Traffic congestion index calculation based on BP neural network

Abstract

Aiming at the problem of traffic congestion, the paper analyzed a large number of traffic congestion data in different regions. After analyzing the data, took speed, day of week, bus count, weather and visibility as the most significant factors of traffic congestion time. These factors were preprocessed with Z-score to unify their dimension. In addition, the Bayesian Regularization training algorithm is selected in the BP neural network model to generate the code for predicting traffic congestion time. There is a high correlation between the result of the model and the real record as expected. Then using the BP neural network to analyze the results, get the prediction results and explore the actual deviation to get the advantages and disadvantages of the model, and put forward the improvement and improvement methods in the future.