new hierarchies
• Implicit hierarchy may not be suitable in all
  cases

11
Results

• HCI metrics
• Scale
• Tends to infinity as user selects the
  visualizations to be currently seen
• User performance
• Every click accomplishes work!
• Learning time
• Good
• Error rate
• Going down the wrong path problems? Yes, Easy to
  recover from it
• Retention time ???
• User satisfaction ???

12
Acknowledgements
Nathan Conklin (bowls for independent study
students) Dr. Chris North Dr. Raman Kumar for
data access