Research On Prediction Methods Of Urban Road Network Connectivity

In today's information age,the amount of data increases exponentially,and the means and tools for data analysis emerge in endlessly.How to calculate the statistical characteristics of data and find the hidden data evolution law in massive and complex data has always been a hot research issue in the field of data analysis.Complex network theory has been widely used in various fields as an important data analysis method for characterizing data structures,calculating statistical characteristics of data,and deducing future development directions in recent years.The connectivity prediction theory in complex networks provides a convenient and effective calculation tool for the evolution and development of real networks.On the one hand,network connectivity prediction helps to promote the study of the internal structure of the network and promote the further development of complex network theory.On the other hand,for real-world network data,connectivity prediction has a certain practicality.Finding out the missing or possible future edges in the network through the prediction method has a positive effect on revealing the development law of the network.With the continuous development of economy,the original road has been far from meeting the demand of the current city for transportation capacity.The construction of a new and convenient transportation network is the only way for every developing city.The connectivity prediction of the network provides a reference for the adjustment of the road network by giving the possible missing sides in the network.In view of the characteristics of the real urban road network,such as complex structure,large-scale data and sparse road data,this thesis uses the theory of complex network to process and analyze the data,and adopts a combination of network embedding algorithm and data feature extraction algorithm to predict the possible missing links in the network.The specific work of this thesis is as follows:(1)Collect real road data of various cities from OpenStreetMap,and map the data into a complex network of roads after induction and processing.(2)In view of the high computational time and space complexity of the traditional adjacency matrix,and the low efficiency of network processing caused by the strong sparsity of the 'road network,this thesis uses a random walk of nodes to obtain the connection information between nodes.In order to emphasize the role of nodes with larger degree in network representation,the traditional random walk equal probability random node selection method is improved,and the embedding vector of nodes is learned by combining the language model.(3)We introduce the theory of self coding network in deep learning,but the processing ability of single layer self coding network is limited.In this thesis,we use stack self coding method to extract the feature of node representation vector,and add the manic factor to the self coding network to improve the robustness of the model.Finally,we use the result to predict the connectivity of the network.(4)In this thesis,the parameter sensitivity experiment of the algorithm and the comparison experiment with the classical algorithm are carried out respectively.The results show that compared with the traditional network connectivity prediction algorithm,the prediction accuracy of the algorithm used in the thesis has obvious advantages in dealing with large-scale network data.