News Stay informed about the latest enterprise technology news and product updates.

Credit risks are sniffed out with help of data mining tool

Japan's largest corporate research company has developed a statistical model to help lending institutions minimize their risk in writing corporate loans.

Even in a good economy, banks have little margin for error when it comes to deciding who gets a corporate loan. In an unstable economy, where a sudden change in the market can greatly influence a company's chance of bankruptcy, banks need all the help they can get in making these high-risk decisions.

Teikoku Databank (TDB), Japan's largest corporate research company, recently developed a statistical model for lenders that predicts which potential corporate borrowers may go bankrupt within a year. The model was created with SAS Institute's data mining and statistical analysis tool, Enterprise Miner.

Katsushige Iwabuchi, a member of the TDB development team, said several vendors were considered, but Cary, N.C.-based SAS was chosen because its product met all of the company's key requirements: speed, scalability, user-friendliness and a high level of technical support. "Speed is absolutely essential for analyzing large volumes of data," said Iwabuchi.

Iwabuchi also said the software had to be intuitive and easy to use because TDB needed to implement it quickly.

Descriptive and predictive modeling

Anne Milley, director of data mining at SAS, said Enterprise Miner was able to meet those requirements because of its greatest strength.

"The main talent of the product is that it has both descriptive and predictive modeling capabilities," according to Milley, who said she often finds that Enterprise Miner customers use both techniques because it provides more thorough results.

To get the answers it needs, a bank can look at its data, look at different attributes about the data and see what kinds of things are highly associated with bankruptcy, especially over time, said Milley. Relevant data can include things such as location of a company, industry, company size or even survey data, said Milley.

To demonstrate how the tool is used, Milley gave an example of a company that is getting into online retailing and might want to borrow money to launch a new venture. One thing a potential lender might want to check is the concentration factor, which is simply a determination of whether the market is saturated with similar ventures. If that is the case, there might not be enough profit to go around, and this could be an index or ratio that the bank could add into its predictive model that could influence the success, or lack of success, of the company.

"With the Teikoku Databank, where they're trying to predict bankruptcy, they have more than enough data," said Herb Edelstein, president of Two Crows, a Potomac, Md.-based data mining consultancy. "So the question becomes: "How do I focus attention on what is the important piece of data, the differentiating piece of data to make these predictions?'

"And I think that's why CRM requires data mining, that's why bankruptcy prediction requires data mining and that's why data mining is growing by leaps and bounds in the CRM arena."


The Best Web Links on data mining

Major issues in data mining

Data mining: Digging user info for gold

Dig Deeper on Customer data management

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.