Skip to main content

Table 2 Comparison of classification results obtained using cross-validation for three different C4.5 parameter settings.

From: Classification tree analysis of second neoplasms in survivors of childhood cancer

Misclassification cost setting 5:1, post-pruning confidence 1%

 

Classified as

Observed

SN

Non-SN

SN

0

34

Non-SN

0

815

Misclassification cost setting 10:1, post-pruning confidence 1%

 

Classified as

Observed

SN

Non-SN

SN

8

26

Non-SN

66

749

Misclassification cost setting 25:1, post-pruning confidence 50%

 

Classified as

Observed

SN

Non-SN

SN

31

3

Non-SN

18

797

  1. The settings comprise of three different misclassification costs, two different post-pruning confidence values, and the default value of 2 for the minimal number of cases in a group parameter. In each table, the number of SN patients being misclassified as non-SN is typed in bold-face. To induce the final tree, presented in Figure 1, we selected a parameter setting that lead to minimal number of such misclassifications.