Research on Architectural Engineering Cost Prediction and Analysis Based on Generative Adversarial Network and Reinforcement Learning

In the context of global economic integration and technological innovation, the construction industry is facing unprecedented opportunities and challenges. Among them, accurately predicting the cost of architectural engineering is the key to ensuring project economic benefits and promoting industry transformation and upgrading. Insufficient data and nonlinear factors limit traditional cost-footing methods, making it hard to meet the increasing demand for modern construction projects. In view of this situation, this study proposes a solution that combines Generative Adversarial Networks (GANs) and Reinforcement Learning (RL), aiming at improving the accuracy and efficiency of cost forecasting. First, GANs are used to generate virtual engineering datasets covering a wide range of variables, which effectively expands the original data scale and enhances the model’s generalization ability. Then, with the help of RL technology, the intelligent agent is trained to learn and optimize the cost forecasting strategy independently in the complex market environment. After comparative analysis, compared with the traditional prediction methods that only use the historical average method and linear regression, the prediction model integrating GANs and RL proposed in this study shows obvious advantages. The test-set performance shows that the new model’s prediction error is reduced by about 20%, especially for large-scale and high-complexity engineering projects.