Labour Productivity Estimation Model for Reinforced Concrete Structural Works Using Artificial Neural Network

Abstract

Labour productivity is a critical factor in determining the success of construction projects because it influences cost efficiency, schedule performance, and project quality. However, accurately estimating labour productivity remains challenging due to the complex interactions among workforce characteristics, management practices, resource availability, environmental conditions, and project-specific factors. Conventional estimation methods often fail to capture nonlinear relationships among these variables, resulting in limited prediction accuracy. This study aimed to develop an Artificial Neural Network (ANN)-based model for estimating labour productivity in reinforced concrete structural works and to identify the contribution of productivity factors influencing model predictions. A quantitative research approach was employed using data collected from 16 multistorey building construction projects in Bali, Indonesia. Productivity factors were identified through a literature review and selected using the Relative Importance Index (RII), while labour productivity was represented by the Productivity Index (PI). A Feedforward Neural Network trained using Bayesian Regularization (FNN-BR) was developed using 12 selected productivity factors and evaluated using the coefficient of determination (R²), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results showed that the optimal ANN architecture, consisting of 15 input neurons, 8 hidden neurons, and 1 output neuron, achieved strong prediction performance with an internal R² value of 0.8861, an RMSE of 0.0671, and a MAPE of 5.58%. External validation produced an R² value of 0.9178, an RMSE of 0.0524, and a MAPE of 4.06%, indicating good generalization capability. The contribution analysis identified site supervision quality as the most influential factor, followed by site accessibility, labour attendance, and workforce health conditions. This study concluded that the proposed ANN model provided an effective approach for estimating labour productivity and supporting data-driven decision-making in reinforced concrete construction projects.