Fig. 10: The model comparison results. | npj Digital Medicine

Fig. 10: The model comparison results.

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

Fig. 10: The model comparison results.

The BOUND model was compared to the traditional logistic regression model across datasets with varying missing data rates. Since the traditional logistic regression model cannot be directly applied to missing data, we employed multiple imputation (MI) and random forest (RF) imputation to handle the missing values, allowing for its application in missing datasets. a Calibration curve intercept, the closer to 0 the better; (b) Calibrate the slope of the curve, the closer to 1 the better. c The performance of AUC. As the missing data rate increases, the advantage of the BOUND model becomes more pronounced.

Back to article page