You are viewing the site in preview mode

Skip to main content

Table 2 Evaluation of the corresponding standard deviations of the five different predictions done by the 5 GNNs and information on the suggested quality and reliability of the predictions

From: Prediction of the water solubility by a graph convolutional-based neural network on a highly curated dataset

Range of SD*

n

rmse

Max. error

Suggestion

0–0.1

38

0.436

1.50

High quality/reliability

0.1–0.2

309

0.559

2.79

Good quality/reliability

0.2–0.3

317

0.593

3.23

Good quality/reliability

0.3–0.4

171

0.748

2.58

Moderate quality/reliability

0.4–0.5

91

0.867

2.72

Moderate quality/reliability

> 0.5

54

0.929

3.32

Low quality/reliability

  1. *Note that the range is given as 0 < SD <  = 0.1 (e.g.)