Intelligent Enterprise

Better Insight for Business Decisions

RSS
Webcasts
Whitepapers
Subscribe
Home

February 1, 2003

Triple Threat

OLAP, visualization, and data mining in one box

by Guy Alexander

Continued from Page 1

The Table Pane is in the lower right quadrant of the application window. It includes an easy-to-use grid that displays the results of the current query. As with the Visualization Pane, you can move elements directly from the Data Pane's Dimension Tree to the Table Pane. The Table Pane lets you nest multiple dimensions in the rows and columns, and it supports hierarchical drill-down. It also supports two ways of sorting: The Hierarchical Sort sorts cell values within their normal hierarchical groupings, and the Absolute Sort represents each individual value overall.

The Analysis Pane (in the lower left quadrant of the application window) distinguishes PAC from other data mining and OLAP tools. User-friendly wizards unlock embedded data-mining algorithms and "automate" the analysis of multidimensional data. These algorithms reveal valuable insights, and the results are displayed in this pane as a list of cross-tabular descriptions. You can drag the mined result directly into the Visualization or Table Pane to further investigate insights through the OLAP tools I described earlier. Available algorithms in this version include:

Discovery (deviation) analysis finds and ranks interesting data anomalies.
Summary discovers the combinations of dimensions that yield the greatest or least measured values, such as what dimensions yield the largest profit when considering 1, 2, 3, or n dimensions at a time.
Market basket rules find events that occur at the same time for a given set of transactions. A classic example is "What products sell well together?"
Association rules find a set of multidimensional rules that can be extracted from the data and their corresponding support and confidence metrics. For example, "IF customers have a high-school education, THEN it can be said with 93-percent confidence and 20-percent support that they make between $30,000 and $50,000 per year."
Probability rules are also known as classification rules and are useful in a predictive sense. For example, "IF customer age = 25 AND customer gender = male THEN movie preference = action." The results include a probability value that is similar to confidence and support.

What's the Catch?

PAC is an analytic tool, which delivers robust out-of-the-box OLAP functionality to the end user. However, end users come in many shapes and sizes, so what's missing is a documented software development kit that would let developers build applications around the PolyVista objects. Applications could then be developed, catering to the more casual end user's needs.

In the Visualization Pane, options are available to "publish" a chart to Word or Excel, either as a PolyVista control or a picture. However, in the Table Pane, cut and paste is the only available option. Minimally, an export routine should be created for the Table Pane that describes and identifies in Excel the "query " (cube orientation) of the grid being exported.

Rate This Article

Comments:

Optional e-mail address:

The documentation isn't mature, but it's sufficient to guide an end user through PAC mechanics. The Analysis Wizards do a nice job of guiding you through the algorithms, but the documentation needs to describe what the algorithms accomplish. Certainly PAC makes it easy to begin mining data, but you still need education to make it a meaningful and fruitful exercise.

OLAP Features, Too

Although I haven't specifically described them, I want to point out that PAC provides many standard OLAP features including, but not limited to, drill down, drill up, exception reporting, dimension member search, dimension splitters, dimension swapping, extensive attribute and formatting controls, and a lot more.

If you're evaluating reporting or analytic applications based on Microsoft Analysis Services, PAC deserves consideration in your selection process.

Guy Alexander [galexander@ADSconsult.com] is a founding partner of consulting company Applied Decision Support.

RESOURCES

Intelligent Enterprise Business Intelligence Info Center