InfoVis 2005 Contest
GeoViKE: GeoVisualization for Knowledge Exploration

Contest webpage:

Authors and Affiliations:

• Geng Sun, School of Information Sciences and Technology, Pennsylvania State University, gengsun@gmail.com
• Xiaolong Zhang, School of Information Sciences and Technology, Pennsylvania State University, lzhang@ist.psu.edu

Tools:

We have developed GeoViKE, a geo-visualization tool to help people analyze and understand the technology trends in the United States. Raw data were first imported into Microsoft Access, and then queries were created to provide various data sets of interest for GeoViKE to process. These data sets characterize company moving and the temporal and geographic distribution of sales in all industries. GeoViKE is Java-based and built upon Piccolo, a zoomable user interface toolkit. In addition to this webpage, we also provide a two-page summary and a video demo.

Tasks:

1. Characterizing Company Moving
2. Understanding Industry Sales Distributions at Different Geographical Levels
3. Unusual Industry Sales
4. Understanding Industry Composition of States and Industry Distribution among States

TASK 1: Characterizing Company Moving

Task 1.1: Understanding Where Companies Moved In and Out

• Process:
  We created a query to find out companies with headquarters located in different states within two consecutive years. These companies were regarded as moving from one state to another. Then, the numbers of companies moving in and out of individual states can be calculated and visualized as circles in different sizes on the top of each state in a US map. Usually, the size of a circle is proportional to a number, unless when a circle is very small, it is enlarged to be visible. Users can check company moving in or out in any industry or industry combination. Topping the cursor over a circle can bring out detailed information about what state the circle represents and how many companies moved in or out.
  
• Image 1.1:
  
• Insight 1.1:
  These three images show the moving of companies in all industries (Users can actually pick a particular industry or a combination of a few industries). As seen, California, Texas, Florida, and New Jersey were the popular destinations, which more than 200 companies from various industries chose to move into. Among these states, California had the most, 380 companies.
  
  California was also the state that lost most companies, totalling 582, within the same time period. States like New York, Massachusetts, and New Jersey also lost many companies.

Task 1.2: Understanding Why Companies Moved

• Process:
  To further understand reasons that may cause a company to move, we chose to analyze three factors provided in raw data: whether a company was in different industries before and after moving, whether its sales was increased, and whether it got more employees. We speculated that these three criteria may be related to company moving.
  
• Image 1.2:
  
  
• Insight 1.2:
  In computer hardware and software industries, companies were more likely to move into states like California, New Jersey, Texas, and Florida.
  
  Among about 150 companies moving into California, almost half, 74 companies, achieved better sales right after the movement; 57 got more employees; and only 5 companies changed industry. Thus, the primary reason for moving might be for more sales or more employees. Actually, close to one-third of companies, 44, had both more employees and better sales. Among these 44 companies, only two changed from a different industry.

Task 1.3: Understanding Company Moving between States

• Process:
  To understand cross-state company moving, users can analyze where they were from, and what may cause their move.
  
• Image 1.3::
  
• Insight 1.3
  Among 380 companies moving into California, Texas contributed most. Down to a particular year, for example in 1996, three companies moved into California from Texas, and two of them achieved better sales.

TASK 2: Understanding Industry Sales Distributions at Different Geographical Levels

• Process:
  Users can also examine industry distribution patterns, in terms of sales, in different years. Sales data can be from one industry or a combination of industries. The distribution information is provided at different geographical levels, from country, region, state, to county. Data at the country level was obtained by adding data from all states. We grouped states into 6 regions. Data of a region can be obtained from corresponding state members. Companies were also grouped into different counties according to their zip codes. Information about zip codes and counties was obtained from a commercial database. GeoViKE allows users to easily zoom in and out to get information at different scale levels.
  
• Image 2:
  
• Insight 2:
  As seen, sales from computer hardware and software industries in 1994 dropped significantly from 1993. At the regional level, both the Mid-Atlantic and the West regions outperformed others.
  
  Down to individual states, we can find California, New York, Texas, and Massachusetts did much better than the rest. Sales from states like Alaska, Hawaii, Montana, Wyoming, and so on were very low. Computer industry, hardware and software, seemed to cluster on the west coast and in the Great-Lake states, in terms of industry sales.
  
  Distribution can also be analyzed at the county level within a state. For the State of Arizona, two counties, Maricopa and Pima, performed much more strongly than others.
  

TASK 3: Unusual Industry Sales

• Process:
  We used parallel coordinate tools to visualize the trends of industries. Each industry is a polyline. Years are parallel coordinates. Users can choose to examine how a parameter (e.g., sales, sales per employee, the number of employees, the number of companies) changes in years.

• Image 3:
  
  
• Insight 3:
  Sales for computer hardware and telecommunication peaked in 1993, and then went flat after. Sales in the telecommunication industry bounced back a little bit after 2000. This may imply a market boom in the industries of hardware and telecommunication in 1993 and then a quick correction thereafter. Although the total sales from the industry of Not Primarily High-Tech seemed much higher than other industries, the average sales per employee in this sector was not particular high. This might be due to a much larger work force in this sector
  
  The energy industry performed consistently better than others, in terms of sales per employee.
  
  A big jump is seen in the total sales of the non-high-tech sector from 2000 to 2001. However, no jump is seen in sales per employee. Examining the change of the company number, we found that in 2001, more than 2,500 companies were added into this industry, and this dramatic increase of company number might contribute to the jump of total sales.

Task 4: Understanding Industry Composition of States and Industry Distribution among States

Task 4.1: Understanding Industry Composition of State Economy: Top Three Industries in A State

• Process:
  With available sales data from 1989 to 2003, we aggregated sales for each state and each industry. GeoViKE ranks all industries based on sales. For a state, top three industries are chosen to create a pie chart to show the industry composition of the state. A pie chart has four pieces: three stand for the percentages of top three industries out of the total sales of all industries, and one for the percentage by all other industries. Each piece of the pie is color-coded for different industries (white color is for all other industries). By moving the cursor over a particular piece, users will see the exact percentage number of an industry. Bar charts in another window provide more detailed information in all 50 states for comparison. In addition to picking a state from a pull-down list in the task panel, users can also drag the slide at the bottom of the bar chart window to quickly see industry compositions in different states.

• Image 4.1
  
• Insight 4.1:
  Economy in states like Arkansas, Delaware, Nebraska, and South Dakota was very concentrated, and sales from top three industries accounted for more than 90% of total sales of all industries. Economy in Arizona, Iowa, Louisiana, Massachusetts, Missouri, New Mexico, and Pennsylvania was more diversified, because sales from top three industries was less than 50% of total sales. In Arkansas, California, Delaware, Maine, Michigan, Mississippi, Montona, Nebraska, and West Virginia, one particular industry strongly dominated.
  

Task 4.2: Finding Other States with Similar Industry Composition

• Process:
  After users choose a particular state to examine its industry composition, the tool also searches other states that have the same top three industries. If such states exist, users can choose to see them all together.

• Image 4.2
  
• Insight 4.2:
  Knowing the composition of top three industries of a particular state, such as Arizona, we can find three other states, Ohio, Alabama, and South Carolina, have the same top three industries.
  

Task 4.3: Understanding Industry Distribution in the United States

• Process:
  GeoViKE ranks all states, also based on sales, in each industry. Top three states are chosen in the visualziation of industry distribution in the United States. After users pick an industry, circles will appear above the top three states on the map. The size of a circle indicates the percentage of the sales from a state out of total sales in an industry.
  
• Image 4.3
  
• Insight 4.3:
  In industries like biotechnology, computer hardware, pharmaceuticals, and photonics, top three states contributed more than 70% of the total sales. However, sales in industries like factory automation, medical, subassemblies and components, and test and measurement were very dispersed. Top three states only make about 30% of total sales.
  
  Among total 18 industries, 13 had California in the top three. California was dominant in computer hardware, software, as well as subassemblies and components.
  
  Minnesota was the leading state in biotechnology; New Jersey in pharmaceuticals; and New York in photonics.
  
  Pharmaceuticals industry clustered in three Mid-Atlantic states: New Jersey, New York, and Pennsylvania. Photonics industry clustered in New York and Massachusetts.