A method and device for predicting dynamic traffic events

Abstract

The invention discloses a dynamic traffic event prediction method and device. The method comprises the following steps: training a neural network model for road condition prediction by using a true value sample of road condition information to obtain a middle and lower layer structure and parameters of the neural network model; establishing a neural network model of dynamic traffic event prediction, wherein a middle and lower layer structure and parameters of the neural network model of dynamic traffic event prediction are fixed as the middle and lower layer structure and parameters of the neural network model of road condition prediction; training the neural network model of the dynamic traffic event prediction using a true value sample of the dynamic traffic event to obtain an upper structure and parameters thereof; performing the dynamic traffic event prediction by the neural network model of the dynamic traffic event prediction. This scheme can effectively reduce the number of freeparameters in the neural network model of dynamic traffic event prediction and fundamentally improve the recall and accuracy of dynamic traffic event prediction.

Description

technical field [0001] The invention relates to the field of traffic technology, in particular to a dynamic traffic event prediction method and device. Background technique [0002] With the continuous increase of vehicles, roads are becoming more and more congested. There are many reasons for road traffic congestion, such as poor road conditions, large traffic volume, etc., and traffic incidents are also one of the main reasons for road congestion. Traffic events are unpredictable, including accidents, temporary traffic restrictions, and unannounced road construction. [0003] In the process of providing navigation services for users, it is necessary to identify these traffic events as soon as possible and accurately, and use the information of these traffic events in navigation, so as to provide users with correct navigation suggestions. To this end, traffic incidents can be detected by training a predictive model to identify anomalies in road traffic. However, it is dif...

AI Technical Summary

Problems solved by technology

However, it is difficult to obtain and confirm the current dynamic traffic incidents, mainly because the traffic incidents are likely to have been resolved before the inspectors arrive at the scene. In addition, the method of verifying traffic incidents through on-site personnel is also expensive in manpower and financial resources, and cannot be popularized.

[0004] Therefore, it is currently difficult to obtain accurate dynamic traffic events as real samples in a timely manner, resulting in the inability to train a suitable prediction model, and thus it is difficult to make accurate predictions on dynamic traffic events

Images