Identification of a minimum number of genes to predict triple-negative breast cancer subgroups from gene expression profiles

Table 2 Comparative overview of seven prediction algorithms according to the 81 downregulated genes

	TP rate	FP rate	Accuracy %	Mean absolute error	Kappa	Precision	Recall	F-measure	MCC	ROC area	PRC area
Naive bayes	0.846	0.037	84.557	0.0515	0.8055	0.848	0.846	0.846	0.809	0.980	0.926
Logistic regression	0.668	0.082	66.8354	0.1091	0.5828	0.672	0.668	0.669	0.587	0.919	0.758
Multilayer perceptron	0.886	0.029	88.6076	0.047	0.8563	0.885	0.886	0.886	0.858	0.988	0.953
Support vector machine	0.861	0.036	86.0759	0.2264	0.8241	0.861	0.861	0.860	0.826	0.958	0.808
k-Nearest neighbours	0.744	0.066	74.4304	0.0884	0.6777	0.744	0.744	0.740	0.677	0.836	0.615
Decision tree	0.618	0.098	61.7722	0.1542	0.5131	0.612	0.618	0.610	0.521	0.847	0.582
Random forest	0.825	0.053	82.5316	0.1386	0.7755	0.827	0.825	0.813	0.781	0.979	0.919

TP, true positive; FP, false positive; MCC, Matthews correlation coefficient; ROC, relative operating characteristic; PRC, precision–recall curve

ISSN: 1479-7364