Skip to main content
. 2024 Apr 16;16:45. doi: 10.1186/s13321-024-00835-y

Table 2.

To ensure that the accuracies achieved for the ECHA dataset were not due to favorable train-test split, we also performed a 5-fold cross validation on the entire dataset. The classification accuracy and weighted F1 scores per fold are summarized here (higher is better)

Fold Accuracy (%) Weighted F1 score
0 77.445 0.8136
1 72.826 0.7955
2 73.250 0.7838
3 75.820 0.7982
4 76.366 0.8080
Average 75.141 0.7998