Information Technology Reference
In-Depth Information
These three approaches to natural concepts can be mapped into Dr. Sadegh-Zadeh's
classical
and
non-classical
concepts. Dr. Sadegh-Zadeh defines the classical con-
cepts as classes “characterized by the 'common nature' of its members, that is, by
a number of properties that are common to all of them.” Dr. Sadegh-Zadeh refers
to this quality of the classical concepts as the “common-to-all postulate.” Thus, a
concept is categorized as classical “if it denotes a category that obeys the common-
to-all postulate.” Dr. Sadegh-Zadeh uses the concept of a square as an example
of a classical concept. The concept of a square is characterized by four properties:
closed figure, four straight sides, all sides equal in length, and equal angles. In
the traditional distinction between artificial and natural concepts, the concept of a
square is classified as an artificial concept since the properties have defining nature.
Accordingly, all members of the class “square” must meet the four conditions. In
contrast, a natural concept has characteristic features rather than defining features.
Dr. Sadegh-Zadeh uses the concept of disease as an example of a non-classical con-
cept. The concept of disease “does not denote a category whose members obey the
common-to-all postulate.”
In general, natural concepts (categories) have two fundamental characteristics:
(1) The members of a natural category do not have to share all features; a natural
category may have some attributes which are common to many members, and some
attributes which may be shared by only a few members; and (2) The members of a
natural category may not be equally representative for a category; thus, the members
may vary in their typicality.
Most medical concepts can be classified as natural concepts rather than classi-
cal concepts. Medical concepts reflect the rapidly expanding and evolving nature
of medical knowledge. They are characterized by a high level of changeability,
context-dependency, and imprecision. Our discussion on the nature of medical con-
cepts has an introductory character, and merely highlights some major issues in the
context of the conceptual models in data mining. Our discussion has also a practical
nature, and it presents an operational definition of obstructive sleep apnea.
13.3
Data Mining and Modeling of Medical Concepts
Data mining is based on a secondary use of existing medical data. Thus, the data are
not purposely collected for data mining, and the meaning of the data should be inter-
preted in the context of the original task. In most cases, medical data are collected
for three distinct reasons: for an individual patient's care, for medical research, and
for patient administration. For the first reason, the data acquisition method is driven
by the diagnostic, prognostic, or treatment process and the data are successively ob-
tained, stored, and used by the healthcare practitioners. Since the intended use of
the data is the patient's care, the data are often incomplete and have varied gran-
ularity. For medical research, data are prospectively acquired through purposely
designed clinical trials, epidemiological studies, or studies of healthy populations.
These data sets are collected by the researchers to answer specific clinical questions
and afterwards are analysed using statistical methods for confirmation or negation
