Urban Highway Freight Volume Prediction Based on the BP Neural Network

Urban highway transportation is the main mode of freight transportation in China, and its proportion has steadily increased year by year. Combined with the current situation of China's social and industrial economic structure, taking into account the nonlinear and random characteristics of road freight volume itself, a neural network model is established. Using the actual data of all urban highway freight volumes and related influencing factors in China's 34 provinces, the input evaluation index system is determined and the test data set is divided. The evaluation system is divided into three main aspects: socio-demographic economy, urban transportation construction and urban industrial construction. By analyzing the error curve between the results and the real value, the feasibility and reliability of the neural network for predicting urban highway freight volume are verified. Finally, the experimental results are compared with the traditional OLS/panel regression prediction method. The RMSE obtained by the neural network model is 8250 and the interpretability R2 is 0.8560. The RMSE obtained by the panel regression method is as high as 13948 and the interpretability R2 is only 0.1843. The excellent comparison results further clarify the accuracy and reliability of the neural network in predicting urban highway freight volumes.