Table 3 The table shows the test MAE of the SC model, proposed TL model, and % error change for each of the target materials properties for prediction task of ‘Other DFT-based Databases’.

From: Structure-aware graph neural network based deep transfer learning framework for enhanced predictive analytics on diverse materials datasets

Dataset

Property

Data Size

Base

MAE of SC Model

MAE of Proposed TL Model

% Error Change

Flla

Deltae (eV)

3936

14.3965

4.0309

3.8936

-3.41

 

Deltae pa (eVatom−1)

3936

0.8478

0.0941

0.0274

-70.88

 

Ehull (eVatom−1)

3927

0.0707

0.0424

0.0238

-43.87

DC

Vol (Å3)

1054

70.989

34.517

33.094

-4.12

 

Bg (eV)

1054

1.1527

0.4318

0.3657

-15.31

 

N

1054

0.7358

0.2872

0.2770

-3.55

 

Poly Elec

1054

4.9939

2.9339

2.2681

-22.69

 

PolyTotal

1054

8.6057

5.2415

4.7807

-8.79

PT

Vol (Å3)

940

65.282

29.099

27.352

-6.00

 

Eij (cm−2)

936

0.4610

0.3297

0.3125

-5.22

  1. The lowest MAE values in each row are highlighted in bold.