Implementation of Process-Based and Data-Driven Models for Early Prediction of Construction Time

Table 2

Accuracy results for the hybrid model BTC-GRNN using three validation methods.

Method for validation and testing	MAPE (%)	R^2 (%)	Coefficient of correlation between actual and predicted target values
10-fold cross-validation	3.34	93.17	0.966
Random percent (16%) (when using the optimization option by reducing the number of neurons)	3.76	93.72	0.970
LOO (when using the optimization option by reducing the number of neurons)	2.596	95.70	0.979