| In Malaysia, construction cost estimation still depends on foreign-derived labour constants, which fail to reflect the local construction practices, leading to discrepancies in tender pricing, budget overruns, and inconsistent formulation of the build-up rate. This paper addresses the limitations by developing a data-driven framework for localised labour constant modelling using both practitioner knowledge and machine learning models. To identify the key issues in current labour constant practices and critical productivity factors in the Malaysian construction projects, a quantitative survey of 110 Quantity Surveyors (QS) was conducted. The results indicate that the biggest problems are budget overruns, inaccurate cost estimation, and lack of context-specific productivity data. Key variables identified include the labour composition, project scale, site conditions, and plant utilisation. Based on these findings, this research paper proposes the Build-Up Rates Optimum System (S-BRO) framework, a multi-layered system architecture designed to support data-driven construction cost estimation. A pilot machine learning modelling study was carried out based on a dataset of construction build-up rate records and extended by controlled scenario-based data expansion. Various algorithms, including multiple linear regression, Random Forest, Gradient Boosting, and Extra Trees, were evaluated on standard performance measures, including RMSE, MAE, and R². The findings show that ensemble-based models perform better than the traditional linear methods, which implies that there are non-linear relationships among the variables of construction cost. The findings demonstrate the potential of machine learning to improved build-up rate estimation in Malaysian construction. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.