Fig. 10: Comparisons of ten candidate models with four metrics.
From: Mechanistic data-driven prediction of as-built mechanical properties in metal additive manufacturing
The error bars indicate the standard deviations of the five model results based on 5-fold cross-validation. Name of a model includes two parts separated by an underline. The first part indicates the method used for reducing the dimension of the input (thermal history). The method includes Discrete Binning (DB) (“Dimension reduction by Discrete Binning (DB)” section), PCA, and WT (“Wavelet transform” section). The second part indicates the regression method including Regression Tree (RT)33, Feed-forward Neural Network (FNN)57, Lasso58, Bayesian Ridge Regression (BRR)59, Gradient Boosting Regression (GBR)34, K-Nearest Neighbors (KNN)60, Random Forest (RF)33,61, and CNN62. a Coefficient of determination (R2), b mean squared error (MSE), c mean relative error (MRE), d mean absolute error (MAE).
