The Hindu Business Line : Patterns hiding in mountains of data

LOOK at these recent reports. If the Food and Drug Administration of the US ties up with major health insurers and reviews patient histories, such a study can reveal possible safety red flags about drugs in use; but according to Harvard Medical School professor Jerry Avorn, cited in www.boston.com, such data mining would cost about $500,000 per drug.

A tax compliance software called GoSystem has a data mining engine by the name FormSource; it helps identify planning opportunities from the simplest to the most complex returns, as www.webcpa.com informs.

A few weeks ago, Fujitsu and France Telecom agreed to launch a joint research project on grid computing technology. According to www.rednova.com, the first phase of the project will focus on facilitating analysis of France Telecom's huge volume of data, such as for data mining and customer billing.

The common thread in the above stories is `data mining'. The topic, therefore, is hot; and so is the second edition of Data Mining Techniques by Michael J.A. Berry and Gordon S. Linoff, published by Wiley Dreamtech India P Ltd (www.wileydreamtech.com) .

The book is aimed at data mining practitioners, not software developers, clarifies the intro. Therefore, "ideas are presented in non-technical language with minimal use of mathematical formulas and arcane jargon."

While data warehouse provides the company with a big memory, you achieve intelligence only when you are able to comb through the bits, notice patterns, devise rules, come up with new ideas, figure out the right questions, and predict the future; to do which you need data mining, advise the authors.

Data mining comes in two flavours, not shallow and deep, but directed and undirected. The former "attempts to explain or categorise some particular target field such as income or response".

The latter is more challenging; it finds patterns among groups of records "without the use of a particular target field or collection of predefined classes."

For their target audience - that is, those in marketing, sales, and customer relationship management - the authors provide numerous real-world examples.

Here is one such, to explain why data may be at the wrong level of detail when using declining customer usage as an early warning before the customer leaves a cellular operator. "For seven months, the subscriber used about 100 minutes per month. Then, in the eighth month, usage went down to about half that. In the ninth month, there was no usage at all."

If you inferred that the customer switched loyalty only in the ninth month, you may be wrong. "Looking at minutes of use by day instead of by month would show that the customer continued to use the service at a constant rate until the middle of the month and then stopped completely, presumably because on that day, he began using a competing service."

"If you torture data sufficiently, it will confess to almost anything," is a Fred Menger quote. But with Data Mining Techniques to help, you will need to torture data less, to know more.

`Secrets' of developing accounting package

WE know that accountants live in a different world, disclosing their secretly sacred work in a format that many don't understand. To know their minds, try a different route: Learn the `secrets of developing an accounting package'! Which is what Bharati and Krishna reveal in Database Programming Using VB.Net & SQL Server 2000, from VK Publishers (www.vkinfotek.com) .

The first chapter running to about a hundred pages explains the components of .Net framework, user interface elements of VB.Net, and OOP implementation.

What follows is a crash course in accounting. Double entry lies at the core of accounting, say the authors, and more flatteringly for bean-counters, it is extolled as "the finest discovery of human intellect."

Auditors should particularly focus on the inputs about SQL, so they can issue a command that may possibly read: USE FinAccounting SELECT *FROM AccountsTable WHERE AccountName LIKE `Osama%'

There are only four types of `new accounts' - viz. creditor, debtor, bank and general. This may not gel with how your munimji operates, yet the discussion that runs across the book guides one to create a chart of accounts, develop `dialog' boxes, `filter' to make subsets of datasets, build a `tran class' as field vs transaction matrix, and so forth.

To the question, "what is a transaction?" an accountant may respond that it is the financial interaction between people.

Not so the duo, Bharati and Krishna: "A transaction is a series of actions that must either succeed, or fail, as a whole. If one of the actions fails, then the entire transaction fails and all the changes made to the database so far, must be reversed or rolled back."

The book, I suppose, can serve a dual purpose: Techies can know how accounting works, and likewise, accountants too get into the minds of system developers.

Tailpiece

"They found that the criminal used a laptop to track his victim and murder... "

"Oh!"

"And the case has been transferred to the IT cops to find the criminal's address from the hard disk!"

Books2Byte@TheHindu.co.in

Article E-Mail :: Comment :: Syndication :: Printer Friendly Page

Stories in this Section
`I'll be there in a jiffy'

Another precious drop...

Yahoo in Outlook Express

Quiz

Patterns hiding in mountains of data