Table 3.

Prediction performance of machine learning algorithms for the estimation of 3 month mortality among bone metastasis patients with lung cancer.

Measures	Approaches
	Logistic regression	XGBoosting machine	Random forest	Gradient boosting machine	Neural network	Decision tree
Mean predicted	0.488	0.487	0.486	0.487	0.461	0.487
Brier score	0.171	0.169	0.178	0.169	0.171	0.175
Intercept	0.05	0.06	0.05	0.06	0.23	0.05
Slope	0.96	0.95	1.49	0.97	0.88	0.88
AUC (95%CI)	0.815 (0.801–0.828)	0.820 (0.807–0.833)	0.811 (0.798–0.824)	0.820 (0.807–0.833)	0.818 (0.805–0.832)	0.806 (0.792–0.820)
Discrimination slope	0.327	0.338	0.228	0.334	0.349	0.349
Specificity	0.812	0.812	0.807	0.809	0.799	0.800
Sensitivity	0.731	0.731	0.733	0.734	0.742	0.735
NPV	0.754	0.754	0.755	0.756	0.759	0.754
PPV	0.793	0.793	0.789	0.791	0.785	0.784
Precision	0.793	0.793	0.789	0.791	0.785	0.784
Recall	0.731	0.731	0.733	0.734	0.742	0.735
Youden	1.543	1.543	1.541	1.543	1.542	1.536
Accuracy	0.772	0.772	0.771	0.772	0.771	0.768
Threshold	0.526	0.488	0.558	0.466	0.382	0.444

AUC, Are under the curve; CI, Confident interval; NPV, Negative predictive value; PPV, Positive predictive value; XGBooting, eXtreme Gradient Boosting.