Database Queries, Data Mining, and OLAP

Background

Today’s commercially available relational database systems now routinely include tools such as SQL database query engines, data mining components, and OLAP (Craig, Vivona, & Bercovitch, 1999; Hamm, 2007; Melomed, Gorbach, Berger, & Bateman, 2006; Scalzo, 2003; Seidman, 2001). These tools allow developers to construct high powered business intelligence (BI) applications which are not only able to retrieve records efficiently but also support sophisticated analyses such as customer classification and market segmentation. However, with powerful tools so tightly integrated with the database technology understanding the differences between these tools and their comparative advantages and disadvantages becomes critical for effective application development. From the practitioner’s point of view questions like the following often arise:

• •

  Is running database queries against large tables considered data mining?

• •

  Can data mining and OLAP be considered synonymous?

• •

  Is OLAP simply a way to speed up certain SQL queries?

The issue is being complicated even further by the fact that data analysis tools are often implemented in terms of data retrieval functionality. Consider the data mining models in the Microsoft SQL server which are implemented through extensions to the SQL database query language (e.g. predict join) (Seidman, 2001) or the proposed SQL extensions to enable decision tree classifiers (Sattler & Dunemann, 2001). OLAP cube definition is routinely accomplished via the data definition language (DDL) facilities of SQL by specifying either a star or snowflake schema (Kimball, 1996).

Main Thrust Of The Chapter

The following sections contain the pair wise comparisons between the tools and components considered in this chapter.