An Improved Stacking Model for Equipment Spare Parts Demand Forecasting Based on Scenario Analysis

Figure 9

Performance of Different Feature Selection Methods. (a) The Proposed Ensemble Feature Selection Method. (b) RF Feature Selection Method. (c) PCA Feature Selection Method. (d) RFECV Feature Selection Method. (R2_Score is the coefficient of determination; MAE is mean_absolute_error, the expected value of the absolute error loss; MSE is mean_squared_error, the expected value of the squared (quadratic) error or loss.).

(a)

(b)

(c)

(d)