Navigation reference method based on AE-LSTM-BO traffic flow prediction

Abstract

The invention relates to a navigation reference method based on AE-LSTM-BO traffic flow prediction. The method comprises the following steps of firstly, constructing and training an auto-encoder-long-short-term memory network model, inputting the historical data of the current intersection and the historical data of the surrounding intersections into the auto-encoder-long-short-term memory networkmodel, outputting the traffic flow of the current intersection in a certain period of time in the future, and finally, publishing the predicted traffic flow result on the navigation app, so that thenavigation app or a driver can reasonably plan travel and select a navigation route according to the congestion of the traffic flow reference road displayed by different colors. When the model is constructed, dimensionality reduction is carried out on historical data of surrounding intersections, and then LSTM network prediction is carried out; during model training, historical data are firstly acquired, then training of an auto-encoder is carried out, then training of an LSTM network is carried out, and finally Bayesian optimization is carried out to adjust hyper-parameters. The accuracy of traffic flow prediction is significantly improved, and more reasonable navigation reference can be provided.

Description

technical field [0001] The invention belongs to the technical field of navigation and relates to a navigation reference method based on AE-LSTM-BO traffic flow prediction. Background technique [0002] With the development of the economy, the number of private cars is increasing, which brings very serious problems to the traffic in the city. In order to solve this problem, intelligent transportation system came into being. Traffic flow forecasting is a very important link in intelligent transportation systems. A short-term forecasting model with better real-time performance and accuracy is helpful for the analysis and control of traffic conditions. The current navigation reminder method is based on the current road traffic flow as a reference. However, when there is a long time to pass, there will be a situation where a certain road is not congested during travel planning but has serious congestion when driving. Therefore, when planning the travel route, it is necessary to...

AI Technical Summary

Problems solved by technology

The existing traffic volume prediction model only considers the historical data of the current intersection, which is comprehensive in the topological model of the road, and the situation of the associated road needs to be added to the analysis with a certain weight, while the historical data of the surrounding intersections After the introduction, it will be found that because the dimensionality in the training set is too high, the learned neural network will overfit, resulting in a decrease in the generalization performance of the neural network. Therefore, it is necessary to reduce the dimensionality of the data in the training set

Images