Table 2 Comparative of AUC between BN-logistic and traditional Bayesian Network parameter learning (i.e., Maximum Likelihood Estimation and Bayesian estimation)

From: Interpretable machine learning model for digital lung cancer prescreening in Chinese populations with missing data

 

BN-logistic (95%CI)

MLE (95%CI)

P-value

Bayes estimation (95%CI)

P-value

Internal validation set

0.866 (0.857–0.875)

0.853 (0.843–0.863)

< 0.01*

0.854 (0.844-0.864)

< 0.01*

Time-external validation set

0.848 (0.844-0.852)

0.824 (0.819-0.828)

< 0.01*

0.825 (0.820-0.829)

< 0.01*

Geographic-external validation set

0.841 (0.827-0.855)

0.813 (0.796-0.830)

< 0.01*

0.815 (0.799-0.832)

< 0.01*

  1. Abbreviations: MLE: Maximum Likelihood Estimation; CI: Confidence Interval.
  2. *P-values reached a significance level of 0.05.
  3. Notes: The P-value in the third column represent the comparison between MLE and BN-logistic, while the P-value in the fifth column represent the comparison between Bayes estimation and BN-logistic.
Back to article page