Table 4.

Accuracy rates in predicting survival status as a result of evaluations of the SVM Kernel machine learning method with the MRMR, Chi2, ANOVA, and Kruskal–Wallis analysis methods.

Machine learning method	Analysis method	Feature count	Features	Accuracy
SVM kernel	MRMR	9	Number of cycles of chemotherapy Neutrophils Chemotherapy response Metastasis location CEA ECOG PS LDH Lymphocytes First-line chemotherapy	87.9%
SVM kernel	Chi2	10	Lymphocytes Tumor location Complaints ECOG PS LDH Number of lines of chemotherapy Number of cycles of chemotherapy Metastasis location Chemotherapy response First-line chemotherapy	87.9%
SVM kernel	ANOVA	3	Number of lines of chemotherapy Number of cycles of chemotherapy First-line chemotherapy	87.9%
SVM kernel	Kruskal–Wallis	8	Neutrophils Tumor location CEA Complaints Number of lines of chemotherapy LDH Number of cycles of chemotherapy First-line chemotherapy	87.3%

ANOVA = analysis of variance, CEA = carcinoembryonic antigen, Chi2 = chi-square, ECOG PS = Eastern Cooperative Oncology Group performance status, LDH = lactate dehydrogenase, MRMR = minimum redundancy–maximum relevance, SVM = support vector machine.