Skip to main content

Table 3 Discrimination results of predicting neonatal jaundice with CRF and GV under CN220 guideline

From: Ensemble learning for the early prediction of neonatal jaundice with genetic features

Variables Method AUC F1-score Precision
CRF Lightgbm 0.792 (0.757–0.828) 0.213 (0.171–0.251) 0.136 (0.109–0.161)
  Cart 0.553 (0.509–0.592) 0.150 (0.074–0.211) 0.137 (0.074–0.191)
  Logistic 0.785 (0.753–0.821) 0.210 (0.178–0.240) 0.122 (0.103–0.141)
  Naive Bayes 0.735 (0.673–0.782) 0.165 (0.129–0.188) 0.091 (0.069–0.104)
  rf 0.766 (0.711–0.806) 0.206 (0.177–0.245) 0.123 (0.106–0.147)
GV36 Lightgbm 0.603 (0.546–0.662) 0.149 (0.105–0.189) 0.105 (0.074–0.131)
  Cart 0.558 (0.522–0.598) 0.149 (0.105–0.191) 0.110 (0.079–0.139)
  Logistic 0.569 (0.519–0.614) 0.118 (0.093–0.141) 0.068 (0.053–0.081)
  Naive bays 0.562 (0.509–0.622) 0.112 (0.106–0.116) 0.059 (0.057–0.062)
  rf 0.587 (0.522–0.652) 0.148 (0.104–0.197) 0.103 (0.074–0.136)
CRF_GV36 Lightgbm 0.820 (0.785–0.857) 0.277 (0.218–0.333) 0.204 (0.160–0.247)
  Cart 0.569 (0.517–0.621) 0.184 (0.103–0.269) 0.175 (0.095–0.250)
  Logistic 0.781 (0.730–0.816) 0.218 (0.185–0.251) 0.129 (0.110–0.150)
  Naive Bayes 0.642 (0.563–0.707) 0.114 (0.105–0.124) 0.061 (0.056–0.067)
  rf 0.792 (0.753–0.833) 0.228 (0.193–0.259) 0.139 (0.118–0.158)
  1. The best performance by algorithms with CRF, GV36 and CRF_GV36 variables are marked in bold
  2. 95% CI is shown in parentheses