Optimizing the Prediction Accuracy of Friction Capacity of Driven Piles in Cohesive Soil Using a Novel Self-Tuning Least Squares Support Vector Machine

Table 3

Comparative results of the predictive methods.

Methods	R value		RMSE value		MAE value		Hyperparameters
	Training	Testing	Training	Testing	Training	Testing
Neural network [6]	0.985	0.956	6.824	5.333	5.111	4.715	N/A
Levenberg–Marquardt BPNN (present study)	0.9848	0.9480	7.6906	7.2189	6.255	6.5531	Final gradient = 205.2903, final μ = 100
LS-SVM (present study)	0.9723	0.8802	13.9526	10.5474	9.351	7.7231	γ = 1, σ^2 = 1
ST-LSSVM (present study)	0.9903	0.9593	5.6678	5.3455	4.2381	4.6812	γ = 81369675.72, σ^2 = 8622.39