Table 3 Overall, balanced and class-wise evaluation of time-split experiments with ChEMBL data.

From: Studying and mitigating the effects of data drifts on ML model performance at the example of chemical toxicity data

 

CV

Predict holdout set

Cal_original

Cal_update1

Cal_update2

Cal_update1_and_2

Validity

0.81 ± 0.01

0.57 ± 0.14

0.75 ± 0.07

0.77 ± 0.09

0.78 ± 0.07

Efficiency

0.93 ± 0.04

0.82 ± 0.14

0.78 ± 0.12

0.74 ± 0.13

0.73 ± 0.15

Accuracy

0.87 ± 0.04

0.68 ± 0.10

0.68 ± 0.08

0.70 ± 0.10

0.70 ± 0.09

Balanced validity

0.81 ± 0.01

0.56 ± 0.11

0.73 ± 0.09

0.76 ± 0.08

0.77 ± 0.08

Balanced efficiency

0.93 ± 0.04

0.83 ± 0.14

0.79 ± 0.12

0.74 ± 0.13

0.73 ± 0.15

Balanced accuracy

0.87 ± 0.04

0.65 ± 0.09

0.65 ± 0.09

0.66 ± 0.10

0.67 ± 0.09

Validity inactive class

0.81 ± 0.01

0.62 ± 0.26

0.76 ± 0.22

0.78 ± 0.22

0.78 ± 0.20

Efficiency inactive class

0.93 ± 0.04

0.84 ± 0.14

0.79 ± 0.14

0.72 ± 0.14

0.73 ± 0.16

Accuracy inactive class

0.87 ± 0.05

0.72 ± 0.26

0.69 ± 0.26

0.68 ± 0.29

0.70 ± 0.24

Validity active class

0.81 ± 0.01

0.50 ± 0.22

0.71 ± 0.19

0.74 ± 0.18

0.75 ± 0.14

Efficiency active class

0.93 ± 0.05

0.81 ± 0.14

0.78 ± 0.13

0.75 ± 0.10

0.73 ± 0.16

Accuracy active class

0.87 ± 0.04

0.59 ± 0.20

0.61 ± 0.26

0.64 ± 0.23

0.64 ± 0.20

Back to article page