Low-Code Application and Practical Implications of Common Machine Learning Models for Predicting Punching Shear Strength of Concrete Reinforced Slabs

Table 4

Performance metrics comparison of machine learning models with different tuning methods.

Model	Set	MAE	RMSE	R^2
GPR model with Bayesian optimisation	Training set	19.61	27.59	0.995
	Validation	47.34	81.19	0.954
	Testing set	40.04	62.36	0.978
SVM model with Bayesian optimisation	Training set	46.46	80.24	0.955
	Validation	58.7	97.91	0.933
	Testing set	44.74	68.57	0.973
Ensemble with Bayesian optimisation	Training set	15.04	20.09	0.997
	Validation	54.94	115.57	0.907
	Testing set	43.87	77.01	0.966
GPR model with random search tuning	Training set	19.05	27.09	0.995
	Validation	43.58	74.01	0.962
	Testing set	42.47	78.908	0.964
SVM model with random search tuning	Training set	46.47	80.3	0.955
	Validation	58.92	102.57	0.926
	Testing set	44.58	67.97	0.974
Ensemble model with random search tuning	Training set	3.29	9.58	0.999
	Validation	56.78	129.19	0.883
	Testing set	47.98	89.29	0.954
SVM model with quadratic kernel	Training set	47.8	80.8	0.95
	Validation	58.9	102.3	0.93
	Testing set	42.8	62.9	0.98
GPR with rational quadratic kernel	Training set	22.8	31.9	0.99
	Validation	48.7	81.6	0.95
	Testing set	44.2	80.5	0.96
Ensemble model with default hyperparameters	Training set	44.6	88.9	0.95
	Validation	64.6	130.2	0.88
	Testing set	34.9	62.9	0.96