In addition to the way of reasoning, there is also a question of how much help the system is retrieving besides simply being given the database. The concepts of supervised and unsupervised learning are helpful at making this clear.
As stated in [DF95], the basic notion of supervised
learning in data mining is that of the classifier. A classifier is a
component (in this case our data mining system) which for a given
input is able to classify it with respect to some kind of
classification. For a system to be using supervised learning, a
teacher must help the system in its model construction by defining
classes and providing positive and negative examples of objects
belonging to these classes. The system is then to find out common
properties of the different classes, and what separates them, in order
to make correct classification for other objects. The resulting rules
say that for certain values of the condition
attributes, the
resulting decision attribute has a certain value.
In the case of unsupervised learning, no teacher defines the classes a priori. Thus, the system itself must find some way of clustering the objects into classes, and also find descriptions for these classes. The resulting rules from such a system will be a summary of some properties of the objects in the database: which classes are present and what discerns them. This will of course only be what the system has found as most prominent, but there may be many other ways of dividing the objects into classes, and many ways of describing each class.