Table 2 Model performance to predict metastasis in ML approaches using 5-fold cross validation in original data and Balance data with oversampling method.

Models	Dataset		Sensitivity%	Specificity%	Precision%	F1	AUC
LR	Original data	Train	.86	.86	.76	.81	.93
	Original data	Test	.86	.78	.77	.81	.88
	Balanced data	Train	.93	.83	.88	.87	.93
	Balanced data	Test	.90	.78	.84	.84	.86
NN	Original data	Train	.93	.93	.88	.88	.98
	Original data	Test	.76	.78	.75	.77	.86
	Balanced data	Train	.98	.87	.95	.95	.99
	Balanced data	Test	.91	.81	.86	.86	.87
RF	Original data	Train	.89	.95	.92	.90	.98
	Original data	Test	.80	.83	.80	.80	.87
	Balanced data	Train	.98	.95	.96	.96	.99
	Balanced data	Test	.91	.78	.85	.85	.87
NB	Original data	Train	.83	.85	.74	.79	.90
	Original data	Test	.89	.78	.77	.83	.88
	Balanced data	Train	.87	.84	.86	.86	.91
	Balanced data	Test	.86	.81	.83	.83	.86
DT	Original data	Train	.82	.97	.94	.88	.96
	Original data	Test	.58	.78	.69	.63	.75
	Balanced data	Train	.94	.96	.95	.95	.98
	Balanced data	Test	.80	.81	.80	.80	.83
SVM	Original data	Train	.94	.86	.77	.84	.94
	Original data	Test	.92	.76	.76	.85	.85
	Balanced data	Train	.98	.87	.93	.92	.93
	Balanced data	Test	.93	.80	.87	.86	.85

Abrivation: Naive Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), Neural Network (NN), Desicion Tree (RT) and Logistic Regression (LR).

Quick links

Search