Determining the construction investment costs of public investment projects at the feasibility study preparation and appraisal stage during the project preparation phase using an optimal machine learning model.

Abstract

The accuracy of determining the construction investment costs of public-funded projects at the feasibility study stage plays a crucial role in the decision-making process for project approval, implementation of subsequent steps, and budget management; especially in the context of the 2025 Construction Law (effective from July 1, 2026), which abolishes the regulation on the appraisal of construction design following the basic design stage by specialized construction agencies (including the appraisal of construction cost estimates). However, determining costs at the initial stage often faces many difficulties due to limited information and high uncertainty. This study proposes a hybrid machine learning model combining least-squares support vector regression (LSSVR) and the Cheetah optimization algorithm using cost data from similar projects and works already implemented to improve the accuracy in determining the construction investment costs of public-funded projects. In this study, the Cheetah optimizer was used to optimize the hyperparameters of the LSSVR model, thereby improving the model's generalization ability and stability. The research dataset includes 50 public investment projects collected in Ho Chi Minh City, reflecting the practical characteristics of construction projects in urban conditions. The effectiveness of the proposed model was evaluated and compared with common methods including support vector machines (SVM) and artificial neural networks (ANNs). The research results showed that the CO–LSSVR model outperformed the comparison models, achieving high prediction accuracy with an average of 49,38, average MAE of 38,15, average MAPE of 11,92%, and R² of 0,883. These results confirm the effectiveness of combining metaheuristic optimization algorithms with machine learning models in estimating construction investment costs during the feasibility study phase of a project. The proposed model provides investment decision-makers, government agencies, and investors with an effective tool to improve the reliability of cost estimates and support decision-making in public investment projects.