Research on a Carbon Emission Prediction Model for Deep Foundation Pit Construction Based on the BP-GA Algorithm

To address the challenges posed by the multi-source heterogeneous characteristics of carbon emissions in deep foundation pit construction and the insufficient generalization capability of traditional prediction models, this study proposes a hybrid model based on a genetic algorithm-optimized BP (Backpropagation) neural network. The model employs an elite retention strategy and a single-point crossover mechanism to optimize the initial weights of the network. Additionally, it utilizes a two-layer neural network structure with 16 and 8 nodes, respectively, to achieve a nonlinear mapping from input features to carbon emissions. Validation using a dataset of over 1,000 engineering cases demonstrates the models strong predictive performance, achieving a Mean Squared Error (MSE) of 2856.34 and a Mean Absolute Error (MAE) of 43.11. These results highlight the models reliability as a prediction tool for carbon emissions in deep foundation pit projects, offering a promising approach for improving the accuracy of carbon footprint assessments in construction engineering.