Table 2 Overall accuracy in terms of the ROC AUC score and individual (twinned and not twinned) accuracy for the five best performing ML methods for the AZ31 dataset.

From: Application of machine learning to assess the influence of microstructure on twin nucleation in Mg alloys

Method

ROC AUC

Twinned accuracy

Not twinned accuracy

Bayesian networks

0.871

0.879

0.863

Gaussian naïve Bayes

0.851

0.834

0.868

GDB

0.825

0.683

0.966

Random forests

0.812

0.660

0.954

AdaBoost

0.807

0.645

0.972

  1. The scores presented here are the mean over the 10 cross-validation tasks. The “individual” scores were obtained by calculating the ratio between correct predictions and total samples (i.e., a value of 0 would mean that all predictions were incorrect and a value of 1 that all predictions were correct). The default settings were used as implemented in the pyAgrum84 and scikit-learn83 Python packages for the Bayesian network and for all other methods, respectively. The highest accuracies are highlighted with bold font.
Back to article page