Road network state predicting method based on recurrent neural network

Abstract

The invention provides a road network state predicting method based on a recurrent neural network. The road network state predicting method comprises the following steps of 1, establishing a sample set; 2, modeling the recurrent neural network; 3, predicting the road network state of the next moment. The road net state predicting method has the advantages that the road network state evolution rule is acknowledged in a macroscopic viewpoint; the time sequence rule of the road network state change is fully considered by a recurrent neural network algorithm, so as to reach better predicting effect.

Description

technical field [0001] The invention relates to the technical field of public transportation information processing, in particular to a road network state prediction method based on a recursive neural network. Background technique [0002] With the acceleration of urbanization, the problem of traffic congestion is becoming more and more serious, especially in big cities, the problem of traffic congestion is more serious, which seriously affects people's daily travel. Urban roads are becoming increasingly saturated, while car ownership is increasing year by year. This unbalanced relationship between supply and demand has aggravated traffic congestion, and traffic congestion prediction is an important way to alleviate traffic congestion. [0003] In the existing patents, there are already some methods for traffic state prediction. The more mainstream methods include Kalman filter model, time series model, neural network model, parameter regression model, etc. The change of tr...

AI Technical Summary

Problems solved by technology

The time series model not only requires a large amount of historical data, but also has poor time sensitivity to traffic state changes, making it difficult to deal with emergencies. The neural network is very sensitive to parameter initialization, and requires multiple predictions to obtain the average value, which requires a large amount of calculations and local problems. Optimal solution, prone to overfitting, poor parameter portability

The parametric regression model is difficult to express the uncertainty, complexity and dynamic characteristics of the traffic state, etc.

[0004] In addition, most of the existing patents are based on the traffic state prediction at the road section level, and it is difficult to grasp the evolution law of the road network state from a macro perspective, and it is difficult to grasp the complexity and uncertainty of the traffic state

Images