Fig. 2
From: Explainable illicit drug abuse prediction using hematological differences
Performance of ML models to predict IDU. (A) ROC curves for all ML models. (B) Internal validation set performance of ML models with a limited number of training features. (C) The overall performance of the LGB model and its standard error in 50 random subsampling cross validation. (D) The performance of RF, LGB, AdaBoost, and GBM models trained with the first 13 features on the internal and external validation sets based on RRSCV, where the box length is the distance between quartiles Q1 and Q3, the dotted line is the mean, the whisker is the inner limit (1.5 times the interquartile range), and the circle is the outlier. IDU: illicit drug use. AUC: the area under the receiver-operating-characteristic curve. RF: random forest. GBM: gradient boosting machine. LGB: light gradient boosting machine. LR: logistic regression. XGB: eXtreme gradient boosting. AdaBoost: adaptive boosting. SVM: support vector machine. BPNN: Backpropagation neural network.
