Prediction method for expressway travel time based on charge data

Abstract

The invention relates to a prediction method for the expressway travel time based on charge data. The prediction method comprises the steps of S1, acquiring charge data of a predetermined expressway; S2, preprocessing the charge data of the expressway; S3, sampling and processing a preprocessed data set so as to acquire a sample set; S4, establishing a prediction model combined by an auto-encoder and a BP neural network by using the sample set; and S5, carrying out prediction on the expressway travel time by using the model. By adopting the technical scheme provided by the invention, the stability of the model is greatly improved because the auto-encoder has a characteristic of adaptive learning and is not sensitive to network initialization.

Description

technical field [0001] The invention relates to the technical field of intelligent transportation, in particular to a prediction method of expressway travel time based on toll data. Background technique [0002] In recent years, in the process of China's economic development, the scale of cities has also continued to expand. Expressways, as an indispensable part of land transportation, play an important role in people's travel and logistics. The continuous increase in the number of motor vehicles has caused traffic congestion, reduced the efficiency of expressway operation, and made the travel time of cars on the expressway uncertain. Accurately predicting the travel time of cars on expressways has become an important research topic at home and abroad. [0003] Expressway travel time prediction can provide expressway users with real-time and accurate information, help expressway users to make decisions about travel time and expressway section selection, plan travel routes r...

AI Technical Summary

Problems solved by technology

Among them, the Kalman filter model mostly uses the data collected by floating cars, and its data volume is limited and relatively simple; the BP neural network model mostly uses the data collected by fixed monitors (including traffic flow, density, speed and other information), and this method has higher Real-time and high precision, but because the neural network is sensitive to network initialization, multiple averages are required to determine the final forecast value, which also increases the amount of calculation, and requires a large amount of money to purchase and install equipment

[0005] These models perform well in short-term predictions. However, as the prediction depth (such as predicting the travel time of a highway section after one hour) continues to increase, the error will increase significantly.

Images