Fig. 2From: A cost-effective machine learning-based method for preeclampsia risk assessment and driver genes discoveryEvaluation and analysis of machine learning classifiers based on different feature selection strategies. AÂ IFS results of five feature selection strategies in four machine learning algorithms. BÂ UMAP shows the clustering of nine placental cell subpopulations in all gene sets (right) and TURF optimal gene sets (left). CÂ The heatmap shows the correlation of subpopulations of placental cells. DÂ Based on the TURF optimal gene set, the Confusion matrix of XGBoost on the independent dataset. EÂ The bar graph shows the mean absolute value of the SHAP values of the first 20 genes for the TURF_XGBBack to article page