The Exposition of Fuzzy Decision Trees and Their Application in Biology - Advancing Artificial Intelligence Through Biological Process Applications - page 388

For the path, SLW = M, the subsethood values
associated with SKW (the paths L, M and H), to
the possible classification to AGE = L, M or H,
are reported in Table 7.
The results in Table 7 relate to the progression
from the main SLW = M path to the SKW = ⋅ paths.
The only largest concomitant subsethood value
greater than the 0.720 minimum truth threshold is
associated with SKW = H and consequent AGE =
M (value 0.797). It follows, the defined path is SLW
= M then SKW = H, which classifies to ADT = L
with 0.797 level of truth. For the other two paths
from SLW, there would be further checks needed
on the possibility of minimising ambiguity by
further augmenting other previously unconsidered
condition attributes to the respective path.
This described process for FDT construction
is repeated until all paths end with a leaf node,
found from either achieving the required truth
level of subsethood defined, or that no further
augmentation is possible. The resultant FDT for
the abalone problem is reported in Figure 6, which
as in the tree found from the tutorial analysis,
shows all the fuzzy decision rules constructed
(each rule denoted by a path down from the root
node to a leaf node).
The reported FDT in Figure 6, constructed
from the analysis given previously, is shown to
have eight levels, indicating a maximum of seven
condition attributes are used in the concomitant
condition parts of the fuzzy decision rules derived.
Indeed, only the attribute LTH (longest shell
length) is not used in the FDT shown. There are
in total 19 fuzzy decision rules, R1 , R2 , .., R19 ,
shown to be constructed. Of these 19 rules, three
of them have an associated truth level labelled
with a *, which indicates they have a truth level
bellow that desired, since no further augmenta-
tion of other condition attributes was able to
still lower the classification ambiguity. One of
the decision rules is next explained more fully,
namely R11 , directly from the FDT the rule can
be written down as;

'If SLW = M, SKW = L and SEX = F then AGE
= H with truth level 0.653.'

Re-written as in a more linguistic form,

'If Shell weight is medium, Shucked weight is low
and Sex is female then Age of abalone is high,
with truth level 0.653.'

Also through using the defining values associ-
ated with each condition attribute,

'If Shell weight is between 0.17 grams and 0.30
grams, Shucked weight is below 0.24 grams and
Sex is female then Age of abalone is above 11.5
years, with truth level 0.653.'

Further interpretation can be given to any of
the fuzzy rules defined in the FDT presented, as
well as the whole FDT itself.

FDT Analysis of Greater Horseshoe
bats

This section presents a second biology based FDT
analysis. Motivation for this analysis comes from a
study investigating patterns of torpor and arousal
from hibernation in free-ranging bats (Park et al .
2000). Moreover, temperature sensitive radiotags
were used to monitor the body temperature and
movements of Greater Horseshoe bats, Rhino-
lophus ferrumequinum for an average of 87±29
days (mean±SD). These measurements, amongst
others, were used as the characteristics associated
with the torpor bouts undertaken by the bats.
Many species of small, temperate zone mam-
mals spend much of the winter in hibernation,
minimizing their energy expenditure when there
is little or no food, and when ambient tempera-
tures are generally low (Lyman et al ., 1982). All
hibernating mammals studied, however, arouse
periodically throughout winter and much atten-
tion has focused on why arousals occur since it
is energetically expensive to do so (Thomas et