A problem diagnosis method for urban traffic networks

Abstract

The present invention relates to a problem diagnosis method for an urban traffic network, including: predefined traffic road network features and traffic problems existing in the traffic road network as entities in a traffic knowledge graph; constructing a traffic knowledge graph by using the predefined entities and diagnosed cases; inputting the traffic knowledge graph into an encoder based on a graph neural network to obtain graph embedding vectors of each node in the traffic knowledge graph; and matching the features of the intersection to be diagnosed with the predefined entities, and inputting the matched entities into a decoder based on entity fusion for the decoder to perform operations and output a diagnosis result.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic network problem diagnosis, and particularly to a problem diagnosis method for urban traffic networks. Background Art

[0002] During the operation of the urban road traffic system, many problems may occur. These problems may affect traffic efficiency, cause traffic congestion, or lead to traffic accidents, reducing traffic safety. These problems can be solved by traffic managers' scheduling of road spatio-temporal resources and increasing the law enforcement intensity of managers. For example, when traffic demand increases, optimizing traffic lights or changing lane functions, or strictly investigating and punishing illegal traffic participants to improve the operation process of the urban road traffic system.

[0003] In response to the above problems, existing solutions mostly rely on expert experience, that is, a large number of experienced experts observe at intersections, diagnose corresponding problems, and give relevant solution strategies. This manual diagnosis solution has the following deficiencies: 1. Low efficiency. Experts cannot diagnose each case in a large-scale road network, so that often only a few intersections with the most serious problems are diagnosed and analyzed; 2. The diagnosis results of each case depend on the experience of experts, and the accuracy and objectivity of the diagnosis results cannot be guaranteed; 3. This solution does not have generalization and transferability, that is, the experience of experts cannot be efficiently inherited and promoted to different cities. Summary of the Invention

[0004] To solve at least some of the above problems in the prior art, the present invention provides a problem diagnosis method for urban traffic networks, including:

[0005] Predefine traffic road network features and traffic problems existing in the traffic road network as entities in the traffic knowledge graph;

[0006] Construct a traffic knowledge graph using the predefined entities and the diagnosed cases;

[0007] Input the traffic knowledge graph into an encoder based on a graph neural network to obtain the graph embedding vectors of each node in the traffic knowledge graph; and

[0008] Match the features of the intersection to be diagnosed with the predefined entities, and input the matched entities into a decoder based on entity fusion for the decoder to perform operations and output a diagnosis result.

[0009] Furthermore, 106 traffic road network features in 6 categories are predefined as F-type entities, and 45 traffic problems in 4 categories with 23 problem subcategories existing in the road network are predefined as P-type entities, among which 23 problem subcategories are C-type entities.

[0010] Further, the construction of the traffic knowledge graph using predefined entities and diagnosed cases includes:

[0011] Match the diagnosed cases with the predefined entities, and each diagnosed case can be regarded as a set F0 of traffic road network feature entities and a set P0 of traffic problem entities;

[0012] Perform Cartesian product operations between and within the sets of traffic road network feature entities F0 and traffic problem entities P0 for each diagnosed case to obtain triples, and integrate the triples in an accumulative manner to construct the traffic knowledge graph, where each triple relationship is regarded as an edge of the traffic knowledge graph, and the number of times the triple appears repeatedly is used as the edge weight; and

[0013] Filter the noise existing in the traffic knowledge graph through edge screening.

[0014] Further, the construction of the traffic knowledge graph using predefined entities and diagnosed cases further includes: adding an edge between C-type entities and P-type entities to obtain the traffic knowledge graph G=(V, E), where P-type entities are traffic problems existing in the predefined road network, and C-type entities are predefined problem subclasses.

[0015] Further, the filtering of the noise existing in the traffic knowledge graph through edge screening includes: combining the total edge weights of the two end nodes of each edge, calculating the normalized edge weight of each edge, and screening the normalized edge weight according to a fixed threshold, and deleting the edges with weights lower than the fixed threshold.

[0016] Further, inputting the traffic knowledge graph into an encoder based on a graph neural network to obtain the graph embedding vector of each node in the traffic knowledge graph includes: the encoder is composed of multiple layers of graph neural networks. In each layer of the graph neural network, the information propagation increment of each neighbor node to the target node is obtained through a message propagation mechanism, and the graph embedding vector of the target node is updated, and finally the updated graph embedding vector of the target node is output.

[0017] Further, the graph embedding vector of the starting node is obtained by random initialization, and then the graph embedding vector of the target node is updated, where the update process of the graph embedding vector of each target node includes:

[0018] Calculate the correlation weight between the neighbor node and the target node through an attention mechanism;

[0019] Normalize the weights of neighbor nodes of the same type;

[0020] Obtain the information propagation increment passed from the neighbor node to the target node through the normalized weight of the neighbor node; and

[0021] Integrate the information propagation increments of all neighbor nodes and update the graph embedding vector of the target node.

[0022] Further, match the features of the intersection to be diagnosed with predefined entities, and then input the matched entities into a decoder based on entity fusion. The decoder performs operations and outputs diagnostic results including:

[0023] Match the features of the intersection to be diagnosed with predefined traffic road network feature entities, and then input the matched traffic road network feature entities into the decoder;

[0024] The decoder mines the joint information contained in the traffic road network features simultaneously included in the intersection to be diagnosed, obtains a vector representing the case, and uses the inner product of this vector with the vectors of traffic problem entities one by one to obtain the correlation between the case and the traffic problems. Screen and rank the traffic problems according to the correlation threshold to obtain the traffic problems of the intersection to be diagnosed.

[0025] Further, the operation process of the decoder includes:

[0026] Add the graph embedding vectors corresponding to the traffic road network feature entities matched with the intersection to be diagnosed to obtain vector a;

[0027] After concatenating the graph embedding vector of each traffic road network feature entity with vector a, input it into a fully connected neural network to obtain the corresponding weight of the graph embedding vector of each traffic road network feature entity;

[0028] Perform weighted summation of the vectors of each traffic road network feature entity and their corresponding weights to obtain a vector representing the case;

[0029] Perform inner product of the vector representing the case with the vectors representing each traffic problem one by one to obtain the correlation between the case and each traffic problem; and

[0030] Through screening and ranking by a preset correlation threshold, a specified number of traffic problems with the highest correlation are used as the diagnostic results of the traffic problem diagnosis model.

[0031] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it executes the steps according to the above method.

[0032] The present invention has at least the following beneficial effects: (1) A problem diagnosis method for urban traffic networks disclosed by the present invention first predefines traffic road network features and problems existing in the traffic road network as entities in a traffic knowledge graph, constructs the traffic knowledge graph by using the predefined entities and diagnosed cases, so that the knowledge in historical case data has generalizability and transferability, and then uses an encoder based on a graph neural network to obtain graph embedding vectors of each node in the traffic knowledge graph. (2) The features of the intersection to be diagnosed are matched with the predefined traffic road network feature entities in a rule-based manner, and then the matched traffic road network feature entities are input into a decoder. The decoder mines the joint information contained in the traffic road network features simultaneously included in the intersection to be diagnosed, obtains a vector representing the case, and uses the vector to perform an inner product with the vector of the traffic problem entity one by one to obtain the correlation between the symbol case and the traffic problem. Screening and sorting according to the correlation threshold, the traffic problems of the intersection to be diagnosed are obtained. By mining the joint information of the traffic road network feature entities that appear simultaneously in the intersection to be diagnosed through a decoder based on entity fusion, a more accurate diagnosis result can be obtained compared with the method of performing inner product for each entity one by one. (3) This method can quickly diagnose traffic problems without manual intervention, has high diagnostic efficiency, and the accuracy and objectivity of the diagnostic results are relatively high. (4) This method has generalizability and transferability and can be extended to different cities. In summary, by applying novel artificial intelligence technologies to the field of intelligent transportation, the present invention greatly improves the diagnostic speed and accuracy of traffic problems, and at the same time reduces the amount of calculation compared with the amount of calculation required for the same accuracy, significantly reducing the hardware occupancy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] To further clarify the above and other advantages and features of the embodiments of the present invention, a more specific description of the embodiments of the present invention will be presented with reference to the accompanying drawings. It can be understood that these drawings only depict typical embodiments of the present invention and thus will not be considered as limiting its scope.

[0034] Figure 1 Shows the flow of a problem diagnosis method for urban traffic networks according to an embodiment of the present invention;

[0035] Figure 2 Shows the structural diagram of a traffic problem diagnosis model according to an embodiment of the present invention;

[0036] Figure 3 Shows a schematic diagram of entities for a traffic knowledge graph according to an embodiment of the present invention; and

[0037] Figure 4 Shows a schematic diagram of the construction of a traffic knowledge graph according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] It should be noted that the components in the respective drawings may be exaggerated for illustrative purposes and not necessarily drawn to scale.

[0039] In the present invention, the various embodiments are merely intended to illustrate the solutions of the present invention and should not be construed as restrictive.

[0040] In the present invention, unless otherwise specified, the quantifiers "a" and "one" do not exclude the scenario of multiple elements.

[0041] It should also be noted here that in the embodiments of the present invention, for clarity and simplicity, only a part of the components or assemblies may be shown. However, those of ordinary skill in the art can understand that, under the teaching of the present invention, the required components or assemblies can be added according to the specific scenario requirements.

[0042] It should also be noted here that within the scope of the present invention, the terms "same", "equal", "equivalent", etc. do not mean that the two values are absolutely equal, but allow for a certain reasonable error. That is to say, these terms also cover "substantially the same", "substantially equal", and "substantially equivalent".

[0043] It should also be noted here that in the description of the present invention, the orientation or positional relationships indicated by the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings. This is only for the convenience of describing the present invention and simplifying the description, and does not explicitly or implicitly imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and should not be construed as explicitly or implicitly indicating relative importance.

[0044] In addition, the numbering of the steps of the various methods of the present invention does not limit the execution order of the method steps. Unless otherwise specified, the method steps can be executed in different orders.

[0045] The present invention provides a problem diagnosis method for an urban traffic network. First, the traffic network features and problems existing in the traffic network are predefined in combination with expert knowledge in the traffic field as entities in a traffic knowledge graph. The predefined entities are used to match the features and problems in the diagnosed cases, knowledge triples are extracted and integrated by accumulation, each triple relationship is regarded as an edge of the traffic knowledge graph, and the noise existing in the graph is further filtered using an edge filtering strategy to obtain a traffic knowledge graph. And by using an encoder based on a graph neural network, a graph embedding vector of each node in the traffic knowledge graph is obtained. The traffic network features of the intersection to be diagnosed are matched with the predefined traffic network feature entities in a rule-based manner, and then the matched traffic network feature entities are input into a decoder. The decoder mines the joint information contained in the traffic network features simultaneously contained in the intersection to be diagnosed, obtains the vector of the symbolic case, and uses the inner product of the vector and the vector of the traffic problem entity to obtain the correlation between the symbolic case and the traffic problem, and screens and sorts according to the correlation threshold to obtain the traffic problem at the intersection to be diagnosed.

[0046] Figure 1 The present invention shows a process of a method for diagnosing problems in an urban transportation network according to an embodiment of the present invention; Figure 2 A structural diagram of a traffic problem diagnosis model according to an embodiment of the present invention is shown.

[0047] like Figure 1 As shown, a problem diagnosis method for urban traffic networks includes two parts: constructing a traffic knowledge graph and using a traffic problem diagnosis model to predict traffic problems.

[0048] The construction of traffic knowledge graph mainly includes entity pre-definition, entity matching, triple knowledge extraction and graph edge screening. The traffic knowledge graph is constructed using pre-defined entities and diagnosed cases.

[0049] Entity pre-definition: pre-define the characteristics of the traffic network and the traffic problems existing in the traffic network as entities in the traffic knowledge graph. Pre-define 6 categories of 106 traffic network characteristics as F-type entities (traffic network characteristic entities) and pre-define 4 categories of 23 problem subcategories and 45 traffic problems existing in the road network as P-type entities (traffic problem entities), of which 23 problem subcategories are C-type entities.

[0050] Entity matching: The diagnosed cases are matched with predefined entities through a rule-based method, so each diagnosed case can be regarded as a traffic network feature entity set F0 and a traffic problem entity set P0.

[0051] The following is an example of matching a diagnosed case with predefined entities through a rule-based method. The diagnosed case includes features and problems, and the features and problems of the diagnosed case are respectively matched with the traffic road network feature entities and traffic problem entities. For example, among the large / small intersection static feature categories in the predefined entities, there are two types: large and small. Here, large and small are determined based on rules, and the intersection scale of the diagnosed case is matched with the large / small intersection scale in the traffic road network feature entities.

[0052] Triple knowledge extraction: For the set F0 of traffic road network feature entities and the set P0 of traffic problem entities of each diagnosed case, perform the Cartesian product operation between sets and within sets respectively to obtain four types of triples, and integrate the triples in an accumulative manner to construct a traffic knowledge graph. Each triple relationship is regarded as an edge of the traffic knowledge graph, and the number of times the triple appears repeatedly is used as the edge weight. The triples include feature - causal relationship - problem, problem - representational relationship - feature, feature - co-occurrence relationship - feature, and problem - co-occurrence relationship - problem.

[0053] Graph edge screening: Filter the noise existing in the traffic knowledge graph through edge screening. Specifically, combine the total edge weights of the nodes at both ends of each edge to calculate the normalized edge weight of each edge, and screen the normalized edge weight according to a fixed threshold, deleting the edges with weights lower than the fixed threshold. In an embodiment of the present invention, the fixed threshold is set to 0.7.

[0054] Finally, add an edge between the C-type entity and the P-type entity to obtain the traffic knowledge graph G=(V, E). The nodes in the traffic knowledge graph refer to entities.

[0055] As Figure 1 shown, predicting traffic problems using the traffic problem diagnosis model includes: inputting the traffic knowledge graph into an encoder based on a graph neural network to obtain the graph embedding vector of each node in the traffic knowledge graph; and matching the traffic road network features of the intersection to be diagnosed with predefined entities, and then inputting the matched entities into the decoder. The decoder performs operations and outputs the diagnosis result. As Figure 2 shown, the traffic problem diagnosis model includes an encoder part based on a graph neural network and a decoder part based on entity fusion.

[0056] Input the traffic knowledge graph G=(V, E) into an encoder based on a graph neural network, and the encoder outputs the graph embedding vector of each node V in the graph.

[0057] The encoder part consists of multiple layers of graph neural networks, for example, two layers. In each layer of the graph neural network, through the message propagation mechanism, the information propagation increment of each neighbor node to the target node is obtained, and the graph embedding vector of the target node is updated. Finally, the updated graph embedding vector of the target node is output. The graph embedding vector of the node at the start is obtained by random initialization, and then the graph embedding vector of the target node is updated. The update process of the graph embedding vector of each target node can be divided into the following 4 steps:

[0058] 1. Calculate the correlation weight between the neighbor node and the target node through the attention mechanism;

[0059] 2. Normalize the weights of neighbor nodes of the same type;

[0060] 3. Obtain the information propagation increment from the neighbor node to the target node through the normalized weight of the neighbor node;

[0061] 4. Integrate the information propagation increments of all neighbor nodes and update the graph embedding vector of the target node.

[0062] The features of the intersection to be diagnosed are matched with the predefined traffic road network feature entities in a rule-based manner, and then the matched traffic road network feature entities are input into the decoder. The decoder mines the joint information contained in the traffic road network features simultaneously included in the intersection to be diagnosed, obtains the vector of the symbolic case, and uses the inner product of this vector and the vector of the traffic problem entity one by one to obtain the correlation between the symbolic case and the traffic problem. The traffic problems of the intersection to be diagnosed are screened and sorted according to the correlation threshold. The decoder consists of a fully connected neural network.

[0063] The following is an example to illustrate the matching of the features of the intersection to be diagnosed with the predefined traffic road network feature entities in a rule-based manner. For example, in the large / small intersection scale in the large category of static features of intersections / road segments in the predefined entities, there are two types: large and small. Here, large and small are determined based on rules, and the intersection scale of the intersection to be diagnosed is matched with the large intersection scale or small intersection scale in the predefined traffic road network feature entities.

[0064] The operation process of the decoder part can be divided into the following 5 steps:

[0065] 1. Add the graph embedding vectors corresponding to the traffic road network feature entities matched with the intersection to be diagnosed to obtain vector a;

[0066] 2. After concatenating the graph embedding vector of each traffic road network feature entity with vector a, input it into the fully connected neural network to obtain the corresponding weight of the graph embedding vector of each traffic road network feature entity;

[0067] 3. Weight-sum the vector of each traffic road network feature entity with the corresponding weight to obtain the vector of the symbolic case;

[0068] 4. Take the inner product of the vector of the symbolic case and the vector of each traffic problem one by one to obtain the correlation between the symbolic case and each traffic problem;

[0069] 5. Through the screening and sorting of the preset correlation threshold, the specified number of traffic problems with the highest correlation are used as the diagnosis results of the traffic problem diagnosis model.

[0070] Figure 3 The figure shows a schematic diagram of the entities for a traffic knowledge graph according to an embodiment of the present invention.

[0071] As Figure 3 shown, combining traffic domain knowledge, 6 major categories and a total of 106 types of traffic road network features (F type entities) and 4 major categories and 23 subcategories (C type entities) of 45 types of traffic problems (P type entities) existing in the road network are designed and proposed for the first time, regarded as the entities of the traffic knowledge graph. The graph contains three types of entities: F, C, and P.

[0072] The 6 major categories of traffic road network features include intersection / section static features, feature engineering features, real-time traffic parameter features, surrounding environment features, arterial coordination features, and intersection signal control schemes. The intersection / section static features include large / small intersection scale, whether it is a conventional intersection, whether there is a waiting area, etc. The feature engineering features include no / partial / all import lanes being congested, lane imbalance coefficient, green light utilization coefficient, etc. The real-time traffic parameter features include no / partial / all import lanes having a large flow, no / partial / all import lanes having a long queue, no / partial / all import lanes having a large number of stops, etc. The surrounding environment features include POI - school, POI - business district, etc. The arterial coordination features include a large number of stops on the main road, a large waste of green lights on the arterial, arterial coordination effect coefficient, etc. The intersection signal control schemes include using symmetric release, using separate release of mixed lanes, overlapping phases, etc. Only some features are listed here as examples.

[0073] Traffic problems include traffic organization problems, background environment problems, single-point signal control scheme problems, and arterial signal control scheme problems, which belong to type C entities. Traffic organization problems include mismatches in import and export functions, lane numbers at import and export, lane mixing at import and export, mismatches in guide lanes, insufficient intersection space allocation, unreasonable bus stop positions, and unreasonable road right planning. Background environment problems include emergencies - traffic accidents, temporary events - serious illegal parking, temporary events - interference from large-scale events, and temporary events - school pick-up and drop-off occupying roads. Single-point signal control scheme problems include unreasonable signal timing - too long cycle, unreasonable signal timing - too short cycle, unreasonable signal timing - phase imbalance, insufficient attention to chronic traffic - pedestrian waiting, insufficient attention to chronic traffic - conflict between motor vehicles and non-motor vehicles, unreasonable signal dispatching - inaccurate division, and unreasonable signal dispatching - no emergency plan. Arterial signal control scheme problems include no arterial coordination, spillover caused by short connections, serious oversaturation of traffic flow, and weak arterial coordination due to large distances between intersections. Only some problems are listed here as examples.

[0074] Figure 4 The schematic diagram of constructing a traffic knowledge graph according to an embodiment of the present invention is shown.

[0075] As Figure 4 shown, a traffic knowledge graph is constructed by using type F entities, type C entities, and type P entities of a certain road network. There is a causal relationship between type F entities and type P entities, a co-occurrence relationship between type F entities and type F entities, and a belonging relationship between type P entities and type C entities. In this embodiment, type F entities include 5 categories: intersection static features, feature engineering features, real-time traffic parameter features, surrounding environment features, and intersection signal control schemes. Type C entities include background environment problems and single-point signal control scheme problems.

[0076] In addition, each embodiment can be provided as a computer program product that may include one or more machine-readable media storing machine-executable instructions. When these instructions are executed by one or more machines such as a computer, a computer network, or other electronic devices, they can cause the one or more machines to perform the operations according to the embodiments of the present invention. Machine-readable media may include, but are not limited to, floppy disks, optical disks, CD-ROMs (compact disc read-only memories), and magneto-optical disks, ROMs (read-only memories), RAMs (random access memories), EPROMs (erasable programmable read-only memories), EEPROMs (electrically erasable programmable read-only memories), magnetic or optical cards, flash memories, or other types of media / machine-readable media suitable for storing machine-executable instructions.

[0077] In addition, each embodiment can be downloaded as a computer program product, wherein the program can be transmitted from a remote computer (e.g., a server) to a requesting computer (e.g., a client) by one or more data signals embodied in and / or modulated by a carrier wave or other propagation medium via a communication link (e.g., a modem and / or a network connection). Thus, the machine-readable medium used herein can include such a carrier wave, but this is not required.

[0078] Although some embodiments of the present invention have been described in this application document, those skilled in the art can understand that these embodiments are merely shown as examples. Those skilled in the art can conceive of numerous variations, alternatives, and improvements under the teachings of the present invention without departing from the scope of the present invention. The appended claims are intended to define the scope of the present invention and thereby cover methods and structures within the scope of these claims themselves and their equivalents.

Claims

1. A problem diagnosis method for urban traffic networks, characterized in that Including: Predefined traffic road network features and traffic problems existing in the traffic road network as entities in the traffic knowledge graph; Constructing a traffic knowledge graph by using the predefined entities and diagnosed cases; Inputting the traffic knowledge graph into an encoder based on a graph neural network to obtain the graph embedding vectors of each node in the traffic knowledge graph; And Matching the features of the intersection to be diagnosed with the predefined entities, and inputting the matched entities into a decoder based on entity fusion for the decoder to perform operations and output a diagnosis result; The step of inputting the traffic knowledge graph into an encoder based on a graph neural network to obtain the graph embedding vectors of each node in the traffic knowledge graph includes: The encoder is composed of multiple layers of graph neural networks. In each layer of the graph neural network, the information propagation increment of each neighbor node to the target node is obtained through a message propagation mechanism, and the graph embedding vector of the target node is updated, and finally the updated graph embedding vector of the target node is output; The graph embedding vector of the node at the start is obtained by random initialization, and then the graph embedding vector of the target node is updated, where the update process of the graph embedding vector of each target node includes: Calculating the correlation weight between the neighbor node and the target node through an attention mechanism; Normalizing the weights of neighbor nodes of the same type; Obtaining the information propagation increment passed from the neighbor node to the target node through the normalized weight of the neighbor node; and Integrating the information propagation increments of all neighbor nodes and updating the graph embedding vector of the target node; Matching the features of the intersection to be diagnosed with the predefined entities, and then inputting the matched entities into a decoder based on entity fusion. The decoder performs operations and outputs a diagnosis result includes: Matching the features of the intersection to be diagnosed with the predefined traffic road network feature entities, and then inputting the matched traffic road network feature entities into the decoder; The decoder mines the joint information contained in the traffic road network features simultaneously included in the intersection to be diagnosed to obtain a vector representing the case, and uses the vector to perform an inner product with the vector of each traffic problem entity to obtain the correlation between the case and the traffic problem, and filters and sorts the traffic problems according to a correlation threshold to obtain the traffic problems of the intersection to be diagnosed; The operation process of the decoder includes: Adding the graph embedding vectors corresponding to the traffic road network feature entities matched with the intersection to be diagnosed to obtain vector a; After concatenating the graph embedding vector of each traffic road network feature entity with vector a, inputting them into a fully connected neural network to obtain the corresponding weight of the graph embedding vector of each traffic road network feature entity; Performing a weighted sum of the vector of each traffic road network feature entity and its corresponding weight to obtain a vector representing the case; Performing an inner product of the vector representing the case with the vector representing each traffic problem one by one to obtain the correlation between the case and each traffic problem; and Through preset correlation threshold screening and correlation ranking, the specified number of traffic problems with the highest correlation are used as the diagnosis result of the traffic problem diagnosis model.

2. The problem diagnosis method for an urban traffic network according to claim 1, characterized in that, A total of 106 traffic road network features in 6 major categories are predefined as F-type entities, and a total of 45 traffic problems in 4 major categories and 23 problem subcategories existing in the road network are predefined as P-type entities, where 23 problem subcategories are C-type entities.

3. The problem diagnosis method for an urban traffic network according to claim 1, characterized in that, The construction of a traffic knowledge graph using the predefined entities and diagnosed cases includes: Matching the diagnosed cases with the predefined entities, then each diagnosed case can be regarded as a traffic road network feature entity set F0 and a traffic problem entity set P0; Performing Cartesian product operations between and within the sets on the traffic road network feature entity set F0 and the traffic problem entity set P0 of each diagnosed case to obtain triples, and integrating the triples in an accumulative manner to construct a traffic knowledge graph, where each triple relationship is regarded as an edge of the traffic knowledge graph, and the number of times the triple appears repeatedly is used as the edge weight; and Filtering the noise existing in the traffic knowledge graph through an edge screening method.

4. The problem diagnosis method for an urban traffic network according to claim 3, wherein The construction of a traffic knowledge graph using the predefined entities and diagnosed cases further includes: adding an edge between the C-type entity and the P-type entity to obtain a traffic knowledge graph G=(V,E), where the P-type entity is the traffic problem existing in the predefined road network, and the C-type entity is the predefined problem subcategory.

5. The problem diagnosis method for an urban traffic network according to claim 3, characterized in that, The filtering of the noise existing in the traffic knowledge graph through the edge screening method includes: combining the total edge weights of the nodes at both ends of each edge, calculating the normalized edge weight of each edge, and screening the normalized edge weight according to a fixed threshold, and deleting the edges with weights lower than the fixed threshold.

6. A computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, executes the steps of the method according to any one of claims 1-5.

Images