Table 5 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 Materials Class Data’.

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

Property

Data Size

Base

MAE of SC Model

MAE of Proposed TL Model

% Error Change

b3lyp gap (eV)

4854

0.00986

0.00062

0.00055

-11.29

b3lyp homo (eV)

4854

0.00632

0.00047

0.00040

-14.89

b3lyp lumo (eV)

4854

0.00968

0.00044

0.00038

-13.64

b3lyp scharber jsc (mAcm−2)

4854

25.05731

1.18286

1.15373

-2.46

b3lyp scharber pce (%)

4854

0.84101

0.03518

0.03443

-2.13

b3lyp scharber voc (V)

4854

0.17182

0.00933

0.00928

-0.54

bp86 gap (eV)

4854

0.00825

0.00056

0.00042

-25.00

bp86 homo (eV)

4854

0.00599

0.00045

0.00034

-24.44

bp86 lumo (eV)

4854

0.00907

0.00048

0.00037

-22.92

bp86 scharber jsc (mAcm−2)

4854

54.61424

2.62765

2.61410

-0.52

bp86 scharber pce (%)

4854

1.99136

0.11712

0.11567

-1.24

bp86 scharber voc (V)

4854

0.14548

0.00846

0.00824

-2.60

m06 gap (eV)

4854

0.01170

0.00068

0.00065

-4.41

m06 homo (eV)

4854

0.00681

0.00048

0.00042

-12.50

m06 lumo (eV)

4854

0.01035

0.00054

0.00045

-16.67

m06 scharber jsc (mAcm−2)

4854

0.74746

0.03624

0.03867

6.71

m06 scharber pce (%)

4854

0.06892

0.00379

0.00363

-4.22

m06 scharber voc (V)

4854

0.18530

0.01045

0.01026

-1.82

pbe0 gap (eV)

4854

0.01026

0.00064

0.00058

-9.38

pbe0 homo (eV)

4854

0.00640

0.00044

0.00036

-18.18

pbe0 lumo (eV)

4854

0.00982

0.00052

0.00045

-13.46

pbe0 scharber jsc (mAcm−2)

4854

18.46749

0.88544

0.86032

-2.84

pbe0 scharber pce (%)

4854

0.82323

0.03391

0.03301

-2.65

pbe0 scharber voc (V)

4854

0.17392

0.00890

0.00904

1.57

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