Cost Prediction of Construction Projects Based on Boruta Algorithm

The application of deep learning algorithms in the field of construction engineering provides new ideas for engineering cost prediction. To address the issues of inefficient feature processing and insufficient prediction accuracy in traditional methods, a prediction model combining Boruta Feature Selection Algorithm (Boruta) and optimized Backpropagation Neural Network (BPNN) has been proposed. Firstly, the Boruta algorithm is applied to select 12 key features from the 25 dimensional original features to reduce feature dimensionality and eliminate noise. By simulating the foraging and movement behavior of snakes through the snake algorithm, the initial weights and thresholds of BPNN are optimized to enhance the global optimization capability of the model. The experimental findings denote that the MAE of the model on the test set is 4.21, and the RMSE is 5.78, which is more than 35% lower than the traditional BPNN error reduction. Meanwhile, the training time of the model is only 19.6 seconds, balancing prediction accuracy and efficiency. The model proposed by the research can effectively complete the task of engineering cost prediction, providing new technical support for intelligent cost management in the field of construction engineering.