Research On Traffic Congestion Prediction Method Based On Data Fusion

With the acceleration of economic development and urbanization,China's motor vehicle ownership has grown rapidly.As of 2017,the number of car ownership in our country reached 217 million,an increase of 23.04 million and an increase of 11.85% compared with 2016,and at the end of 2017,the total mileage of roads was 477.35 million kilometres,an increase of 7.82 million kilometres and an increase of 1.64% compared with 2016.The speed of China's automobile ownership is much higher than that of highways,which has caused the problem of urban increasingly serious traffic congestion.Intelligent Transportation System(ITS)is the application of science and technology in transportation,service control and other fields to strengthen the connection between vehicles,roads and users,thus forming an integrated transportation system that guarantees safety,improves efficiency and saves resources.In the system,accurately predicting the traffic congestion of urban roads and rationally conducting traffic dispatching is currently the main solution to traffic congestion.In this paper,by analyzing various factors that may affect traffic congestion,the research on traffic congestion prediction method combining denoising autoencoder and long-short memory model(LSTM)has important academic value and application prospects.The main research work of the thesis is as follows:(1)In order to accurately predict traffic congestion,the first problem to be solved is the quality of traffic data.Therefore,this paper firstly conducts data fusion research,uses geomagnetic sensors to collect traffic data,and proposes a state machine vehicle detection algorithm based on fixed threshold.Through data fusion analysis and processing,it detects the road segment vehicles and obtains information such as vehicle speed,traffic volume and vehicle proportion.(2)The existing forecasting of traffic congestion rarely analyzes various environmental and social factors.This paper analyzes the impact of traffic flow characteristics,weather and holidays on traffic congestion,combines these factors and uses LSTM models to forecast traffic congestion in 12 sections which effectively improves the accuracy of prediction.(3)Based on the above research work,this paper proposes a traffic congestion prediction method combining denoising autoencoder and LSTM.Using LSTM's characteristics of longterm memory historical data and autoencoder model's core features of extracting data predicts traffic congestion Finally,compared with the existing traffic congestion prediction model,it is verified that the traffic congestion prediction method has higher prediction accuracy and can reach more than 92%.The experimental results show that taking the weather,holiday,time period and other factors into account,the prediction model combined with denoising self-encoding and LSTM can effectively improve the accuracy and robustness of traffic congestion prediction.