Construction cost prediction system based on Random Forest optimized by the Bird Swarm Algorithm.

Area of Science:

• Construction Management
• Predictive Modeling
• Artificial Intelligence in Engineering

Background:

• Construction cost prediction faces challenges including low accuracy, poor efficiency, and high uncertainty due to project complexity.
• Existing methods often struggle with the dynamic and multifaceted nature of construction cost factors.

Purpose of the Study:

• To develop an advanced prediction index system and a robust prediction model for construction costs.
• To enhance the accuracy and efficiency of construction cost forecasting by addressing limitations of traditional methods.

Main Methods:

• Identification of key factors influencing construction costs and development of a 14-index prediction system.
• Construction of a Random Forest (RF) prediction model optimized using the Bird Swarm Algorithm (BSA) for parameter tuning.
• Validation using engineering data from a Chinese construction company, comparing BSA with other optimization algorithms and RF with other forecasting methods.

Main Results:

• The proposed model achieved a maximum relative error of only 1.24%, meeting engineering practice requirements.
• The Bird Swarm Algorithm demonstrated faster convergence to optimal parameters compared to other metaheuristic algorithms.
• The RF-BSA model outperformed traditional and advanced forecasting methods in accuracy and efficiency.

Conclusions:

• The developed prediction model offers a significant improvement for construction cost forecasting.
• The findings provide a basis for optimizing cost management in construction projects through more reliable predictions.