This article focuses on the cost management of highway tunnel engineering. Under the background of vigorous development of transportation infrastructure, the number of highway tunnel projects has increased, and its accurate cost prediction and reasonable dynamic optimization are of great significance. In this article, machine learning (ML) technology is introduced to build a cost prediction model based on multi-layer perceptron (MLP), and reinforcement learning (RL) is used to realize dynamic optimization of cost. The experimental results show that the MLP model performs well in the training set and the test set by collecting 200 groups of data, 150 groups of training and 50 groups of testing. The training set RMSE is 35.62, MAE is 28.45, and R² is 0.92. The test set RMSE is 38.75, MAE is 31.20, and R² is 0.90. The experiment of dynamic optimization method shows that the cost deviation after optimization is reduced by about 62.0% on average in the face of changes such as project schedule delay and material price increase. The research shows that the method based on ML is feasible and reliable in the cost prediction and dynamic optimization of highway tunnel engineering, which provides scientific and effective support for cost management.
Jinquan Lin (Sun,) studied this question.