Ambient Diagnostics - Ambient Intelligence for Scientific Discovery - page 238

Fig. 8. Pattern of the cancer cases displayed in parallel coordinates.
algebraic operations (e.g. equations, integrals), logic models rely on logical con-
nectives (and, or, if-then), often with linguistic parameters, which give rise to
rule-based and knowledge-based systems. Fuzzy logic models can combine both
of these types of modeling via the fuzzification of algebraic and logical opera-
tions. There are three common classes of fuzzy logic models: information pro-
cessing model, which describes probabilistic relationships between sets of inputs
and outputs; control models, which control the operations of systems governed
by many fuzzy parameters; and decision models, which model human behavior
incorporating subjective knowledge and needs, by using decision variables. For
some applications, fuzzy systems often perform better than traditional systems
because of their capability to deal with non-linearity and uncertainty. While
traditional systems make precise decisions at every stage, fuzzy systems retain
the information about uncertainty as long as possible and only draw a crisp
decision at the last stage. Another advantage is that linguistic rules, when used
in fuzzy systems, not only make tools more intuitive, but also provide better
understanding and appreciation of the outcomes.
As more tongue data is available, it would be more appropriate to treat the
extent of each feature value for each diagnostic category as a fuzzy set. The
membership function for this fuzzy set can be computed from the frequency of
each value. This membership value gives an indication of the confidence level
that each value belongs to this set. Hence, the level of overall confidence that
a given case belongs to a particular diagnostic category is the minimum of the
membership values for all features. Since we have not yet obtained a large enough
set of tongue data, we demonstrate this technique using the Iris data example.
Figure 9 shows a 3D parallel coordinates display for this data set. The advantages
of integrating fuzzy sets are two-fold. First, it provides an intuitive match with
the way doctors fuzzily assess the condition of the tongues. Secondly, it is possible
to select the tightness of clusters through the use of an alpha cut plane to discard
those cases whose feature values have too low membership values (i.e. the level
of confidence that a particular case belongs to a specific class is low).