Table 2.

Performance metrics of the Machine Learning models and comparisons with conventional Logistic Regression for the prediction of Critical Care Unit admission following cytoreductive surgery for advanced high grade serous ovarian cancer.

Predictors	Model	Set	Accuracy	Sensitivity	Specificity	F-Score
All variables (n = 13)	KNN (K = 4)	Train	0.94	0.78	0.97	0.86
		CV LOO	0.94	0.78	0.97	0.86
		Test	0.80	0.45	0.92	0.60
	ANN	Train	0.97	0.96	0.97	0.96
		CV LOO	0.88	0.85	0.88	0.86
		Test	0.82	0.86	0.81	0.83
	LDA	Train	0.97	0.96	0.97	0.96
		Test	0.90	0.93	0.89	0.91
	QDA	Train	0.97	1.00	0.97	0.98
		Test	0.93	0.93	0.93	0.93
	LR	Train	0.96	0.85	0.98	0.91
		Test	0.84	0.59	0.93	0.72
Selected * Variables (p < 0.05)	KNN (K = 6)	Train	0.94	0.89	0.95	0.92
		CV LOO	0.94	0.89	0.95	0.92
		Test	0.86	0.69	0.92	0.79
	ANN	Train	0.90	0.89	0.90	0.99
		CV LOO	0.89	0.89	0.89	0.89
		Test	0.76	0.79	0.74	0.76
	LDA	Train	0.97	0.96	0.97	0.96
		Test	0.89	0.93	0.88	0.90
	QDA	Train	0.89	0.96	0.87	0.91
		Test	0.75	0.97	0.68	0.80
	LR	Train	0.95	0.78	0.98	0.87
		Test	0.82	0.55	0.92	0.69

* Surgical complexity score; pre-surgery albumin; blood loss; operative time; bowel resection with stoma. KNN; k-Nearest Neighbors, ANN; Artificial Neural Network, CV-LOO; Leave-one-out-cross-validation; LDA; Linear Discriminant Analysis, QDA; Quadratic Discriminant Analysis, LR; Logistic Regression.