Improving Residential Construction Cost Prediction with PSO-BP Neural Network Optimization

Abstract

The prediction of costs for residential construction projects has been a prominent research focus for scholars both domestically and internationally. To identify a more efficient and precise prediction method, this study introduces the particle swarm optimization (PSO) algorithm to enhance the Backpropagation (BP) neural network, resulting in the establishment of the PSO-BP neural network model. Seven influencing factors were identified as feature indicators through literature review and expert interviews. The study utilized 35 sample data points from publicly completed residential projects in Anhui Province over the past five years. The implementation was carried out using MATLAB software coding. The findings reveal that the PSO-BP neural network prediction has an average absolute relative error of 0.30019%, which is significantly lower than the 0.41029% error of the standard BP neural network. This indicates that the PSO-BP neural network offers superior accuracy and precision in predicting construction costs.