Coping with Complexity When Predicting Surface Roughness in Milling Processes: Hybrid Incremental Model with Optimal Parametrization

Table 4

Main parameters of each modeling strategy considered in this study.

Algorithms	BN	ANN	HIM + SA
Model	Seven inputs; (Fr, fz, D, ae, HB, Geom, rpm); Output: Ra	Seven inputs: (Fr, fz, D, ae, HB, Geom, rpm); Output: Ra	Seven inputs; (Fr, fz, D, ae, HB, Geom, rpm); Output: Ra
Clustering	No (TAN)	No (BP)	Fuzzy k-NN
Structure	Three layers (4-11-4 neurons)	Tree-like structure	Second-order polynomial, two neighbors, and 1.92 for fuzzy strength parameter
Inference	Global	Causal and abductive	Global and local
Main parameters to adjust	No tuning parameters	Many (MLP)	m, k, p, optimal setting by SA
Training algorithms	No (estimation of conditional probability)	BP	LSE + IBL
Interpretability	Yes	No	Partially

BN: Bayesian network; ANN: artificial neural network; HIM + SA: hybrid incremental model with simulated annealing; TAN: tree augmented Naive-Bayes; BP: backpropagation; MLP: multilayer perceptron; LSE: least square error; IBL: instance based learning.