Neural Network-Based Model for Predicting Preliminary Construction Cost as Part of Cost Predicting System

Table 1

Results for the training and validation data (DTREG software).

Estimators of the model accuracy (DTREG)	Value
Training data
Mean target value for input data	13.358369
Mean target value for predicted values	13.356284
Variance in input data	4.4677631
Residual variance after model fit	0.0024144
Proportion of variance explained by model R^2	0.99946 (99.946%)
Coefficient of variation (CV)	0.003678
Normalized mean square error (NMSE)	0.000540
Correlation between actual and predicted R	0.999731
Maximum error	0.3219897
RMSE (root mean squared error)	0.0491365
MSE (mean squared error)	0.0024144
MAE (mean absolute error)	0.0288461
MAPE (mean absolute percentage error)	0.2199448
Validation data
Mean target value for input data	13.358369
Mean target value for predicted values	13.35876
Variance in input data	4.4677631
Residual variance after model fit	0.0199458
Proportion of variance explained by model R^2	0.99554 (99.554%)
Coefficient of variation (CV)	0.010572
Normalized mean square error (NMSE)	0.004464
Correlation between actual and predicted R	0.997882
Maximum error	0,5402981
RMSE (root mean square error)	0.1412296
MSE (mean square error)	0.0199458
MAE (mean absolute error)	0.0984472
MAPE (mean absolute percentage error)	0.7326534