Skip to main content
. 2020 Apr 3;10:5866. doi: 10.1038/s41598-020-62821-2

Figure 9.

Figure 9

Classification performance as a result of hyperparameter tuning for accuracy, precision, recall, and F1 for Hecht, Troiano, Choi, and Hees. Each figure also shows the effect that optimizing for one metric has on the remaining metrics. For example, optimizing for recall results in a very low precision score for all four algorithms. Besides optimized scores, we also show the default classification scores obtained by using the default hyperparameter values for easy comparison; which can also be viewed in Fig. 4. Error bars indicate the 95% confidence interval.