Fig. 4: ML performance of respective temperature sub-datasets. | npj Computational Materials

Fig. 4: ML performance of respective temperature sub-datasets.

From: Coupling physics in machine learning to predict properties of high-temperatures alloys

Fig. 4

Accuracy of trained ML models (random forest as an example) as a function of temperature and numbers of top-ranking features from the a MIC and b PCC analyses. The models were trained for ten times to determine the average accuracy and its standard deviation (error bar). The hyperparameters of each model were tuned up to 150 iterations to obtain optimum parameters. The vertical dash line indicates where issues of lack of data and relevant features start to occur.

Back to article page