Urban Road Network Congestion Prediction Method Based On Multi-source Data Fusion

With the rapid development of the economy and the acceleration of urbanization,the demand for urban transportation has increased substantially,and the rational dispatch of urban traffic is conducive to the efficiency and convenience of residents.The prediction of the occurrence of traffic congestion and the effective prediction of the spread can be used to divert and regulate the impending urban congestion.Accurate and robust traffic state prediction is an important research significance and application value of urban traffic operation.This paper takes the future traffic conditions of urban road network as the research object,mainly uses the whole-day running track of Shanghai taxi as the research object,and obtains the POI data of Shanghai urban area,the weather data of each urban area and the network structure of urban road network.First,data cleaning and preprocessing of multi-source data is performed using Python multitasking parallelism.After that,road matching is performed on the trajectory point of the taxi operation,and the point of the taxi travel record is mapped to the traffic road,and the current state of the road at the current time is congested or unblocked by the traveling speed of the taxi traveling on the road.Due to differences between roads,such as different road widths and road types,it is not possible to divide the state of different roads by only the same speed threshold.This paper proposes a road index to describe the traffic conditions of the road.The clustering algorithm is further used to classify the road state.First of all,due to the frequent congestion of roads and the structure of the road network,the congestion of the roads is cyclical,with periodic and weekly periodicity.Therefore,the time series model is used to mine the rules,and the roads are based on the historical information of the road to predict Future status.In order to study the interaction between congestion between roads and adjacent roads,Bayesian networks were used to study road conditions.This method uses the historical traffic conditions of the road and the spatial structure of the road to predict the future traffic conditions of the road.Since the Bayesian network is directly used for the complete urban road network,the traffic network is first divided by the snake clustering method.The clustering rules of the snake make the roads classified into the same class must be the same as the other ones in the class.The roads are directly adjacent,and the roads with the same road running state are preferentially divided into one class.After the classification of the road is completed,the mixed integer linear programming is used to fine-tune the division result.Finally,the correlation and influence of the traffic state between the roads in the same class are used,and the future road traffic state is predicted based on the historical information of the road peers in the same class.Finally,due to the complexity of the urban traffic network,the road network status is affected by many factors.Road traffic is affected by factors such as working days,precipitation,temperature,wind and spatial distribution of POI.In the case of weather such as rainfall or snowfall,traffic conditions are more likely to be congested than other weather roads.The spatial distribution of POI greatly affects the passage of roads.For example,in the early peak hours,the roads leading to the work area in the residential area are more prone to congestion,and the late peaks are the opposite.Based on the above situation,this paper uses RNN and deep learning network based on feature fusion to predict the road state.RNN input feature is a sequence of time periods,in addition to considering the historical state of road traffic,and adding the influence of weather and other factors on the current state.The deep learning network based on feature fusion pays more attention to the fusion of different features,constructs high-order combination features,and predicts the future road state.This paper uses the data of Shanghai for one month to experiment with the above methods,and verifies the validity and accuracy of the proposed road congestion prediction method.