Fig 1.
Statistics and projections of population aging in Taiwan.
Fig 2.
DNN-based biological age prediction architecture for different populations [9].
Fig 3.
Biological age prediction architecture based on traditional machine learning methods [9].
Table 1.
Comparison of related studies and this work.
Fig 4.
Distribution of health examination years.
Fig 5.
Age distribution of health examination participants.
Table 2.
Sample of a single record from the merged dataset.
Fig 6.
Features with high missing rates filtered out.
Fig 7.
MICE flow chart [13].
Fig 8.
MICE example [11].
Fig 9.
Imputation process.
Table 3.
Table of information for the filtered complete dataset.
Fig 10.
The tree growth diagram of leaf-wise method.
Fig 11.
Example of converting residual life to biological age.
Table 4.
Average life expectancy by city.
Fig 12.
Survival analysis data example [18].
Fig 13.
Illustration of SHAP.
Fig 14.
PCC combined with SHAP.
Table 5.
Hardware environment table.
Table 6.
Software environment table.
Table 7.
Complete dataset obtained by imputation.
Table 8.
Complete dataset obtained by filtering.
Table 9.
Model parameters for biological age prediction with MICE imputation.
Table 10.
Model parameters for biological age prediction with filtered dataset.
Fig 15.
Comparison of different residual life prediction models.
Fig 16.
Residual life prediction using the imputed dataset (Male).
Fig 17.
Residual life prediction using the imputed dataset (Female).
Table 11.
Residual life prediction metrics using the imputed dataset.
Fig 18.
Residual life prediction using the filtered dataset (Male).
Fig 19.
Residual life prediction using the filtered dataset (Female).
Table 12.
Residual life prediction metrics using the filtered dataset.
Fig 20.
K-M curve evaluation of biological age prediction model (Male).
Fig 21.
K-M curve evaluation of biological age prediction model (Female).
Table 13.
Summary of aging-related biomarkers.
Fig 22.
PCC combined with SHAP experiment results.