Table 8 Prediction accuracy of the DNN models.

From: Deep learning-based framework for time-dependent reliability analysis of a cable-stayed bridge with corroded PSC box girders under gravity loads

DNN Models

Dataset

MSE

MAE

Relative error (%)

R2

DNN 1

Training

0.034

0.073

9.29

0.997

Test

0.034

0.073

9.27

0.997

DNN 2

Training

0.030

0.084

9.09

0.995

Test

0.030

0.084

9.07

0.996

DNN 3

Training

0.019

0.049

5.67

0.997

Test

0.020

0.050

5.74

0.997

Back to article page