Table 6 Model configurations and LOOCV performance.

From: Predictive modeling and optimization of surface roughness in Reverse-µEDM fabricated microeletrode arrays using ML models

Model

Key hyperparameters

R2

MAE (µm)

RMSE (µm)

Training time (s)

ANN

Hidden layer sizes=(5), α = 0.0001

0.971

0.221

0.31

13.54

SVR

Kernel = RBF, C = 1.0, ε = 0.1, γ=‘scale’

0.743

0.682

0.927

1.36

RF

n_estimators = 50, max depth = None, min samples leaf = 1

0.975

0.178

0.287

65.62

GBT

n_estimators = 50, learning rate = 0.01, max depth = 1

0.482

1.086

1.317

12.92

GPR

RBF kernel (auto length-scale), α = 1e-10

0.787

0.597

0.844

0.004

Back to article page