models. The third interface can be used to compute test metrics on a
previously saved model.
Whereas Chapter 12 was directed to developers supporting business
users, this chapter is directed to users and developers of data mining
Data Mining Tools
The previous chapter focused on practical application implementation.
This chapter focuses more on showing the JDM features for the entire
modeling life cycle, and on a vendor-independent implementation, by
heavily using the JDM notion of capability introduced in Chapter 8.
One purpose of this chapter is to show how some traditional data
mining functionalities, or modules, can be implemented using JDM.
Tools are often aligned with data mining methodologies such as
CRISP-DM (discussed in Chapter 3). This chapter shows how to easily
build tools to cover aspects of:
• Data exploration
• Model building
• Model validation
• Model and task management
To illustrate these, we build graphical user interfaces (GUI). The
purpose of these demonstration interfaces is not to show off fancy
graphics but to be functionally usable. Hence, the user interfaces are
straightforward and simple. We have based these GUIs on Java
Swing classes because Swing is a portable graphical layer; it is rela-
tively simple and powerful. None of the Java Swing classes are
detailed in this implementation; only the code referring to JDM is
commented on in the following sections.
The first user interface is designed to allow basic maintenance
operations on an MOR, such as viewing saved objects like models,
build settings, and tasks on a per class basis. This interface also allows
deleting objects from the MOR and renaming them. DME capabilities
are used to detect what classes can be persisted since implementa-
tions are not required to persist all MOR objects. The second user
interface is designed to show the mining functions supported by the
DME, to allow creating models for each of these functions, and using
the general settings associated with each function. The third user