Gene Expression Profiles for Predicting Metastasis in Breast Cancer: A Cross-Study Comparison of Classification Methods

Table 2

Number of features in the models.

Dataset	AM	RO
Method	(𝑛=151)	(𝑛=286)
RF	21	21
LR	5	11
R-SVM	20	25
L-SVM	8	11
P-SVM	4	7
S-SVM	17	35
NNET	21	16

AM and RO are the Amsterdam and Rotterdam training sets and 𝑛 shows the number of samples in the respective datasets. Methods used are as follows: RF: random forest, LR: logistic regression, R-, L-, P-, and S-SVM: support vector machine with a radial basis function, linear, polynomial, or sigmoid kernel, and NNET: neural network with a single hidden layer.