A systematic intelligent prediction model for residential construction cost based on fuzzy AHP and GA-BP neural network

Abstract

Accurate and early-stage cost prediction remains a significant challenge in the construction industry due to the nonlinear, high-dimensional, and interdependent nature of influencing factors. Traditional models often rely on oversimplified assumptions or limited feature representation, leading to suboptimal accuracy and generalization.To address these gaps, this study proposes an integrated intelligent prediction framework that combines the Fuzzy Analytic Hierarchy Process (FAHP) for feature selection with a Genetic Algorithm-optimized Backpropagation Neural Network (GA-BPNN). The FAHP systematically quantifies expert judgment to identify the most relevant construction features, while the GA improves the convergence and robustness of the neural network by globally optimizing its parameters.The model was validated using a real-world dataset of 4,552 residential construction projects obtained from the Glodon platform. Results show that the FAHP-GA-BPNN framework significantly outperforms benchmark models, achieving an RMSE of 79.5, MAE of 65.2, and R² of 0.93 on the validation set. This study not only contributes a scalable and adaptable methodology for intelligent cost estimation but also offers practical insights for enhancing decision-making in residential project planning and budgeting.