FIGURE 8.

The accuracy of RF, MLP, SVM, and E–S algorithms for predicting soybean yield using full and RFE selected variables (-VS) measured at R4 (A) and R5 (B) soybean growth stages in four environments. The mean performance was shown as × in each figure. MLP, multilayer perceptron; SVM, support vector machine; RF, random forest; E–S, ensemble–stacking strategy; RFE, recursive feature elimination.