Optimized machine learning modelling for predicting the construction cost and duration of tunnelling projects

Predicting duration and cost of tunnelling projects is an essential factor in determining the usefulness of a decision-making system. Therefore, research on the duration and cost of tunnels' construction and the possibility of their distribution is essential, which can be a valid basis for concluding a contract and a tool to optimize the construction plan and reduce costs. For this purpose, we employ four machine learning techniques of linear regression (LR), Gaussian process regression (GPR), support vector regression (SVR), and decision tree (DT). The models were optimized using the grey wolf optimization algorithm. 350 datasets, including 16 input parameters, were used in the models. The four models' prediction performance for predicting the duration and cost parameters from high to low was LR, GPR, SVR, and DT. The sensitivity analysis revealed that the most influential parameters on the duration and cost of tunnels are the drilling machinery system and groundwater, respectively.

Automated summary

Predicting the duration and cost of tunnelling projects is crucial for determining the effectiveness of a decision-making system. Research on these factors can serve as a valid basis for contract negotiation and construction optimization. Four machine learning techniques were employed, with linear regression performing the best in predicting duration and cost.