Table 3.

Screening results of five methods for identifying predictors.

Characteristics	Univariate Cox regression		LASSO regression		Besubset regression		Random forest		XGboost
	Preliminary	Final	Preliminary	Final	Preliminary	Final	Preliminary	Final	Preliminary	Final
Age	+	+	+	+	+	+	+	+	+	+
Gender	+		+				+		+
Race	+						+		+
Grade	+	+	+	+	+	+	+	+	+	+
Stage	+				+		+	+
T stage	+	+	+	+	+	+	+	+	+	+
N stage	+	+	+	+	+	+	+		+	+
Radiation	+	+	+	+	+	+			+	+
Chemotherapy	+	+	+	+	+	+	+	+	+	+
Laterality			+						+
Marriage			+	+	+	+	+	+	+	+
Lymph node ratio(%)	+		+				+
Total number	10	6	10	7	8	7	10	6	10	7
AUC(5 years)		87.7		88.4		88.4		88.0		88.4
AIC	2131.3	2122.3	2123.9	2118.9	2122.6	2118.9	2129.0	2120.0	2124.2	2118.9

Final, the results obtained after the model undergone Multivariate Cox stepwise backward regression.

Statistical significance indicated by P < 0.05 has been denoted in bold.