Table 1 Summary of model performances for hMOF dataset and CO2 adsorption.

From: Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks

Descriptor

0.01 bar

0.05 bar

0.1 bar

0.5 bar

2.5 bar

Structural

0.45

0.55

0.61

0.67

0.71

Topological

0.57

0.64

0.68

0.75

0.80

T + S

0.70

0.70

0.72

0.78

0.84

T + WE

0.71

0.84

0.85

0.90

0.93

T + S + WE

0.70

0.86

0.88

0.92

0.94

Best model, Fanourgakis et al.11

–

0.65

–

0.90

0.93

  1. Machine learning results for carbon dioxide adsorption predictions on the hMOF dataset at different pressures, represented by R2 score. The best performing model for a given pressure is highlighted.
Back to article page