Prediction of postoperative complications of pediatric cataract patients using data mining

Table 2 Performance indicators in three binary classification problems

Method		Accuracy	FNR	FPR
Problem 1: Whether a patient suffers from complications
Random forest	–	0.757 ± 0.025	0.414 ± 0.031	0.128 ± 0.013
Random forest	SMOTE	0.762 ± 0.019	0.231 ± 0.013	0.220 ± 0.037
Naïve Bayesian	–	0.748 ± 0.025	0.465 ± 0.042	0.887 ± 0.023
Naïve Bayesian	SMOTE	0.751 ± 0.032	0.270 ± 0.043	0.208 ± 0.044
Problem 2: Whether a patient suffers from SLPVA
Random forest	–	0.810 ± 0.014	0.621 ± 0.089	0.071 ± 0.023
Random forest	SMOTE	0.753 ± 0.069	0.257 ± 0.054	0.258 ± 0.044
Naïve Bayesian	–	0.782 ± 0.014	0.155 ± 0.043	0.449 ± 0.100
Naïve Bayesian	SMOTE	0.782 ± 0.043	0.244 ± 0.065	0.267 ± 0.025
Problem 3: Whether a patient suffers from AHIP
Random forest	–	0.838 ± 0.024	0.580 ± 0.050	0.015 ± 0.014
Random forest	SMOTE	0.813 ± 0.016	0.228 ± 0.055	0.265 ± 0.025
Naïve Bayesian	–	0.847 ± 0.033	0 ± 0	0.321 ± 0.043
Naïve Bayesian	SMOTE	0.816 ± 0.037	0.225 ± 0.047	0.265 ± 0.074

ISSN: 1479-5876