Decision Tree Induction - Gene Expression Programming - page 362

a.

0123456789012345678901234567890
ABbBbaFDCIbababaaabbb3894830502

C = {0.297699, 0.667999, 0.392853, 0.896088, 0.149322,
0.392120, 0.183807, 0.289062, 0.461792, 0.235870}

b.

CLUMP_THICK

d > 0.896088 0.896088

UNI_CELL_SIZE Mal

d >

0.461792

0.461792

UNI_CELL_SIZE Mal

d >

0.149322 0.149322

Ben BARE_NUCLEI

d >

0.297699 0.297699

MARG_ADHESION UNI_CELL_SHAPE

d d >

0.39212

>

0.39212 0.297699 0.297699

MITOSES Mal Ben Mal

d >

0.392853

0.392853

Ben Mal

Figure 9.15. Model designed by the EDT-RNC algorithm to diagnose breast
cancer. a) The linear representation of the decision tree. b) The corresponding
decision tree. This model is extremely accurate and generalizes outstandingly well
(it has a training set accuracy of 97.43% and a testing set accuracy of 98.28%).