Software for the 4th Dimension
Visual interface lets users easily query time-ordered data.
News Story by Sami Lais
JUNE 04, 2001
Both things happen incrementally over time, says Harry Hochheiser, a computer science graduate research assistant at the University of Maryland in College Park.
Using conventional data mining tools, finding patterns of cell death and stock performance or any other time-series data has been difficult.
Hochheiser's TimeFinder software has a visual interface that lets users manipulate simple graphical tools to query huge databases and then lets them see the results represented on a graph.
One mouse click replaces the typing of multiple parameters, reducing procedure time and the chance of introducing errors.
A user clicks the mouse to draw a box on an empty graph with defined axes, say, with time on the X axis and stock price on the Y. Thus, the user poses a query, such as, "Show me all stocks with prices between $50 and $75 at the end of last September." A second window then shows a list of each stock that meets that criterion.
By drawing more boxes, a user can refine his query to get only stocks whose prices then rose 50% by January and fell by 75% by March.
Andrew Woelflein, a London-based financial analyst, recently saw the software demonstrated. "This is great," he says. "With this, I could query the data directly. Right now, this is the kind of query you submit to the analysts, and they get back to you in maybe a day. You change the parameters, then it's another day. With this, I could get the answer immediately."
To build TimeFinder, Hochheiser used Jazz, an open-source software development kit that was built on the Java2D application programming interface and created at the university's Human-Computer Interaction Lab.
"Harry's work is a great example of the improved visual interface that lets a novice perform at the level of an expert," says Hochheiser's adviser, Ben Shneiderman, the laboratory's founding director and a leader in visual interface design.
Working with Shneiderman, Hochheiser is still developing TimeFinder. He demonstrates a feature he's recently added to the software that lets users superimpose results for comparison. The lines representing performance of most of the stocks follow a similar path. One peak stands out much higher than the others.
"Presenting the information visually allows for serendipity," Hochheiser says. He points out the spike - with conventional data mining tools, there would be no way you'd know about this stock. But now that you know about it, he says, you can ask, "What's different about this one that caused this performance anomaly?"
At SmartMoney.com in New York, Martin Wattenberg, director of research and development, is developing similar software for use on the company's Web site.
Wattenberg says he first watched "how people use the existing query tools. They'd type in the criteria, generate the price graph and immediately look at the price graph."
"I thought, if the graph is what they're interested in, then this works backwards," he says. "So I looked to see how I could eliminate the drudge work of entering the criteria and let people go directly to the graph."
On a blank graph, users of Wattenberg's software do a freehand sketch of a performance curve they seek. The drawing represents a query, which returns stocks whose performance curve matches the sketch.
"Harry's software is a much more sophisticated tool for searching in a more complex way," Wattenberg says. "Ours is aimed at the novice user, letting them quickly find stocks that match a target graph."
Although Hochheiser uses stock prices to demonstrate TimeFinder, the software began with a search by Eric Baehrecke, an assistant professor at the University of Maryland's Center for Agricultural Biotechnology, for a way to interpret data on cell death.
"The type of data we collect is not very different from temporal stock data," Baehrecke says. "Only for us, it's how genes get turned on and off over time: gene expression."
Baehrecke conducts DNA microarray experiments using fruit flies, which have approximately 14,000 genes. The amount of data recorded from the experiments is vast.
"Our brains are not capable of looking at it and making sense of it. We need tools like Harry's software," Baehrecke says. "Other people have development tools to look at this, but they're not as good as they should be."
First, Baehrecke identifies patterns of gene expression in normal cells, then in mutant cells. "Then we overlay the data and compare the two," he explains.
"That's the beauty of data visualization," Wattenberg says. "You nail your method in one discipline, but you immediately have something that will work in another."
Hochheiser's work has been partly funded by a grant from America Online Inc. "But we are actively seeking partners for further development," Shneiderman says.
"I mean, we've talked about biological research and stocks, but the generalization doesn't have to be time-related data. It can be any linear-ordered data. It's equally applicable to something like oil-well log analysis," Shneiderman says. "We're pretty excited about Harry's work."
See your link here
Delivering Information Services Executive Program: Turn your IT strategy into real shareholder value at Harvard Business School
With Hyperion, Visionary CIOs show their companies how to innovate.
Financial Markets Professionals: Help us understand you financial computing needs, and enter into a drawing win an HP Photosmart digital camera
Boost your company's performance with SAP. Learn more at www.sap.com/results
Free Trial Download Webroot�s award winning Spy SweeperTM Enterprise NOW!
EMC - Build your enterprise content management platform.
Riverbed: Consolidation WP
Capture, transform, transport & present your business information
Save Data EMC recovery management - when backup alone isn't enough.
Save 57% on Omaha Steaks Clearance Sale and get 3 FREE Gifts
Get simplified access to people, information and tools with IBM Lotus software.
Microsoft SQL Server 2005. Learn More.
VoIP: Converged voicemail & email for an IP world � free whitepaper
Help Desk How to Choose the Best Help Desk Software white paper
Leader in internet filtering and web security software
Integrate any systems source-to-sale. Click to see the top retailers that do
Get a FREE IDC White Paper on Portals and SOA Portals represent a logical first step in the SOA process. Find out more. Click here
Get JBoss Understand the financial impact of open source. Go!
Linux or Windows? Get the facts.
Copyright © 2005 Computerworld Inc. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of Computerworld Inc. is prohibited. Computerworld and Computerworld.com and the respective logos are trademarks of International Data Group Inc.