Knowledge Discovery Mine - Tons of links related to data mining. Contact Mark Gibbs

By Mark Gibbs
Intranet, 3/24/97

The idea of data mining on intranets is taking on star status. And if the data mining field had the equivalent of the Academy Awards, Level Five Research, Inc.'s Quest Version 2.5 would be nominated for best supporting tool.

An intranet is first and foremost a generalized data collection and distribution system. But how do users find what they want in the collection of data? While general-purpose searching is available in any industrial-strength database product, there haven't been many choices for making decisions based on how well the data fit specific criteria.

This is an area in an intranet environment to which Level Five's Quest Server, a back-end Web server application, brings tremendous value.

The Quest system comprises the Quest Desk-top, which IT staffers and savvy users can employ for data mining and Web content construction; and the Quest Server, which provides sophisticated intranet search facilities through a Web server.

Installation is straightforward. The Quest system comes with its own Web server or can be installed on a number of popular Web servers. I recommend the latter option for a production environment.

Here's a an example of how Quest Desktop and Quest Server, a Common Gateway Interface application, might be used on an intranet.

Consider a company that has a vendor- and product-tracking system using a database that complies with Microsoft Corp.'s Open Database Connectivity (ODBC). In addition to using an ODBC-compliant database, the IT staff has developed Web server scripts that accept input and updates from Web browser forms. The problems IT faces now are making the system accessible via the intranet and allowing users to decide which vendors provide the best combination of price and reliability.

Prepare for stardom

In this example, we use the sales database and from the table select the record number, product code, quality, price, reliability and warranty fields. We are then presented with a table of our data.

In the fields beneath each column header, we can enter target values for which we want to allow searching. If we enter 'MAX' in the value field of, say, the Reliability column, the table will be sorted with the highest value at the top.

Quest Desktop allows the relative score of each cell in a column to be displayed in the cell. Grading icons make the ranking of values more visual.

We also can factor in the quality measure. We need to assign relative importance to text strings, such as 'Grade 1' and 'Grade 2,' if our quality measure uses phrases rather than numbers. Quest supports this through value maps.

To refine our criteria even further, we can assign relative importance to the column for which we have set target values and value maps. If our primary concerns are reliability and quality, for example, we could adjust their weights to each other and to price.

It's show time!

The default presentation of the data at the browser is a table with the database fields listed as column headings. Below each heading is an entry field for setting the target value and another for assigning the importance of the field.

Using the Quest Desktop, we can control whether the target and importance fields for a given database field are displayed through Quest Server. This is important for fields that are not intended to be part of the search criteria but are required for display purposes.

When we set the target values and importance of one or more fields and click on the Quest! button, the database is reordered and returned to the browser. If we disable automatic searching (the default state), the first time the user retrieves a project the records in the database will not be displayed.

This is useful because simply presenting the field names, target values and importance means users won't be distracted by the unranked data. With Quest, we also can specify 'HTML wrappers' to create headers and footers for the pages so we can add instructions and graphics.

It also is possible to drive the search specification from an HTML form, instead of the default table display, so we can improve the look and feel of the user interface. What's more, URLs can be built into the results.

A star is born

Level Five stands out for the multimedia training it includes with the Quest system. The company uses 11 Lotus Development Corp. Screen-Cam presentations with a 66-minute run-time to cover basic information and advanced topics related to Quest Desktop and Quest Server.

For a product with the richness of Quest, these tutorials make a huge difference in how fast a user can learn the system. The paper manuals are adequate, though somewhat repetitious.

I am impressed with the Quest products, as they provide sophisticated, flexible, user-friendly database mining in an intranet environment.