Skip to main content
. 2021 Mar 24;11:6770. doi: 10.1038/s41598-021-86361-5

Table 2.

Evaluation for the best model in untrained data (test data) for each group. Group 1 is made up of patients without text data, and Group 2 patients with text data.

Support (%) Precision Recall F1-score
Group 1
RF accuracy score (76.36%)
 Discarded cases 35 (63.6%) 0.76 0.91 0.83
 Somewhat probable 20 (36.4%) 0.77 0.50 0.61
Group 2
GB accuracy score (83.33%)
 Discarded cases 135 (60.9%) 0.91 0.91 0.91
 Somewhat probable 40 (18.0%) 0.70 0.53 0.60
 Moderately probable 36 (16.2%) 0.76 0.86 0.81
 Highly probable 11 (4.9%) 0.56 0.82 0.67