GM To Make Major Data Mining Investment
ACTION ITEMS

Detroit, MI -- As Information Week reported, General Motors Corp. is currently evaluating several data mining products for projects the company plans for this, and the coming year, involving several GM divisions, including its GMAC loan division and its credit-card division.

GM has experimented with data mining on a few scientific projects, but only in the past two years has the automaker tried to use data mining in production work to solve problems like database marketing, using customer data and warranty information, says Ramasamy Uthurusamy, director of emerging technologies in GM's information systems and services division.

"Some people think data mining is getting software and starting to run some programs, but it's not," explains Uthurusamy. "The biggest difficulty in implementing any data mining strategy is trying to extract clean and reliable data before you do the mining. We have spent a lot of money on that alone."

GM's first pilot, planned for rollout late this year, is the mining of warranty information. "What we'd like to learn from warranty data is how to identify problems early enough, so we can go back stream and fix them," says Uthurusamy. "For example, if I can learn that a particular [car] part is going bad in a specific vehicle, I can go back to the supplier that provides the part, and avoid future warranty expenses."

Since the first of the year, Uthurusamy has been testing three data mining products:IBM's Business Discovery Solutions, which includes its Intelligent Miner product; SAS Institute's Enterprise Miner, slated to ship in April; and Silicon Graphics' MineSet.

Uthurusamy says each product has strengths. "If our business problem requires a good visualization tool, Silicon Graphics' MineSet has that strength," he says. "If I'm looking for associations between different variables, IBM's Intelligent Miner works well. And if I'm looking at statistical techniques, SAS's Enterprise Miner is useful." GM has IBM DB2, Oracle, and SAS databases. "We won't ever have a situation where we'll use one tool, but perhaps a combination of each."

Uthurusamy continued: "Data mining is very important to GM. We are still in exploratory stages. But we hope to seriously start investing (in data mining) once we show some results."