The study initiated with underlying principles of construction production which is an impetus to ill-conditioned prediction of project determinants at the early phases of building projects. To enhance the precision of these estimations, unique solutions relying on the statistical evidences were offered. Two alternative methods of analysis, namely linear regression and artificial neural networks, were employed to recognize the patterns in the sampled projects. Comparison was conducted on the basis of prediction measurements that were computed with the help of unseen test sample. The evidences of the empirical investigation suggest offered solutions provide superior prediction accuracy when compared to current practices. Last but not least, implementation of the solutions was illustrated on a random office development.

The study initiated with underlying principles of construction production which is an impetus to ill-conditioned prediction of project determinants at the early phases of building projects. To enhance the precision of these estimations, unique solutions relying on the statistical evidences were offered. Two alternative methods of analysis, namely linear regression and artificial neural networks, were employed to recognize the patterns in the sampled projects. Comparison was conducted on the basis of prediction measurements that were computed with the help of unseen test sample. The evidences of the empirical investigation suggest offered solutions provide superior prediction accuracy when compared to current practices. Last but not least, implementation of the solutions was illustrated on a random office development.

The study initiated with underlying principles of construction production which is an impetus to ill-conditioned prediction of project determinants at the early phases of building projects. To enhance the precision of these estimations, unique solutions relying on the statistical evidences were offered. Two alternative methods of analysis, namely linear regression and artificial neural networks, were employed to recognize the patterns in the sampled projects. Comparison was conducted on the basis of prediction measurements that were computed with the help of unseen test sample. The evidences of the empirical investigation suggest offered solutions provide superior prediction accuracy when compared to current practices. Last but not least, implementation of the solutions was illustrated on a random office development.