Slide 10
Slide 10 text
Nested cross-validation with Bayesian hyperparameter optimization
10
Matthew’s Correlation Coefficient (MCC) on test data in nested cross-validation
Elastic Net
SVM
Random Forest
XGBoost
MLP
0,50
0,55
0,60
0,65
0,70
0,75
0,80
0,85
0,90
0,95
1,00
Feature-based models
𝑀𝐶𝐶 =
𝑡𝑛 ∙ 𝑡𝑝 − 𝑓𝑛 ∙ 𝑓𝑝
(𝑡𝑝 + 𝑓𝑝)(𝑡𝑝 + 𝑓𝑛)(𝑡𝑛 + 𝑓𝑝)(𝑡𝑛 + 𝑓𝑛)
» +1 best agreement between predicted and actual values
» 0 no agreement
» -1 perfect misclassification
» Measurement is unaffected by unbalanced class
ratios