Skip to main content

Table 3 Simulation study results. Mean and coefficient of variation (in parenthesis) of precision (when recall = 0.70), PPV (when NPV = 0.95), AUC, NO. (number of features in predictive models) of five prediction models in testing set in 100 repetitions

From: The application of unsupervised deep learning in predictive models using electronic health records

Scenarios

Prediction Models

Precision (%)

(recall = 0.70)

PPV (%) (NPV = 0.95)

AUC

NO.

1. Raw Data

Autoencoder

24.23 (0.18)

19.93 (0.07)

0.749 (0.01)

50 (0.00)

LASSO (λmin)

28.25 (0.17)

25.09 (0.05)

0.788 (0.01)

300 (0.06)

Random Forest

25.63 (0.18)

21.93 (0.06)

0.767 (0.01)

100 (0.00)

Simple Reg

20.96 (0.20)

15.73 (0.11)

0.708 (0.02)

12 (0.00)

Enhanced Reg

24.62 (0.18)

20.45 (0.07)

0.754 (0.01)

57 (0.03)

2. Correct Categories

Autoencoder

25.07 (0.18)

21.45 (0.07)

0.757 (0.03)

50 (0.00)

LASSO (λmin)

26.25 (0.17)

22.94 (0.05)

0.771 (0.01)

132 (0.02)

Random Forest

24.93 (0.18)

21.57 (0.06)

0.759 (0.01)

136 (0.00)

Simple Reg

21.36 (0.18)

17.10 (0.09)

0.713 (0.01)

16 (0.00)

Enhanced Reg

25.77 (0.17)

22.32 (0.06)

0.766 (0.01)

60 (0.06)

3. Incorrect Categories

Autoencoder

22.73 (0.18)

18.82 (0.08)

0.732 (0.01)

60 (0.00)

LASSO (λmin)

24.07 (0.17)

20.25 (0.06)

0.748 (0.01)

132 (0.02)

Random Forest

22.70 (0.18)

18.67 (0.07)

0.733 (0.01)

136 (0.00)

Simple Reg

19.83 (0.19)

15.31 (0.12)

0.690 (0.02)

16 (0.00)

Enhanced Reg

23.61 (0.18)

19.69 (0.07)

0.743 (0.01)

69 (0.03)

4. Incorrect Categories and Missing Data

Autoencoder

24.16 (0.18)

20.45 (0.07)

0.748 (0.03)

60 (0.00)

LASSO (λmin)

25.32 (0.17)

21.67 (0.06)

0.761 (0.01)

175 (0.08)

Random Forest

23.61 (0.18)

19.92 (0.07)

0.745 (0.01)

226 (0.00)

Simple Reg

20.92 (0.19)

16.31 (0.10)

0.706 (0.02)

28 (0.00)

Enhanced Reg

24.89 (0.17)

21.25 (0.07)

0.756 (0.02)

81 (0.04)