Prediction of the risk of developing hepatocellular carcinoma in health screening examinees: a Korean cohort study

Table 3 Performances of machine learning models in prediction of the probability of and the time to the development of HCC

Model	Evaluation metric	CV or OOB error in the training cohort	Validation in the test cohort
Model	Evaluation metric	Value (±SD)	AUC (95% CI)
Probability of developing HCC within 10 years
Random survival forest	AUC	0.871 (±0.019)
Random survival forest	BSS	0.062
Extreme gradient boosting	AUC	0.882 (±0.013)
Extreme gradient boosting	BSS	0.109
Ensemble of two models	AUC	0.892 (±0.011)	0.873 (0.860–0.885)
Ensemble of two models	BSS	0.112	0.078
Time to cancer occurrence if HCC develops
Cox proportional hazard	C-index	0.843 (±0.006)	0.828 (0.819–0.838)
Random survival forest	C-index	0.881 (±0.010)	0.857 (0.850–0.864)

HCC hepatocellular carcinoma, CV cross validation, OOB out-of-bag, SD standard deviation, AUC area under receiver operating characteristics curve, CI confidence interval, BSS Brier skill score, C-index concordance index

ISSN: 1471-2407