We also have collected
results on the usage of the default rule while using Minimum OSR in Tables
12 and 14, which
show the percentage of testing observations covered by the default rule
when using Minimum OSR
(with and without 
-test, respectively), and Table
13 and 15,
which show the prediction accuracies of the default rule on different data
sets.
Table 12: Mean Coverage (%) of the Default Rule in BruteDL
with Minimum OSR (with 
-test) on Different Data Sets
Table 13: Mean Prediction Accuracies (%) of the Default Rule in BruteDL
with Minimum OSR (with 
-test) on Different Data Sets.
``N'' Means the Default Rule is Never Used.
Table 14: Mean Coverage (%) of the Default Rule in BruteDL with Minimum OSR
(without 
-test) on Different Data Sets
Table 15: Mean Prediction Accuracies (%) of the Default Rule in BruteDL with
Minimum OSR (without 
-test) on Different Data Sets.
``N'' means the default rule is never used.
As can be seen in Table 12 and Table
14, the default rule is rarely used on all the
data sets when including Minimum OSR as part of the default
strategy
. Most of the testing
observations that are not covered by the existing rules in the decision list
are classified by the rules learned by the Minimum-OSR.
