Predicting Final Construction Costs of Hospitals Based on Initial Project Attributes: An Advanced Regression Approach

Abstract

Accurate estimation of construction costs at the early stages of hospital projects is critical for effective budgeting and planning in healthcare infrastructure. Given the complexity of hospital design and the sensitivity of healthcare systems to cost overruns, advanced modeling techniques are required to improve forecast accuracy. This study aims to predict the final construction cost of hospital projects based on initial project attributes using multiple regression approaches, including Linear Regression, Support Vector Regression (SVR), Random Forest Regression, and Artificial Neural Networks (ANN). A synthetic dataset of 100 hospital projects was generated, capturing variables such as built-up area, number of beds, seismic zone, contract type, prefabrication method, and sustainability certification. Each model was trained and evaluated using standard performance metrics including RMSE, MAPE, and R². Results revealed that Random Forest Regression outperformed all other models, achieving the lowest prediction error and highest coefficient of determination (R² = 0.65), while SVR and ANN underperformed due to overfitting and insufficient data. The findings underscore the effectiveness of ensemble learning techniques in capturing the non-linear, multi-dimensional nature of hospital construction costs. This study provides a practical, data-driven framework for improving cost forecasting during the pre-construction phase, supporting better decision-making and risk mitigation in healthcare infrastructure development.