Urban multi-modal traffic supernetwork situation prediction method based on hypergraph deep network

Abstract

The application discloses a kind of urban multi-mode traffic hypernetwork situation prediction methods based on hypergraph deep network, constructs the traffic graph network of each traffic mode according to the spatial topological relation of subway station, bus stop, road section in urban traffic;The high-order semantic correlation of each traffic mode is analyzed, and the traffic hypergraph network of different traffic modes is constructed;The coupling relationship between different traffic modes is analyzed, and the traffic hypergraph network that fuses urban multi-mode traffic is constructed;Through hypergraph neural network, the hypergraph deep learning model for urban multi-mode traffic hypernetwork collaborative prediction is established, and the model is trained using historical data;Finally, the multi-mode traffic flow situation in future time period is predicted using the trained hypergraph deep learning model.The application more objectively and comprehensively reflects the operation law of urban multi-mode traffic system, and improves the multi-mode traffic flow situation prediction accuracy.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of traffic flow prediction, and particularly relates to a city multi-mode traffic super network situation prediction method based on a super graph deep network. BACKGROUND

[0002] With the construction and continuous promotion of city comprehensive transportation systems, city traffic flow prediction is of great significance to the operation and management of city transportation systems, and has attracted extensive attention in recent years. Graph neural networks (GNNs) have shown significant advantages in traffic prediction problems due to their strong relationship data modeling capabilities, and have been widely applied. However, traditional graph network models model the correlation according to the spatial topological relationship, and can only construct a binary connection relationship between network nodes, that is, a single edge connecting two nodes, and cannot construct high-order correlation between nodes, such as a bus line. At the same time, city comprehensive transportation systems are a complex large system, and there is significant coupling between multiple transportation modes, such as bus stations and subway stations on the same road or bus stations and subway stations with the same name (usually located in the same place), and the traditional graph network model cannot model the coupling relationship between multiple transportation modes. In order to accurately analyze the multi-mode traffic situation of city comprehensive transportation systems, it is urgent to seek a more comprehensive and efficient modeling method to accurately predict the traffic state and development trend of different transportation modes. SUMMARY

[0003] Therefore, the purpose of the present application is to provide a city multi-mode traffic super network situation prediction method based on a super graph deep network to solve the problems of modeling high-order correlation between nodes and coupling between multiple transportation modes in multi-mode traffic prediction.

[0004] The technical scheme for achieving the purpose of the present application is: a city multi-mode traffic super network situation prediction method based on a super graph deep network, the specific steps of which are as follows:

[0005] Step 1: analyze the traffic network to obtain multi-mode traffic network data;

[0006] Step 2: construct a traffic graph network for each transportation mode using spatial topological relationships;

[0007] Step 3: analyze the high-order semantic correlation of each transportation mode and construct a traffic super graph network for different transportation modes;

[0008] Step 4: analyze the coupling relationship between different transportation modes and construct a traffic super graph network that integrates city multi-mode traffic;

[0009] Step 5: establish a super graph deep network for city multi-mode traffic super network collaborative prediction.

[0010] Step 6: Use a hypergraph depth network for collaborative prediction of urban multimodal transportation to predict the multimodal transportation passenger flow trend in the future time period.

[0011] Preferably, the specific method for analyzing the traffic network in step 1 to obtain multimodal traffic network data is as follows:

[0012] Step 1.1: Analyze the multi-modal transportation network of subway, bus, and taxi to obtain the node data of each transportation mode network;

[0013] Step 1.2: Analyze the topology of the multi-modal transportation network of subway, bus, and taxi to obtain spatial topology data for each mode of transportation.

[0014] Preferably, for subway transportation modes, subway stations are extracted as nodes in the subway transportation mode graph network, forming a set of subway transportation graph network nodes. I (m) The number of subway stations is given; for public transportation modes, bus stops are extracted as nodes in the public transportation mode graph network, forming a set of nodes for the public transportation graph network. I (b) The number of bus stops is used; for taxi transportation, the road segments that taxis can travel on are extracted as nodes in the taxi transportation mode graph network, forming a set of nodes in the taxi transportation graph network. I (r) This refers to the number of road segments.

[0015] Preferably, the specific method for constructing a traffic map network for each mode of transportation using spatial topological relationships is as follows:

[0016] For subway transportation, utilize subway transportation map network node V (m) Adjacency matrix A with subway transportation modes (m) Constructing a transportation map network G for subway modes (m) =(V (m) A (m) );

[0017] For public transportation modes, utilize public transportation map network node V (b) Adjacency matrix A with public transport modes (b) Constructing a transit map network G for public transport modes (b) =(V (b) A (b) );

[0018] For taxi transportation, the taxi traffic map network node V is used. (r) Adjacency matrix A of taxi transportation modes (r)Constructing a transit map network G for public transport modes (r) =(V (r) A (r) ).

[0019] Preferably, the specific steps for analyzing the high-order semantic relevance of each mode of transportation and constructing a transportation hypergraph network for different modes of transportation are as follows:

[0020] Step 3.1: For subway transportation, establish a hyperedge ε using higher-order semantic relations. (m) Constructing a metro transit supermap network Specifically:

[0021] Step 3.1.1: For subway transportation modes, establish a hyperedge using the semantic relation "subway stations belong to the same subway line". J (m) For super-edge Quantity;

[0022] Step 3.1.2: Utilize the hyperedge set ε (m) Constructing a metro transit supermap network in, This is the set of nodes in the metro transportation supermap network. The incidence matrix is ​​defined as follows:

[0023]

[0024] in, and They are the i-th node and the j-th hyperedge in the metro transportation hypergraph network, respectively. (m) With J (m) These represent the number of nodes and the number of superedges, respectively. (m) (i,j) is The element in the i-th row and j-th column;

[0025] Step 3.2: For public transportation modes, establish a hyperedge ε using higher-order semantic relations. (b) Constructing a public transport supermap network Specifically:

[0026] Step 3.2.1: For public transportation modes, establish a hyperedge using the semantic relation "bus stops belong to the same bus route". J (b) For super-edge Quantity;

[0027] Step 3.2.2: Utilize the hyperedge set ε (b) Building a public transport supermap network in, This is the set of nodes in the public transport hypergraph network. The incidence matrix is ​​defined as follows:

[0028]

[0029] in, and They are the i-th node and the j-th hyperedge in the public transport hypergraph network, respectively. (b) With J (b) These represent the number of nodes and the number of superedges, respectively. (b) (i,j) is The element in the i-th row and j-th column.

[0030] Preferably, the specific steps for analyzing the coupling relationships between different modes of transportation and constructing a transportation hypergraph network that integrates urban multi-modal transportation are as follows:

[0031] Step 4.1: Analyze the semantic relationships between multimodal transportation, establish hyperedges between pairs of subnetworks, and construct a transportation hypergraph network between pairs of subnetworks;

[0032] Step 4.2: Analyze the coupling relationships among the three sub-networks of subway, bus, and taxi. Establish hyperedges using the semantic relation "subway and bus stops have the same name and are on the same road segment," and construct a transportation hypergraph network integrating the three modes of transportation (subway, bus, and taxi) based on these hyperedges. in For a set of nodes, This is the set of nodes in the metro transportation supergraph network. This is the set of nodes in the public transport hypergraph network. The incidence matrix is ​​defined as follows:

[0033]

[0034] in, and These represent the i-th node and the j-th hyperedge in the subway-bus-taxi transportation hypergraph network, respectively. (m,b,r) =I (m) +I (b) +I (r) With J (m,b,r) These represent the number of nodes and the number of superedges, respectively. (m ,b,r) (i,j) is The element in the i-th row and j-th column.

[0035] Preferably, the specific method for analyzing the semantic relationships between multimodal transportation, establishing hyperedges between pairs of subnetworks, and constructing a transportation hypergraph network between pairs of subnetworks is as follows:

[0036] Step 4.1.1: For metro-bus network, hyper-edges are established by semantic relation "station name of metro station is same as station name of bus station" wherein and a metro-bus traffic hypergraph network is constructed wherein is a node set, is an incidence matrix defined as shown below,

[0037]

[0038] wherein, and are the ith node and jth hyper-edge in the metro-bus traffic hypergraph network, respectively, I (m,b) = I (m) + I (b) and J (m,b) are the number of nodes and hyper-edges, respectively, h (m,b) (i,j) is the ith row and jth column element of

[0039] Step 4.1.2: For metro-taxi network, hyper-edges are established by semantic relation "metro station is on a certain road segment" wherein and a metro-taxi traffic hypergraph network is constructed wherein is a node set, is an incidence matrix defined as shown below,

[0040]

[0041] wherein, and are the ith node and jth hyper-edge in the metro-taxi traffic hypergraph network, respectively, I (m,r) = I (m) + I (r) and J (m,r) are the number of nodes and hyper-edges, respectively, h (m,r) (i,j) is the ith row and jth column element of

[0042] Step 4.1.3: For bus-taxi network, hyper-edges are established by semantic relation "bus station is on a certain road segment" wherein and a bus-taxi traffic hypergraph network is constructed wherein is a node set, is an incidence matrix defined as shown below,

[0043] ​​

[0044] in, and These represent the i-th node and the j-th hyperedge in the bus-taxi transportation hypergraph network, respectively. (b,r) =I (b) +I (r) With J (b,r) These represent the number of nodes and the number of superedges, respectively. (b,r) (i,j) is The element in the i-th row and j-th column.

[0045] Preferably, the specific method for establishing a hypergraph depth network for collaborative prediction of urban multimodal transportation hypernetworks is as follows:

[0046] Step 5.1: Construct a transportation hypergraph network that integrates multiple modes of transportation, including subway, bus, and taxi. Specifically:

[0047] Step 5.1.1: Combine the graph network G constructed in steps 2 to 4. (m) G (b) G (r) With Hypergraph Network To achieve this, a hyperedge is established in the integrated hypergraph network of multiple transportation modes, including subway, bus, and taxi, as shown below.

[0048] ε={e j |j=1,2,…,J}=ε (1) ∪ε (2) ∪ε (3) ∪ε (m) ∪ε (b) ∪ε (m,b) ∪ε (m,r) ∪ε (b,r) ∪ε (m,b,r)

[0049] in,

[0050]

[0051]

[0052]

[0053] Among them, A (m) (i,j), A (b) (i,j) and A (r) (i,j) represent matrix A. (m) A (b) With A (r) The element in the i-th row and j-th column;

[0054] Step 5.1.2: Constructing the traffic hypergraph network of fused subway-bus-taxi multi-modal traffic by using the super-edge set ε wherein is the node set, is the incidence matrix, which is defined as shown below,

[0055]

[0056] wherein, and e j are the i-th node and the j-th hyperedge in the multi-modal traffic hypergraph network, respectively, I = I (m) + I (b) + I (r) is the number of nodes of the hypergraph network; J = I (m) + I (b) + I (r) + J (m) + J (b) + J (m,b) + J (m,r) + J (b,r) + J (m,b,r) is the number of hyperedges of the hypergraph network;

[0057] Step 5.2: Constructing the feature matrix X(t) and the prediction label y(t) by using the historical data;

[0058] Step 5.3: Establishing the hypergraph deep learning model f for multi-modal traffic collaborative prediction, and the optimization objective function of the hypergraph deep learning model is as shown below,

[0059]

[0060] wherein, is the empirical loss function, Ω reg (·) is the regularization function;

[0061] Step 5.4: Inputting the sample data into the hypergraph deep network model to train the same, so as to obtain the multi-modal traffic collaborative prediction model

[0062] Preferably, the method for constructing the feature matrix X(t) and the prediction label y(t) by using the historical data is specifically as follows:

[0063] Step 5.2.1: For the time period t, the hypergraph network in the current time period t and T-1 historical time periods is used wherein is the feature vector composed of the traffic flow of each node of the hypergraph network in the time period t;

[0064] Step 5.2.2: for time period t, using the traffic flow on each node of the hypergraph network in the next time period t+1 to constitute the prediction label of the current time period

[0065] Preferably, the specific method for predicting the multi-modal traffic flow situation in the future time period by the hypergraph deep network for urban multi-modal traffic hypernetwork collaborative prediction is as follows:

[0066] Step 6.1: for any time period t', using the traffic flow data on each node of the hypergraph network in the current time period t' and T-1 historical time periods as the hypergraph network feature matrix of the current time period wherein is the feature vector composed of the traffic flow on each node of the hypergraph network in the time period t';

[0067] Step 6.2: inputting the feature matrix X(t') and the hypergraph correlation matrix constructed in step 5.1 into the multi-modal traffic collaborative prediction model obtained in step 5.4 to predict the multi-modal traffic flow in the future time period, as follows,

[0068]

[0069] wherein, is the predicted value of the traffic flow on each node of the hypergraph network in the next time period t'+1.

[0070] Compared with the prior art, the present application has the following advantages: the present application simultaneously models the high-order correlation between the traffic network nodes and the coupling relationship between the multi-modal traffic in a unified manner, simultaneously performs collaborative prediction on the multi-modal traffic flow situation, can more objectively and comprehensively reflect the operation rules of the urban multi-modal traffic system, and thus improves the accuracy of the multi-modal traffic flow situation prediction.

[0071] The present application will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 is the flow chart of the method of the present application.

[0073] Figure 2 is the schematic diagram of the subway and bus hypergraph network in the present application.

[0074] Figure 3 is the schematic diagram of the hypergraph network fusing the subway, bus and taxi multi-modal traffic in the present application.

[0075] Figure 4 ​​A schematic diagram of the fusion method of the adjacency matrix and the incidence matrix in the application is shown. DETAILED DESCRIPTION

[0076] Referring to Figure 1 The application provides a city multi-mode traffic super network situation prediction method based on a hypergraph deep network, and the specific steps are as follows:

[0077] Step 1: analyzing the traffic network in the research area to obtain multi-mode traffic network data;

[0078] Step 2: constructing a traffic graph network of each traffic mode by using spatial topological relations;

[0079] Step 3: analyzing the high-order semantic correlation of each traffic mode to construct a traffic hypergraph network of different traffic modes;

[0080] Step 4: analyzing the coupling relationship between different traffic modes to construct a traffic hypergraph network that fuses city multi-mode traffic;

[0081] Step 5: establishing a hypergraph deep network for city multi-mode traffic super network collaborative prediction;

[0082] Step 6: predicting the multi-mode traffic flow situation in a future time period.

[0083] In this embodiment, step 1 is specifically as follows:

[0084] Step 1.1: analyzing the subway, bus, and taxi multi-mode traffic network in the area near Xinjiekou in Nanjing to obtain node data V (m) , V (b) , and V (r) of the graph network of each traffic mode;

[0085] Step 1.2: analyzing the topological structure of the subway, bus, and taxi multi-mode traffic network to obtain spatial topological structure data A (m) , A (b) , and A (r) of each traffic mode.

[0086] In this embodiment, step 1.1 is specifically as follows:

[0087] Step 1.1.1: for the subway traffic mode, extracting subway stations as nodes of the graph network of the traffic mode to form a subway traffic graph network node set V 4 is the number of subway stations;

[0088] Step 1.1.2: for the bus traffic mode, extracting bus stations as nodes of the graph network of the traffic mode to form a bus traffic graph network node set V 7 is the number of bus stations;

[0089] Step 1.1.3: For the taxi traffic mode, extract the road segments that it can pass through as the nodes of the traffic mode graph network, to form a set of taxi traffic graph network nodes 16The number of road segments in the study area where taxis travel.

[0090] In this embodiment, the step 1.2 is specifically:

[0091] Step 1.2.1: For the subway traffic mode, use the spatial topological structure relationship between subway stations to construct the adjacency matrix of the traffic mode graph network

[0092] Step 1.2.2: For the bus traffic mode, use the spatial topological structure relationship between bus stations to construct the adjacency matrix of the traffic mode graph network

[0093] Step 1.2.3: For the taxi traffic mode, use the spatial topological structure relationship between road segments to construct the adjacency matrix of the traffic mode graph network

[0094] In this embodiment, the step 2 is specifically:

[0095] Step 2.1: For the subway traffic mode, use the network nodes V (m) and the adjacency matrix A (m) to construct the traffic graph network G (m) =(V (m) ,A (m) ) of the traffic mode;

[0096] Step 2.2: For the bus traffic mode, use the network nodes V (b) and the adjacency matrix A (b) to construct the traffic graph network G (b) =(V (b) ,A (b) ) of the traffic mode;

[0097] Step 2.3: For the taxi traffic mode, use the network nodes V (r) and the adjacency matrix A (r) to construct the traffic graph network G (r) =(V (r) ,A (r) ) of the traffic mode.

[0098] In this embodiment, the step 3 is specifically:

[0099] Step 3.1: For the subway traffic mode, use the high-order semantic relationship to establish the super edge ε (m), constructing the subway traffic hypergraph network

[0100] Step 3.2: for the bus traffic mode, the hyperedge ε is established by using high-order semantic relationship (b) , constructing the bus traffic hypergraph network

[0101] In this embodiment, the step 3.1 is specifically:

[0102] Step 3.1.1: for the subway traffic mode, the hyperedge is established by using the semantic relationship "subway station belongs to the same subway line"

[0103] Step 3.1.2: the hyperedge set ε is used (m) constructing the subway traffic hypergraph network wherein, is the node set of the subway traffic hypergraph network, is the incidence matrix, which is defined as shown below,

[0104]

[0105] wherein, and are the i th node and the j th hyperedge in the subway traffic hypergraph network respectively.

[0106] In this embodiment, the step 3.2 is specifically:

[0107] Step 3.2.1: for the bus traffic mode, the hyperedge is established by using the semantic relationship "bus station belongs to the same bus line"

[0108] Step 3.2.2: the hyperedge set ε is used (b) constructing the bus traffic hypergraph network wherein, is the node set of the bus traffic hypergraph network, is the incidence matrix, which is defined as shown below,

[0109]

[0110] wherein, and are the i th node and the j th hyperedge in the bus traffic hypergraph network respectively.

[0111] In this embodiment, the step 4 is specifically:

[0112] Step 4.1: analyzing the semantic relationship between multi-mode traffic, establishing the hyperedge ε between two two sub-networks (m,b) , ε(m,r) with ε (b,r) , constructing traffic hypergraph network between each pair of sub-networks with

[0113] Step 4.2: Analyzing the coupling relationship between the subway, bus and taxi sub-networks, and using the semantic relationship "the station names of subway stations and bus stations are the same and on the same road section" to establish hyper-edges

[0114] wherein and thereby constructing a traffic hypergraph network integrating subway-bus-taxi traffic modes

[0115] wherein is a node set, is an association matrix, which is defined as shown below,

[0116]

[0117] wherein, and are the i-th node and the j-th hyper-edge in the subway-bus-taxi traffic hypergraph network, respectively.

[0118] In this embodiment, the step 4.1 is specifically:

[0119] Step 4.1.1: For the subway-bus network, using the semantic relationship "the station names of subway stations and bus stations are the same" to establish hyper-edges wherein and thereby constructing a subway-bus traffic hypergraph network wherein is a node set, is an association matrix, which is defined as shown below,

[0120]

[0121] wherein, and are the i-th node and the j-th hyper-edge in the subway-bus traffic hypergraph network, respectively;

[0122] Step 4.1.2: For the subway-taxi network, using the semantic relationship "subway stations are on a certain road section" to establish hyper-edges wherein and thereby constructing a subway-taxi traffic hypergraph network wherein is a node set, is an association matrix, which is defined as shown below,

[0123]

[0124] wherein, with respectively the i-th node and the j-th hyperedge in the subway-taxi traffic hypergraph network;

[0125] Step 4.1.3: for the bus-taxi network, the hyperedge is established by using the semantic relationship "bus station on a certain road section" wherein and a bus-taxi traffic hypergraph network is constructed wherein is a node set, is an association matrix, which is defined as shown below,

[0126]

[0127] wherein, with respectively the i-th node and the j-th hyperedge in the bus-taxi traffic hypergraph network;

[0128] In this embodiment, the step 5 is specifically:

[0129] Step 5.1: constructing a traffic hypergraph network integrating subway-bus-taxi multi-mode traffic

[0130] Step 5.2: constructing a feature matrix X(t) and a prediction label y(t) by using historical data;

[0131] Step 5.3: establishing a hypergraph deep learning model f oriented to multi-mode traffic collaborative prediction;

[0132] Step 5.4: inputting sample data into the hypergraph deep network model to train the same, and obtaining a multi-mode traffic collaborative prediction model

[0133] In this embodiment, the step 5.1 is specifically:

[0134] Step 5.1.1: integrating the graph networks G (m) , G (b) , G (r) and the hypergraph network to establish a hyperedge of a hypergraph network integrating subway-bus-taxi multi-mode traffic, as shown below,

[0135] ε={e j |j=1,2,…,J}=ε (1) ∪ε (2) ∪ε (3)∪ε (m) ∪ε (b) ∪ε (m,b) ∪ε (m,r) ∪ε (b,r) ∪ε (m,b,r)

[0136] wherein,

[0137]

[0138]

[0139]

[0140] wherein, A (m) (i,j), A (b) (i,j) and A (r) (i,j) are the i-th row and j-th column elements of matrix A (m) , A (b) and A (r) , respectively.

[0141] Step 5.1.2: Constructing the traffic hypergraph network integrating metro-bus-taxi multi-modal traffic by using the hyperedge set ε wherein is the node set, is the incidence matrix, which is defined as shown below,

[0142]

[0143] wherein, and e j ∈ε are the i-th node and j-th hyperedge in the multi-modal traffic hypergraph network, respectively, I = I (m) + I (b) + I (r) = 27 and J = I (m) + I (b) + I (r) + J (m) + J (b) + J (m,b) + J (m,r) + J (b,r) + J (m,b,r) = 49 are the number of nodes and hyperedges of the hypergraph network, respectively.

[0144] In this embodiment, the step 5.2 is specifically:

[0145] Step 5.2.1: For the time period t, using the hypergraph network in the current time period t and 5 historical time periods and the traffic flow data on each node as the hypergraph network feature matrix wherein t is 15 minutes, is a feature vector composed of traffic flow of each node of the hypergraph network in the current 15 minutes;

[0146] Step 5.2.2: For the time period t, the traffic flow on each node of the hypergraph network in the next time period t+1 (i.e. the next 15 minutes) is used to construct the prediction label of the current time period

[0147] In this embodiment, the step 5.3 is specifically that a hypergraph deep learning model f for multi-modal traffic collaborative prediction is established by using the hypergraph convolutional neural network model, and the optimization objective function is as follows,

[0148]

[0149] wherein, is an empirical loss function, and Ω reg (·) is a regularization function.

[0150] In this embodiment, the step 5.4 is specifically that the feature matrix X(t) constructed in the step 5.2, the prediction label y(t) and the hypergraph association matrix A constructed in the step 5.1 are input into the hypergraph deep learning model f established in the step 5.3, The hypergraph deep learning model f established in the step 5.3 is input, and the model is trained by using the gradient descent method to obtain an optimized multi-modal traffic collaborative prediction model

[0151] In this embodiment, the step 6 is specifically that:

[0152] Step 6.1: For any time period t', the traffic flow data on each node of the hypergraph network in the current time period t' and 5 historical time periods are used as the hypergraph network feature matrix X(t') of the current time period wherein, is a feature vector composed of traffic flow of each node of the hypergraph network in the time period t';

[0153] Step 6.2: The feature matrix X(t') and the hypergraph association matrix A constructed in the step 5.1 are input into the multi-modal traffic collaborative prediction model obtained in the step 5.4 The multi-modal traffic collaborative prediction model obtained in the step 5.4 is input, The multi-modal traffic flow in the future time period is predicted, as follows,

[0154]

[0155] wherein, is a prediction value of traffic flow of each node of the hypergraph network in the next time period t'+1.

[0156] ​The above merely describes preferred embodiments of the present application, and any equivalent changes and modifications made within the scope of the present application should be included in the scope of the present application.

Claims

1. A city multi-mode traffic hypernetwork situation prediction method based on hypergraph deep network, characterized in that, The specific steps are: Step 1: Analyze the traffic network to obtain multi-modal traffic network data; Step 2: Construct traffic graph networks for each mode of transportation using spatial topological relationships; Step 3: Analyze the high-order semantic correlation of each mode of transportation and construct traffic supergraph networks for different modes of transportation; Step 4: Analyze the coupling relationship between different modes of transportation and construct traffic supergraph networks that integrate multi-modal urban transportation, with the specific steps being: Step 4.1: Analyze the semantic relationship between multi-modal transportation and establish superedges between two subnetworks to construct traffic supergraph networks between two subnetworks, with the specific method being: Step 4.1.1: For the metro-bus network, the hyperedge is established using the semantic relationship "the station name of the metro station is the same as the bus station" wherein and the metro-bus traffic hypergraph network is constructed wherein is the node set, is the incidence matrix, which is defined as shown below, wherein, and are the ith node and jth hyperedge in the subway-bus transport hypergraph network, respectively, I (m,b) = I (m) + I (b) and J (m,b) are the number of nodes and hyperedges, respectively, h (m,b) (i,j) is the element in the ith row and jth column of . Step 4.1.2: For the subway-taxi network, the hyperedge is built using the semantic relation "subway station on a certain road segment" where and the subway-taxi traffic hypergraph network is constructed where is the node set, is the incidence matrix, which is defined as shown below, wherein, and are the ith node and jth hyperedge in the subway-taxi traffic hypergraph network, respectively, I (m,r) = I (m) + I (r) and J (m,r) are the number of nodes and hyperedges, respectively, h (m,r) (i,j) is the element in the ith row and jth column of . Step 4.1.3: For the bus-taxi network, the hyperedge is built using the semantic relation "bus stop is on a certain road segment" wherein and the bus-taxi traffic hypergraph network is built wherein is the set of nodes, is the incidence matrix, which is defined as shown below, wherein, and are the ith node and jth hyperedge in the bus-taxi traffic hypergraph network, respectively, I (b,r) = I (b) + I (r) and J (b,r) are the number of nodes and hyperedges, respectively, h (b,r) (i,j) is the element in the ith row and jth column of . Step 4.2: Analyze the coupling relationship between the three sub-networks of subway, bus, and taxi, establish hyper-edges using the semantic relationship "subway station and bus station have the same station name and are on the same road segment", and construct a traffic hypergraph network that integrates the three traffic modes of subway, bus, and taxi according to the hyper-edges wherein is a node set, is a node set of the subway traffic hypergraph network, is a node set of the bus traffic hypergraph network, is a correlation matrix, which is defined as shown below, wherein, and are the ith node and jth hyperedge in the subway-bus-taxi traffic hypergraph network, respectively, I (m,b,r) = I (m) + I (b) + I (r) and J (m,b,r) are the number of nodes and hyperedges, respectively, h (m,b,r) (i,j) is the element in the ith row and jth column of . Step 5: Establish a supergraph deep network for multi-modal urban traffic super-network collaborative prediction; Step 6: Use the supergraph deep network for multi-modal urban traffic super-network collaborative prediction to predict the future traffic flow situation in the future time period.

2. The urban multi-modal traffic hypernetwork situation prediction method based on hypergraph deep network according to claim 1, characterized in that, The specific method for analyzing the traffic network in Step 1 to obtain multi-modal traffic network data is: Step 1.1: Analyze the subway, bus, and taxi multi-modal transportation network to obtain node data for each mode of transportation graph network; Step 1.2: Analyze the topological structure of the subway, bus, and taxi multi-modal transportation network to obtain spatial topological structure data for each mode of transportation.

3. The urban multi-modal traffic hypernetwork situation prediction method based on hypergraph deep network according to claim 2, characterized in that, For the subway traffic mode, extract the subway station as the node of the subway traffic mode graph network, and constitute the subway traffic graph network node set I (m) The number of subway stations; for the bus traffic mode, extract the bus station as the node of the bus traffic mode graph network, and constitute the bus traffic graph network node set I (b) The number of bus stations; for the taxi traffic mode, extract the road segment that the taxi can pass through as the node of the taxi traffic mode graph network, and constitute the taxi traffic graph network node set I (r) The number of road segments.

4. The urban multi-modal traffic hypernetwork situation prediction method based on hypergraph deep network according to claim 1, characterized in that, The specific method for constructing traffic graph networks for each mode of transportation using spatial topological relationships is: For the subway traffic mode, the subway traffic graph network node V (m) The subway traffic mode adjacency matrix A (m) The subway traffic mode traffic graph network G (m) =(V (m) ,A (m) ) For the public transport mode, the public transport graph network node V (b) The public transport mode adjacency matrix A (b) Construct the public transport graph network G (b) =(V (b) ,A (b) ) For the taxi traffic mode, the network nodes V (r) of the taxi traffic graph are determined by the taxi traffic graph adjacency matrix A (r) For the bus traffic mode, the network nodes V (r) of the bus traffic graph are determined by the bus traffic graph adjacency matrix A (r) , (r) .

5. The urban multi-modal traffic hypernetwork situation prediction method based on hypergraph deep network according to claim 1, characterized in that, The specific steps for analyzing the high-order semantic correlation of each mode of transportation and constructing traffic supergraph networks for different modes of transportation are: Step 3.1: For the subway transportation mode, the hyperedge is established using high-order semantic relations Constructing the subway transportation hypergraph network Specifically: Step 3.1.1: For the metro transportation mode, the hyperedge is built using the semantic relation "a subway station belongs to the same subway line" J (m) The number of hyperedges ​ Step 3.1.2: Utilizing the super-edge set Constructing the metro traffic hypergraph network wherein, is a set of nodes of the metro traffic hypergraph network, is a incidence matrix defined as shown below, wherein, and are the ith node and jth hyperedge in the subway traffic hypergraph network, respectively, I (m) and J (m) are the number of nodes and hyperedges, respectively, h (m) (i,j) is the element in the ith row and jth column of . Step 3.2: For the public transportation mode, the hyper-edge is established using the high-order semantic relationship Constructing the public transportation hypergraph network Specifically: Step 3.2.1: For the public transportation mode, hyper-edges are built using the semantic relation "bus stop belongs to the same bus line" J (b) The number of hyper-edges ​ Step 3.2.2: Utilizing the super-edge set Constructing the public transit supergraph network wherein, is a set of nodes of the public transit supergraph network, is a connection matrix defined as shown below, wherein, and are the ith node and jth hyperedge in the public transit hypergraph network, respectively, I (b) and J (b) are the number of nodes and hyperedges, respectively, h (b) (i,j) is the element in the ith row and jth column of .

6. The urban multi-modal traffic hypernetwork situation prediction method based on hypergraph deep network according to claim 4, characterized in that, The specific method for establishing a supergraph deep network for multi-modal urban traffic super-network collaborative prediction is: Step 5.1: Constructing the traffic hypernetwork of fusion subway-bus-taxi multi-modal traffic Specifically: Step 5.1.1: Constructing the graph network G from steps 2-4 (m) , G (b) , G (r) and the hypergraph network fusion, the hyperedge of the fusion subway-bus-taxi multi-mode traffic hypergraph network is established as follows, Wherein, where A (m) (i,j), A (b) (i,j), and A (r) (i,j) are the elements of matrix A (m) , A (b) , and A (r) , respectively, in the i-th row and j-th column. Step 5.1.2: Constructing the traffic hypergraph network of fused subway-bus-taxi multi-modal transportation with hyper-edge set ε wherein is a node set, is an incidence matrix, which is defined as shown below, wherein, with are the i-th node and the j-th hyperedge in the multi-modal transportation hypergraph network, respectively, I = 1, 2,..., I (m) +I (b) +I (r) is the number of nodes of the hypergraph network; J = 1, 2,..., J (m) +I (b) +I (r) +J (m) +J (b) +J (m,b) +J (m,r) +J (b,r) +J (m,b,r) is the number of hyperedges of the hypergraph network; Step 5.2: Use historical data to construct a feature matrix X(t) and a prediction label y(t); Step 5.3: Establish a supergraph deep learning model f for multi-modal transportation collaborative prediction, with the optimization objective function of the supergraph deep learning model being as follows, wherein, is an empirical loss function, Ω reg (·) is a regularizing function; Step 5.4: input sample data into the hypergraph deep network model to train it, and obtain a multi-mode traffic collaborative prediction model 7. The urban multi-modal traffic hypernetwork situation prediction method based on hypergraph deep network according to claim 6, characterized in that, The method for constructing a feature matrix X(t) and a prediction label y(t) using historical data is: Step 5.2.1: For time period t, utilize the current time period t and T-1 historical time periods of the hypergraph network The traffic flow data on each node as the hypergraph network feature matrix of the current time period Wherein is the feature vector composed of the traffic flow of each node of the hypergraph network in the time period t; Step 5.2.2: For a time period t, utilize the traffic flow on each node of the hypergraph network in the next time period t+1 to form the prediction label for the current time period t 8.The urban multi-modal transportation hypernetwork situation prediction method based on hypergraph deep network according to claim 7, characterized in that, The specific method for using the supergraph deep network for multi-modal urban traffic super-network collaborative prediction to predict the future traffic flow situation in the future time period is: Step 6.1: For any time period t', utilize the current time period t' and T-1 historical time periods of the hypergraph network Traffic flow data on each node as the hypergraph network feature matrix of the current time period wherein is the feature vector composed of traffic flow of each node of the hypergraph network in time period t'; Step 6.2: The feature matrix X(t') is associated with the supergraph matrix constructed in step 5.1 The multi-modal traffic collaborative prediction model obtained in step 5.4 is input The multi-modal traffic passenger flow in the future time period is predicted, as follows, wherein, is the predicted value of the traffic flow of each node of the hypergraph network for the next time period t' + 1.

Images