Figure 2
From: Comparative performances of machine learning algorithms in radiomics and impacting factors
Heat map of median [Q1; Q3] AUC scores for all 9 × 14 combinations of feature selection algorithms and classifiers. All : No feature selection (non-informative); Random : Random feature selection (non-informative) ; Pearson : Pearson correlation coefficient; Spearman : Spearman correlation coefficient; RfVarImp : Random Forest Variable Importance; RfPermImp : Random Forest Permutation Importance; JMI : Joint Mutual Information; JMIM : Joint Mutual Information Maximization; MRMR : Minimum-Redundancy-Maximum-Relevance ; KNN:K-Nearest Neighbors; Lr : Linear Regression; LrL1 : Lasso Penalized Linear Regression; LrL2 : Ridge Penalized Linear Regression; LrElasticNet : Elastic-net Linear Regression; LDA : Linear Discriminant Analysis; RF: Random Forest; AdaBoost : AdaBoost; XGBoost : XGBoost; Linear SVC : Linear Support Vector Classifier; Poly SVC : Polynomial Support Vector Classifier; RBFSVC : Radial Support Vector Classifier; BNB : Binomial Naive Bayes; GNB : Gaussian Naive Bayes.
