The Construction and Application of a Rapid Estimation Model for Engineering Cost based on BP Algorithm

Traditional cost estimation methods in construction projects struggle with efficiency and accuracy, especially in early stages with limited data. This study develops a rapid estimation model using a Backpropagation (BP) neural network. Key factors (building area, project type, design complexity, material costs) were quantified, and data preprocessing (cleaning, standardization, feature selection) was applied. The BP model was optimized via hyperparameter tuning and cross-validation. Comparative experiments with Support Vector Regression (SVR) and Decision Tree Regression evaluated performance. The BP model achieved superior accuracy with MSE = 350,000, MAE = 400, and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.92, outperforming SVR (MSE = 420,000, R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.85) and Decision Tree Regression (MSE =490,000, R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.82). However, BP’ s computational efficiency (training: 1.25s, prediction: 0.42s) lagged behind SVR (0.38s/0.09s). The BP model offers high accuracy for early-stage cost prediction but requires efficiency improvements. Future work should optimize network structures or integrate parallel computing. This data-driven approach enhances decision-making in construction management, balancing precision and practicality.