Research On Prediction Method Of Urban Road Traffic State Oriented To Temporal And Spatial Features

With the continuous and rapid development of China’s social economy,the current number of motor vehicles in my country is growing rapidly.On the one hand,it has witnessed the development of China’s social economy and the improvement of people’s living standards.The challenge of urban traffic conditions is deteriorating.Although the road traffic infrastructure in various cities is constantly improving and various traffic travel policies are continuously released,the traffic congestion situation is still not optimistic.Traffic congestion is one of the problems faced by cities,and it is also a major problem that needs to be solved in the process of urban development.With the advent of the mobile Internet era,all mobile devices can become a powerful portrayal of road traffic capacity.With the continuous development of Internet technology,the road traffic data we can collect on mobile devices and road traffic status predictions have entered The era driven by big data.Based on the real-time and historical road condition information and other data provided by the travel platform,this thesis analyzes and processes multi-dimensional data with the help of deep learning related technologies,and builds a neural network-based prediction model to predict the traffic state of urban roads.The specific work of this thesis as follows:Due to the complexity of traffic problems,there are many factors that affect the road traffic state.A single factor cannotreflect the road traffic state,and it is difficult to accurately predict the urban road traffic state.In view of the problems that some prediction models have not fully processed the input data and have not fully explored the influence of factors on the traffic state,this thesis will analyze the factors affecting the traffic state from various aspects,and mine the relationship between the data,such as road length,The road attribute information such as the number of lanes,road grade and road speed limit,the connection information between roads and the historical and real-time road state information and other factors on the traffic state provide ideas for the design of traffic state prediction models.Aiming at the problem that most of the current traffic state prediction models only focus on the time dependence between traffic data information,and do not consider the influence of multi-dimensional features such as spatial features and road attribute features,in this thesis,a spatiotemporal feature-oriented road traffic state prediction method is proposed by integrating time,space,road attribute information and other features.The temporal dimension is modeled based on Gated Recurrent Unit(GRU),the spatial dimension is modeled with a subgraph-based Graph Convolutional Neural Network(GCN),and the road attributes are modeled with a Neural Networked Factorization Machine Model(NFM).Equal category feature modeling.The experimental results based on real data sets show that the model based on GRU+GCN+NFM proposed in this theis has a good prediction effect and can provide auxiliary decision-making for predicting the road traffic state.