6.2. Results
The results obtained are shown in Table 7. The first column shows the dataset code, the second the error in the
sample using C4.5 (first value) and the error using our approach PostP (second value). In columns 3, 4, and 5,
the results obtained for the dataset with 10%, 20%, and 30% of missing values, respectively, are shown. For each
percentage, the mean error of the experiments is shown in the first row of each dataset, and the standard deviation of
this error appears in the second row.
There are no essential differences between decision trees generated with complete data and those with missing
values; only one more branch could appear in each level but, in general, the size of the trees in both methods is similar.