The MiningMart Approach
MiningMart provides support for knowledge discovery applications. Thus the system is aimed at those people in an institution who actually work with the institution's data and process it in various ways in order to gather statistics or other higher-level information. While the system provides an intuitive access to data and easy handling of processing steps, users should have a certain knowledge about how their data is stored before the application of MiningMart.

MiningMart works with relational databases. It assumes that all input data is given in tables in a relational database and its output are new tables in this database. It also stores its own data in relational tables. Thus, there are no limitations to the amount of data that MiningMart can handle.

Referring to the definitions at the beginning of this chapter, MiningMart supports the whole knowledge discovery process but focusses clearly on preprocessing. That is, the system provides a few common data mining algorithms which can be applied directly from the system, but its main value is the support for the technical steps that are needed to bring the data into a format which can be used for data mining. Like the input, the output of the system is a number of relational database tables, but in the output tables the data is stored in a representation suitable for data mining. Thus, you can use your favourite data mining algorithm easily because the input data for it is stored in a table in your database in exactly the right format after the application of MiningMart.

MiningMart supports preprocessing by applying a number of data processing steps to its input. Each step is graphically represented in the MiningMart workspace. The complete sequence of steps is stored in the database and can also be exported to other sites where MiningMart is in use. In this way, a documentation of the whole knowledge discovery process is achieved. All the details of a discovery process can be easily saved for later usage, can be modified using a graphical user interface, and can be transferred from one discovery process to another.

MiningMart uses a layer of abstraction of the actual data to model the knowledge discovery process. This abstraction allows to publish successful discovery applications for the benefit of other users, while sensitive details are hidden. This means that you can benefit easily from the work done by other MiningMart users. The MiningMart web pages provide a central platform for the exchange of successful discovery processes, called cases. On this platform, such cases are described both in terms of their relevance to a business and in technical terms, which allows you to find cases which are similar to the application you have in mind. You can then download such cases into your MiningMart system and make the necessary modifications towards your own data.

The following topic describes these central ideas in more detail by explaining the basic MiningMart terminology. Once you have become familiar with those basic notions, you can start your own MiningMart application easily.