Research on Risk Assessment and Decision Optimization Model for Construction Projects Based on Transformer-Improved Reinforcement Learning

Abstract As construction projects become more complex and unpredictable, traditional risk assessment and decision-making methods often fall short in meeting the demands of modern engineering management. Our study introduces a novel model for construction project risk assessment and decision optimization, which combines the Transformer architecture with reinforcement learning. The Transformer identifies risk factors, and reinforcement learning dynamically optimizes risk assessment and decision-making. The process treats risk assessment as a sequential decision task, leveraging the Transformer to analyze historical data for identifying risk patterns and employing reinforcement learning to adjust risk responses in accordance with real-world project conditions, aiming for minimized risks and maximized benefits. Tests show that this model enhances risk assessment accuracy by 15% and decision-making efficiency by 20% compared with traditional methods, leading to improved outcomes in various risk scenarios and a reduction in potential losses. In summary, this model offers an innovative and practical solution for construction project risk management, with significant application potential.