Assessment of performance of survival prediction models for cancer prognosis

Table 6 The numbers of misclassifications for five binary classifiers using the support vector machine (SVM), random forest (RF) and logistic regression (LR) classification algorithms

	Survival Threshold	Risk	Number of Training	Number of Test samples	Number of Misclassification
	Survival Threshold	Risk	Number of Training		A	B	C	D	E
	Year	Group	samples		A	B	C	D	E
SVM	4	high	28	10	10	6	7	7	7
		low	50	9	1	2	3	2	3
	5	high	34	12	9	1	3	2	5
		low	44	7	1	2	2	2	2
	6	high	43	14	5	2	3	2	2
		low	35	5	1	1	2	1	1
RF	4	high	28	10	10	6	5	5	7
		low	50	9	2	1	3	1	3
	5	high	34	12	10	1	5	5	6
		low	44	7	1	2	2	0	2
	6	high	43	14	6	1	5	3	6
		low	35	5	2	1	2	1	2
LR	4	high	28	10	7	5	6	6	6
		low	50	9	1	2	3	4	4
	5	high	34	12	8	3	6	7	8
		low	44	7	0	4	2	3	1
	6	high	43	14	6	1	3	3	7
		low	35	5	0	3	2	2	1

The binary classifiers are developed based on the 4-year, 5-year, and 6-year metastasis-free times to define the high and low risk classes.

ISSN: 1471-2288