What We Learned

While there are mountains of data available in databases all over the planet, we need a way to determine any underlying meanings or trends. We examined one of these methods, the ID3 algorithm, which looks at the relative importance of a classification. We can set up our own criteria if we wish - such as region of the globe and language spoken – and let ID3 tell us if these are valid to consider when looking at sales growth. We also set up a two dimensional array to store an entropy value and a classification. By writing a variation of the standard quick sort routine, we were able to sort our array in ascending order based on the entropy value of the classification.

We concentrated a bit on the user interface, by designing a sophisticated system that requires absolutely no user input. The product manager simply points and clicks to get results. We also learned a bit more about the MSFlexGrid control. We used two of these in the program to their best advantage. They can group items together automatically to provide a powerful and compelling way to look at reams of complex data.

Finally, we went beyond the literal approach to database access. Not only did we use ADO recordsets, cursors, and filters, but we also used the data in a new and meaningful way by creating the

ID3

© 1998 by Wrox Press. All rights reserved.