Randomized trial demonstrates forecasting's impact on contingency estimation in a bridge project, indicating improved budgeting approaches.
Key Points
This study aims to explore the integration of construction input price forecasting and probabilistic simulation for contingency estimation in a road infrastructure project.
Empirical application based on a Peruvian bridge project
Forecasting using Bi-GRU and Random Walk models
Monte Carlo simulation with monthly series of construction inputs and exogenous variables
Random Walk model showed lower RMSE values for most inputs in seven of eight comparisons
Bi-GRU resulted in a 7.24% lower RMSE for diesel
Estimated P95 contingency at 3.92% under Bi-GRU and 3.96% under Random Walk, indicating similar contingencies.