CN117592006B - Smart city data processing method, device, equipment and readable storage medium - Google Patents
Smart city data processing method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN117592006B CN117592006B CN202410078615.XA CN202410078615A CN117592006B CN 117592006 B CN117592006 B CN 117592006B CN 202410078615 A CN202410078615 A CN 202410078615A CN 117592006 B CN117592006 B CN 117592006B
- Authority
- CN
- China
- Prior art keywords
- smart city
- node
- nodes
- map
- heterogeneous
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 69
- 238000003860 storage Methods 0.000 title claims abstract description 44
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 33
- 238000004364 calculation method Methods 0.000 claims description 101
- 238000012545 processing Methods 0.000 claims description 87
- 238000000034 method Methods 0.000 claims description 72
- 238000010586 diagram Methods 0.000 claims description 46
- 238000004422 calculation algorithm Methods 0.000 claims description 43
- 230000008569 process Effects 0.000 claims description 30
- 239000011159 matrix material Substances 0.000 claims description 29
- 238000012544 monitoring process Methods 0.000 claims description 28
- 238000005096 rolling process Methods 0.000 claims description 25
- 238000012549 training Methods 0.000 claims description 19
- 230000006870 function Effects 0.000 claims description 17
- 238000005065 mining Methods 0.000 claims description 17
- 238000010276 construction Methods 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 14
- 238000013507 mapping Methods 0.000 claims description 12
- 230000004927 fusion Effects 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 8
- 238000005295 random walk Methods 0.000 claims description 7
- 230000003068 static effect Effects 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 description 18
- 238000005516 engineering process Methods 0.000 description 9
- 239000000284 extract Substances 0.000 description 7
- 238000007405 data analysis Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 102100036962 5'-3' exoribonuclease 1 Human genes 0.000 description 2
- 101000804879 Homo sapiens 5'-3' exoribonuclease 1 Proteins 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000009877 rendering Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000004308 accommodation Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000003930 cognitive ability Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000004382 visual function Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/26—Discovering frequent patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Tourism & Hospitality (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to the technical field of artificial intelligence, and particularly discloses a smart city data processing method, device, equipment and readable storage medium.
Description
Technical Field
The present invention relates to the field of artificial intelligence technology, and in particular, to a smart city data processing method, device, apparatus and readable storage medium.
Background
Smart cities, smart farms, smart transportation, etc. utilize Information and Communication Technology (ICT) and other intelligent technologies to plan cross-scene management schemes. The application of artificial intelligence techniques in smart cities is also related to aspects across scenarios. However, the current smart city still stays in the informatization stage, that is, information is visually represented by a mode of collecting cross-domain data and establishing an information base on a large scale, but the information of each scene is an information island, and the intelligent processing scheme of the cross-scene cannot be realized due to lack of deep association. The existing artificial intelligent models such as a target detection model, a human body posture estimation model and the like can only play a role in a single scene, have weak cognitive ability and cannot meet the construction requirements of an intelligent platform.
The intelligent platform data processing scheme suitable for cross-scene is provided, and is a technical problem to be solved by a person skilled in the art.
Disclosure of Invention
The invention aims to provide a smart city data processing method, device and equipment and a readable storage medium, which are used for realizing a cross-scene smart city data processing scheme.
In order to solve the above technical problems, the present invention provides a smart city data processing method, including:
collecting data from information sources in multiple fields corresponding to the smart city and storing the data into a smart city storage system;
Extracting data of the information sources in each field from the smart city storage system, and constructing a smart city heterogeneous graph taking information source entities as nodes and the association relationship among the information source entities as edges;
constructing a smart city map calculation model based on the smart city heterogeneous map, and extracting features in the smart city heterogeneous map to be converted into a low-dimensional embedded representation and then embedding the low-dimensional embedded representation into the smart city map calculation model in the construction process;
and deploying the application program interface of the smart city map calculation model to user equipment to receive the smart city data processing task of the user equipment and processing the smart city data based on the smart city map calculation model.
In some implementations, the building a smart city heterogeneous graph with information source entities as nodes and association relationships between the information source entities as edges includes:
And constructing the smart city heterogeneous graph by taking the ontology of the information source entity as an entity node, the category of the information source entity as a virtual node, and the association relationship between different entity nodes and the association relationship between the entity node and the virtual node as the edges.
In some implementations, the building a smart city heterogeneous graph with information source entities as nodes and association relationships between the information source entities as edges includes:
Taking two nodes as a node pair, and excavating association relations between the two nodes in each node pair by utilizing at least one of a network document, map software and an artificial intelligence large model;
and constructing the smart city heterogeneous graph by utilizing the information of the nodes and the association relation between the nodes.
In some implementations, mining the association between two of the nodes in each of the node pairs using the network document includes:
And extracting association relations between two nodes in each node pair from the network document by adopting a multi-strategy Chinese open relation extraction algorithm.
In some implementations, mining, with the map software, an association between two of the nodes in each of the node pairs includes:
searching for a geographic distance between two of the nodes in each of the node pairs from the map software, respectively;
Comparing the geographic distance with a neighbor node distance threshold, and if the geographic distance is greater than or equal to the neighbor node distance threshold, determining that no geographic neighbor relation exists between the two nodes; and if the geographic distance is smaller than the distance threshold value of the neighbor nodes, determining that the geographic neighbor relation exists between the two nodes.
In some implementations, mining the association between two of the nodes in each of the node pairs using the artificial intelligence large model includes:
generating corresponding association relation identification tasks according to the information of the two nodes in the node pair;
inputting the association relation identification task into the artificial intelligent large model, and outputting the association relation of two nodes in the node pair.
In some implementations, the mining the association between two nodes in each node pair by using at least one of a network document, a map software and an artificial intelligence big model with the two nodes as one node pair includes:
Initializing a node interface, a network document interface, a map software interface and an artificial intelligence large model interface;
generating a first association list, a second association list and a third association list;
receiving information of each node pair from the node interface;
extracting first association relations between two nodes in each node pair from the network document according to the information of the node pairs, and writing the first association relations into the first association relation list;
searching for a geographic distance between two of the nodes in each of the node pairs from the map software, respectively;
Comparing the geographic distance with a neighbor node distance threshold, and if the geographic distance is greater than or equal to the neighbor node distance threshold, determining that no geographic neighbor relation exists between the two nodes; if the geographic distance is smaller than the distance threshold value of the neighbor node, determining that the geographic neighbor relation exists between the two nodes;
Writing information of two nodes with the geographic neighbor relation into the second association relation list;
generating corresponding association relation identification tasks according to the information of the two nodes in the node pair;
Inputting the association relation identification task into the artificial intelligent large model, and outputting a third association relation of two nodes in the node pair;
writing the third association into the third association list;
and outputting the updated first association list, the updated second association list and the updated third association list.
In some implementations, the building a smart city heterogeneous graph with information source entities as nodes and association relationships between the information source entities as edges includes:
extracting the characteristics of the information source entity as the characteristics of the nodes by using a pre-trained first language model, and constructing the nodes and the edges into the smart city heterogeneous diagram;
the characteristics of the node comprise a node name, static information of the node and dynamic information of the node.
In some implementations, the building a smart city map computation model based on the smart city heterogeneous map, extracting features in the smart city heterogeneous map for conversion into a low-dimensional embedded representation during the building, and embedding the smart city map computation model includes:
extracting semantic features and/or structural features from the smart city heterogeneous map, embedding the semantic features and/or the structural features into a map convolution network to learn the features of the smart city heterogeneous map, and generating the smart city map calculation model.
In some implementations, the extracting semantic features and/or structural features from the smart city heterogeneous map, embedding the semantic features and/or the structural features into a map convolution network to learn features of the smart city heterogeneous map, generating the smart city map computation model includes:
Acquiring the characteristics of the node;
Extracting the semantic features and the structural features from the smart city heterogeneous map;
fusing the characteristics of the nodes, the semantic characteristics and the structural characteristics to obtain fused characteristics;
and embedding the fusion features into the graph rolling network to learn the features of the smart city heterogeneous graph, and generating the smart city graph calculation model.
In some implementations, extracting the semantic features from the smart city heterogeneous map includes:
Based on the smart city heterogeneous diagram, obtaining triple data consisting of a first node h, a second node t and an association relation r between the first node and the second node;
employing a knowledge-graph embedding algorithm to satisfy the vector representation For optimization purposes, learning the semantic features of the smart city heterogeneous map.
In some implementations, the employing a knowledge-graph embedding algorithm such that the vector representation satisfiesFor optimization purposes, learning the semantic features of the smart city heterogeneous map, comprising:
Representing the triplet data as a random initialization vector;
Sampling the random initialization vector to obtain a negative sample;
To be used for Learning the negative sample for an objective function and a semantic loss function to obtain the semantic features of the smart city heterogeneous diagram;
Wherein the semantic loss function is:
；
wherein, Is an interval parameter,/>To take positive number calculate,/>Is a negative sample set,/>For the sampled header entity,/>The tail entity obtained for sampling.
In some implementations, extracting the structural features from the smart city heterogeneous map includes:
And aiming at converting the nodes in the smart city heterogeneous map into embedded representations with low dimension and density and enabling similar nodes in the smart city heterogeneous map to be similar in low dimension space distance, learning the map structure embedded representations of the smart city heterogeneous map by adopting a map embedding algorithm, and obtaining the structural characteristics in the smart city heterogeneous map.
In some implementations, the learning the graph structure embedded representation of the smart city heterogeneous graph using a graph embedding algorithm with the goal of converting the nodes in the smart city heterogeneous graph into a low-dimensional dense embedded representation, such that similar nodes in the smart city heterogeneous graph are close in low-dimensional spatial distance, to obtain the structural features in the smart city heterogeneous graph includes:
the method comprises the steps of taking the network neighborhood which can be observed with maximum probability based on the embedding of a central node as a target, and adopting a bias random walk model to learn the relation structure of each node;
and calculating the structural characteristics in the smart city heterogeneous graph according to the vector representation of each node and the vector representation of the relation structure of each node.
In some implementations, the fusing processing is performed on the features of the nodes, the semantic features and the structural features to obtain fused features, and the fused features are obtained through the following formula:
；
wherein, For the fusion feature,/>For the node characteristics of the ith node,/>For the structural feature of the ith node,/>For the semantic features of the ith node,/>For matrix connection operations.
In some implementations, the training method of the smart city map calculation model includes:
acquiring an initial graph rolling network;
Setting initial parameters for the initial graph rolling network;
Training the graph convolution network according to the semantic features and/or the structural features and the features of the nodes to obtain a prediction category of the nodes;
And optimizing model parameters of the graph convolution network by using the actual category of the node and the loss of the predicted category of the node until the category prediction target of the node is met, so as to obtain the smart city graph calculation model.
In some implementations, the class prediction target is represented by:
；
wherein, Is a negative log likelihood loss,/>For model parameters of the graph convolution network,/>For the model parameter/>The corresponding negative log likelihood loss is the smallest,/>W is a parameter matrix for the number of nodes,/>Is the transposed matrix of the parameter matrix W,/>For the hidden state representation corresponding to the ith node,And (5) the probability of each category corresponding to the ith node.
In some implementations, the receiving smart city data processing tasks of the user device and performing smart city data processing based on the smart city map computing model includes:
Receiving a smart city monitoring task sent by the user equipment;
identifying node attributes and monitoring items in the smart city monitoring task;
Invoking the smart city diagram calculation model to perform node classification and/or node clustering processing on the nodes belonging to the node attribute in the smart city heterogeneous diagram according to the monitoring item to obtain a node classification result and/or a node clustering result;
And outputting the node classification result and/or the node clustering result.
In some implementations, the receiving smart city data processing tasks of the user device and performing smart city data processing based on the smart city map computing model includes:
receiving a to-be-solved problem sent by the user equipment;
Invoking the smart city diagram calculation model to search information of related nodes of the to-be-solved problem and neighborhood information of the related nodes;
Invoking a pre-trained second language model to generate answer information according to the information of the related nodes and the neighborhood information of the related nodes;
And outputting the answer information.
In order to solve the above technical problems, the present invention further provides a smart city data processing device, including:
The acquisition unit is used for acquiring data from information sources in multiple fields corresponding to the smart city and storing the data into the smart city storage system;
the mapping unit is used for extracting the data of the information sources in each field from the smart city storage system and constructing a smart city heterogeneous map which takes information source entities as nodes and the association relationship between the information source entities as edges;
the modeling unit is used for constructing a smart city map calculation model based on the smart city heterogeneous map, and extracting features in the smart city heterogeneous map to be converted into a low-dimensional embedded representation and then embedded into the smart city map calculation model in the construction process;
And the processing unit is used for deploying the application program interface of the smart city map calculation model to the user equipment so as to receive the smart city data processing task of the user equipment and perform smart city data processing based on the smart city map calculation model.
In order to solve the above technical problem, the present invention further provides a smart city data processing device, including:
a memory for storing a computer program;
a processor for executing the computer program, which when executed by the processor implements the steps of the smart city data processing method as described in any of the above.
To solve the above technical problem, the present invention also provides a readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the smart city data processing method as described in any one of the above.
According to the smart city data processing method provided by the invention, the informatization processing of the smart city information across the fields is realized by collecting data from the information sources of the multiple fields corresponding to the smart city and storing the data into the smart city storage system; the method comprises the steps of extracting data of information sources in various fields from a smart city storage system, constructing a smart city heterogeneous map taking information source entities as nodes and the association relation between the information source entities as edges, and constructing a smart city map calculation model in a mode of extracting features in the smart city heterogeneous map to be converted into low-dimensional embedded representation and then embedded into the smart city map calculation model, so that implicit association modeling across scenes is realized, and the tasks of global optimal regulation, timely discovery of problems, resource optimization scheduling and the like are supported; through deploying the application program interface of the smart city map calculation model to the user equipment, the smart city data processing task of the user equipment is received, and smart city data processing is performed based on the smart city map calculation model, so that the correlation calculation and analysis of smart cities under different scenes in the cross-domain are realized, and the traditional smart city is promoted to be converted from an informatization scheme to an intelligent scheme.
According to the smart city data processing method, when the smart city heterogeneous diagram is constructed, the entity of the information source entity is taken as the entity node, the category to which the information source entity belongs is taken as the virtual entity, the hierarchical concept body structure is constructed, and the entity node and the virtual node are added into the smart city heterogeneous diagram, so that the semantic understanding capability of the smart city heterogeneous diagram analysis is enhanced.
According to the smart city data processing method provided by the invention, the two nodes are taken as a node pair, and the association relation between the two nodes in each node pair is mined by utilizing information sources such as network documents, map software, artificial intelligence large models and the like, so that the association relation between each node and each node is utilized to construct a smart city heterogeneous graph, and the deep mining of the multidimensional association relation is realized.
According to the smart city data processing method, semantic features and structural features are extracted from the smart city heterogeneous map and embedded into the neural network for learning to conduct node classification training, so that the generalization capability of the graph rolling network on node representation is enhanced, the graph rolling network depth analysis can be conducted based on the constructed smart city heterogeneous map, and the smart city data processing method can be widely applied to various intelligent platform scenes.
The invention also provides a smart city data processing device, equipment and a readable storage medium, which have the beneficial effects and are not repeated here.
Drawings
For a clearer description of embodiments of the invention or of the prior art, the drawings that are used in the description of the embodiments or of the prior art will be briefly described, it being apparent that the drawings in the description below are only some embodiments of the invention, and that other drawings can be obtained from them without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of a smart city data processing system according to an embodiment of the present invention;
FIG. 2 is a flowchart of a smart city data processing method according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a modeling process of a smart city map calculation model according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a smart city data processing device according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a smart city data processing device according to an embodiment of the present invention.
Detailed Description
The core of the invention is to provide a smart city data processing method, a device, equipment and a readable storage medium, which are used for realizing a cross-scene smart city data processing scheme.
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The following describes an embodiment of the present invention.
FIG. 1 is a schematic diagram of a smart city data processing system according to an embodiment of the present invention.
For easy understanding, the smart city data processing system provided by the embodiment of the invention is first described.
As shown in fig. 1, an embodiment of the present invention provides a smart city data processing system for implementing the smart city data processing method provided in the embodiments of the present invention, where the smart city data processing system mainly includes a smart city storage system, a smart city computing system, and a smart city client device.
The smart city storage system is used for storing data acquired from information sources in various fields corresponding to a smart city, as shown in fig. 1, the data in different fields can be stored in different storage areas, corresponding storage strategies and backup strategies can be determined according to importance degrees of different data, and used storage equipment can be determined according to data updating frequency. In the process of constructing the smart city data processing system to the operation, the smart city data processing system can continuously receive updated data in various fields through the smart city storage system so as to ensure the data instantaneity of the smart city data processing system.
The smart city client device is user equipment provided with the smart city client, can be a mobile phone and a personal computer held by a personal user, and can also be a server of a data center of a group such as an enterprise, a unit and the like. By applying the smart city data processing method provided by the embodiment of the invention, an intelligent smart city platform is built, which can comprise sub-platforms such as smart traffic, smart travel, smart farms, smart public security, smart government affairs and the like, a plurality of different platform products can be derived, and the smart city data processing method is provided for a user application program interface to be deployed on user equipment as a user client for users to access the smart city platform, and analysis and processing of the smart city data are realized by using a smart city map calculation model.
The smart city computing system is used for constructing a smart city heterogeneous map and a smart city map computing model, and solving smart city data processing tasks, such as monitoring and early warning tasks, user question-answering services and the like, provided by the user equipment. The smart city computing system may include one or more computing devices that may employ, but are not limited to, a central processing unit (Central Processing Unit, CPU), a graphics processor (Graphics Processing Unit, GPU), a field programmable gate array (Field Programmable GATE ARRAY, FPGA), or the like as the computing module. The invention provides a light smart city implementation scheme which can be implemented without large-cost input of computing equipment.
On the basis of the above architecture, the smart city data processing method according to the embodiment of the present invention is described below with reference to the accompanying drawings.
The second embodiment of the present invention will be described below.
Fig. 2 is a flowchart of a smart city data processing method according to an embodiment of the present invention.
The information technology solution of the smart city generally adopts a cloud, side and end collaborative computing architecture, and the information technology solution covers a lot of contents including the Internet of things, artificial intelligence and mobile Internet development technology stacks. Most of traditional smart city solutions are in an informatization stage, have no good intelligence realization, and still face a lot of limitations:
(1) The cross-scene analysis is difficult, the information of each scene becomes an information island, the depth association is lacking, and the global regulation and control of each scene corresponding to the whole platform cannot be realized. Artificial intelligence can only play a role in a single scene, for example, a license plate recognition model and a travel scenic spot model are independently operated, but a road section with traffic jam is unfavorable for travel, and a traditional smart city scheme cannot model the relevance of information in various scenes.
(2) The universality is poor, the traditional smart city scheme only has an artificial intelligent model trained for a single scene, and each artificial intelligent model can only solve a single task, for example, a traffic flow prediction model can only predict traffic flow and cannot execute a target detection task. Because of the high cost of training a single model and poor versatility, models cannot be trained specifically for each requirement and each scene of a smart city.
(3) The degree of intellectualization is low. Traditional machine learning models have low success rates for solving problems in smart city solutions, for example, only provide simple target detection functions and face recognition functions, and the requirements for building smart cities are increasingly unsatisfied.
(4) The deployment and maintenance costs are high, the traditional smart city system is mostly the scheme of the artificial intelligence of the previous generation and the Internet of things (AIoT), the organization structure is complex, faults frequently occur, the adaptability to new scenes and new demands is poor, new functions cannot be flexibly added, and the system deployment and maintenance costs are high.
Aiming at the problems faced by the traditional smart city scheme, the embodiment of the invention provides a smart city data processing method which comprises a modeling method based on graph calculation and a data searching method based on a graph calculation model.
As shown in fig. 2, the smart city data processing method provided by the embodiment of the invention includes:
S201: data is collected from information sources in multiple fields corresponding to the smart city and stored in a smart city storage system.
S202: and extracting data of information sources in various fields from the smart city storage system, and constructing a smart city heterogeneous graph which takes information source entities as nodes and takes association relations among the information source entities as edges.
S203: and constructing a smart city map calculation model based on the smart city heterogeneous map, and extracting features in the smart city heterogeneous map to be converted into a low-dimensional embedded representation and then embedding the low-dimensional embedded representation into the smart city map calculation model in the construction process.
S204: an application program interface of the smart city map calculation model is deployed to the user equipment to receive a smart city data processing task of the user equipment and to perform smart city data processing based on the smart city map calculation model.
In some implementations of the embodiments of the present invention, for S201, an entity that can actively or passively send information to the outside is defined as an information source. For example, a building has information such as a building name, a geographic location, a functional attribute, a time of construction, etc., which are known from the outside, and then a building can be regarded as an information source entity.
For a smart city, categories of information sources may include: cities, buildings, institutions, vehicles, user terminals, etc.
For S202, the embodiment of the present invention provides a multi-source multi-scene information modeling method, by constructing multi-source multi-scene information as a heterogram g= (V, E), where V represents a node set and E represents an edge set. Each node represents an information source in each scene faced by the smart city, such as ancient building 1, ancient building 2, college 1 and college 2 in the city B can be used as nodes in the heterogeneous map of the smart city. Edges represent the connection relationships between nodes, e.g. (historic building 1, neighbor, historic building 2).
In order to expand the association relationship between information source entities, in the smart city data processing method provided by the embodiment of the invention, virtual nodes between the information source entities can be also constructed. The constructing a smart city heterogeneous map with information source entities as nodes and the association relationship between the information source entities as edges in S202 may include: and constructing a smart city heterogeneous graph by taking the ontology of the information source entity as an entity node, the category of the information source entity as a virtual node and the association relationship between different entity nodes and the association relationship between the entity node and the virtual node as edges. That is, assuming that there is no surface association between college 1 and college 2, since they all belong to colleges, a virtual node defined as "college" may be constructed, that is, a category of "college 1" may be obtained as a (college 1, category, college), and an association between (college 2, category, college) may be obtained as well, thereby establishing an association between college 1 and college 2.
By collecting the characteristics of the nodes and the characteristics of the association relationship between the nodes, a smart city heterogeneous map corresponding to the smart city can be constructed and formed.
The heterogeneous graph (Heterogeneous Graph) is a graph structure that includes multiple types of nodes and edges, and only one type of node and edge is in the homogeneous graph (Homogeneous Graph) corresponding thereto. The knowledge graph construction technology aims at relating entities and concepts through different types of relations by means of ontology modeling and instance layer modeling technologies, so that a semantic network is constructed. The semantic network is usually a heterogeneous diagram, the process of knowledge map construction is a heavy-weight engineering, the embodiment of the invention adopts a lightweight method to construct a smart city heterogeneous diagram, and the embodiment of the invention adopts the ontology modeling thought in the knowledge map, and adds virtual nodes, such as category concepts of 'colleges and universities', 'scenic spots', 'companies', 'parking lots', 'communities', 'subways', and the like, to the heterogeneous diagram by constructing a hierarchical concept ontology structure, so as to help enhance the semantic understanding capability of the heterogeneous diagram analysis, unlike most heterogeneous diagrams constructed according to real nodes.
For S203, to implement solving the problems related to the smart city based on the smart city heterogeneous diagram, the expression modeling of the smart city heterogeneous diagram is required. In the embodiment of the invention, the expression of the smart city heterogeneous map is realized by constructing the smart city map calculation model, and in the process of constructing the smart city map calculation model, the tasks of global optimal regulation and control, timely discovery of problems, resource optimization scheduling and the like are supported by extracting the characteristics in the smart city heterogeneous map, converting the characteristics into low-dimensional embedded representation and then embedding the low-dimensional embedded representation into the smart city map calculation model, so that the data of information sources in each field in the smart city are converted into data with different dimensions into the same dimension, and the implicit associated modeling of the cross-scene is realized.
For S204, an application program interface of the smart city map calculation model is created and deployed to the user device, and the user can access the smart city platform by downloading the smart city client in the user device to obtain the required smart city data processing service. By the smart city data processing method, the smart city data processing task input by the user through the user equipment is received, and downstream tasks such as classification, clustering, link prediction and the like of nodes in the smart city heterogeneous map can be executed based on the smart city map calculation model, so that the smart city is really converted from a traditional informatization stage to an intelligent platform.
The smart city platform provided by the embodiment of the invention is a comprehensive platform across fields, and can solve the problem of data analysis tasks of users on any cross fields. Under the smart city platform provided by the embodiment of the invention, the sub-platforms such as a smart traffic sub-platform, a smart travel sub-platform, a smart farm sub-platform, a smart public security sub-platform, a smart government sub-platform and the like can be further arranged, each sub-platform can correspond to one or more fields, and under the smart city platform, if the smart city data analysis service required by a user is one or a part of data analysis service of the fields, a smart city map calculation model is called to execute data analysis tasks of a single field or a cross field based on node information and neighborhood information corresponding to the sub-platform.
According to the smart city data processing method provided by the embodiment of the invention, the informatization processing of the smart city information across the fields is realized by collecting data from the information sources of the multiple fields corresponding to the smart city and storing the data into the smart city storage system; the method comprises the steps of extracting data of information sources in various fields from a smart city storage system, constructing a smart city heterogeneous map taking information source entities as nodes and the association relation between the information source entities as edges, and constructing a smart city map calculation model in a mode of extracting features in the smart city heterogeneous map to be converted into low-dimensional embedded representation and then embedded into the smart city map calculation model, so that implicit association modeling across scenes is realized, and the tasks of global optimal regulation, timely discovery of problems, resource optimization scheduling and the like are supported; through deploying the application program interface of the smart city map calculation model to the user equipment, the smart city data processing task of the user equipment is received, and smart city data processing is performed based on the smart city map calculation model, so that the correlation calculation and analysis of smart cities under different scenes in the cross-domain are realized, and the traditional smart city is promoted to be converted from an informatization scheme to an intelligent scheme.
The smart city data processing method provided by the embodiment of the invention realizes the multi-source and multi-scene modeling of the smart city through one smart city map calculation model, does not need to train an artificial intelligent model for each scene of each information source, obviously reduces the model training difficulty, namely reduces the computational power resources and storage resources required for building the smart city, supports a general central processor and graphic processor scheme in practical application, does not need to adjust server hardware of a data center, and greatly reduces the cost of model training and the period of new service development and maintenance. In addition, the smart city data processing method provided by the embodiment of the invention provides a low coupling scheme, can rapidly adapt and access new business data, and does not need manual marking of data.
The following describes a third embodiment of the present invention.
On the basis of the above embodiments, the embodiment of the present invention further describes a method for constructing a heterogeneous map of a smart city.
In the smart city data processing method provided by the embodiment of the present invention, in S202, using information source entities as nodes and using association relations between the information source entities as edges, a smart city heterogeneous graph is constructed, which may include:
Taking two nodes as a node pair, and utilizing at least one of a network document, map software and an artificial intelligence large model to mine the association relation between the two nodes in each node pair;
and constructing a smart city heterogeneous graph by utilizing the information of each node and the association relation between each node.
In some implementations of the embodiments of the present invention, for the association between nodes, on one hand, an association may be established at a conceptual level according to ontology modeling, and in addition, deep association between nodes may be mined by using information sources such as network documents, map software, and artificial intelligence large models.
The mining of the association relationship between two nodes in each node pair by using the network document may include: and extracting association relations between two nodes in each node pair from the network document by adopting a multi-strategy Chinese open relation extraction algorithm. The network document can be sourced from each Internet platform, information of two nodes is input into a search engine, the network document associated with the two nodes is obtained through searching, and then the association relationship between the two nodes is extracted from the network document by utilizing a multi-strategy Chinese open relationship extraction algorithm.
Mining association relationships between two nodes in each node pair by using map software can comprise: searching the geographic distance between two nodes in each node pair from map software respectively; comparing the geographic distance with a neighbor node distance threshold, and if the geographic distance is greater than or equal to the neighbor node distance threshold, determining that no geographic neighbor relation exists between the two nodes; if the geographic distance is smaller than the neighbor node distance threshold, determining that a geographic neighbor relationship exists between the two nodes. And judging whether a geographic neighbor relation exists between the two nodes or not by setting a neighbor node distance threshold value. Further, a smaller neighbor node distance threshold can be set for a densely distributed area and a larger neighbor node distance threshold can be set for a sparsely distributed area in combination with the distribution situation of all nodes.
Mining the association relationship between two nodes in each node pair by using the artificial intelligence big model can comprise: generating corresponding association relation identification tasks according to the information of two nodes in the node pair; and inputting the association relation identification task into the artificial intelligent large model, and outputting the association relation of two nodes in the node pair. The artificial intelligence large model may employ chatgpt. For each node pair, information of two nodes can be input into a preset problem template, a prompt (prompt) is generated and input into an artificial intelligence large model, and the association relation between the nodes is obtained. If the output result of the artificial intelligence large model is not null (null and null indicate that the answer contains conditions such as no association, no recognition and the like), the association relation of the two nodes is obtained.
Based on the information source, using two nodes as a node pair, and utilizing at least one of a network document, map software and an artificial intelligence large model to mine the association relationship between the two nodes in each node pair, the method can comprise the following steps:
Initializing a node interface, a network document interface, a map software interface and an artificial intelligence large model interface;
generating a first association list, a second association list and a third association list;
Receiving information of each node pair from a node interface;
Extracting a first association relation between two nodes in each node pair from the network document according to the information of each node pair, and writing the first association relation into a first association relation list;
Searching the geographic distance between two nodes in each node pair from map software respectively;
comparing the geographic distance with a neighbor node distance threshold, and if the geographic distance is greater than or equal to the neighbor node distance threshold, determining that no geographic neighbor relation exists between the two nodes; if the geographic distance is smaller than the distance threshold value of the neighbor node, determining that a geographic neighbor relation exists between the two nodes;
writing information of two nodes with geographic neighbor relations into a second association relation list;
Generating corresponding association relation identification tasks according to the information of two nodes in the node pair;
Inputting the association relation identification task into the artificial intelligent large model, and outputting a third association relation of two nodes in the node pair;
writing the third association into a third association list;
And outputting the updated first association list, the updated second association list and the updated third association list.
The above steps can be realized by the following algorithm:
The input of the algorithm is a smart city node set V, a network document set D, geographic coordinate position information Loc of each node and an artificial intelligent large model interface chatgpt; the output is three association relations between two nodes in each node pair: a first association (relationship 1), a second association (relationship 2), and a third association (relationship 3).
Line 1 of the algorithm is a list of sources that initialize three associations.
Lines 2-5 are association relations mined in the document set based on encyclopedia, news and the like, wherein line 3 can be calculated by adopting a multi-strategy Chinese open relation extraction algorithm. Line 4 is to add the extracted information to the first association table.
Lines 6-12 are association relation methods based on geographic coordinate information mining. The 6 th row combines the node sets in pairs, the 7 th-8 th rows are used for carrying out geographic distance calculation on the nodes based on the map software interface API, the 9 th-10 th rows are used for adding the node pairs with geographic neighbor relations into a second association relation table, and t is a set neighbor node distance threshold.
Lines 13-18 are the excavation of the association relationship between the nodes by using the open API of the artificial intelligence large model chatgpt in the form of prompt question-answering, and line 14 is the question asking by designing a question template for each pair of nodes to obtain the relationship between the nodes. Lines 15-16 are for judging that the content of the answer of the artificial intelligence large model chatgpt is not null (null represents the condition that the answer refers to "no association exists" or "sorry cannot answer", etc.), and adding the association relationship into the third association relationship table.
The association relationship between the nodes obtained by the algorithm can be directly used for constructing the data of the smart city heterogeneous diagram and executing subsequent analysis.
In addition, the characteristics of the nodes are extracted and added into the smart city heterogeneous map. In S202, constructing a smart city heterogeneous graph with information source entities as nodes and association relations between the information source entities as edges may include: extracting characteristics of information source entities as characteristics of nodes by using a pre-trained first language model, and constructing the nodes and edges into a smart city heterogeneous diagram; the characteristics of the nodes comprise node names, static information of the nodes and dynamic information of the nodes.
That is, the characteristics of each node are extracted by using a pre-trained language model, and the characteristics of the node can be expressed by the following formula:
x=F(<[CLS],Name,[SEP0],text,[SEP1],dynamic,[SEP]>)；
Wherein F (-) represents a pre-trained language model, x represents the features extracted from the input of the node by the pre-trained language model, [ CLS ] represents the first special token (token) of the sequence, name represents the Name of the node, text represents the static information of the node, and dynamic represents the dynamic information of the node. [ SEP0] and [ SEP1] represent templates designed manually and are composed of token sequences, which aim to prompt the model to better understand the input components. The node characteristics obtained by the method can be used as an important component of the node characteristics in the subsequent calculation model analysis process of the smart city diagram.
According to the smart city data processing method provided by the embodiment of the invention, the two nodes are taken as one node pair, and the association relation between the two nodes in each node pair is mined by utilizing information sources such as network documents, map software, artificial intelligence large models and the like, so that the smart city heterogeneous graph is constructed by utilizing the association relation between each node and each node, and the deep mining of the multidimensional association relation is realized. Node characteristics are extracted through a pre-trained language model, so that the learning of the smart city map calculation model is facilitated.
The fourth embodiment of the present invention will be described below.
Fig. 3 is a schematic diagram of a modeling process of a smart city map calculation model according to an embodiment of the present invention.
On the basis of the above embodiments, the present embodiment further describes a process of training a smart city map calculation model.
In the smart city data processing method provided by the embodiment of the present invention, in S203, a smart city map calculation model is constructed based on a smart city heterogeneous map, and in the construction process, features in the smart city heterogeneous map are extracted and converted into a low-dimensional embedded representation, and then the low-dimensional embedded representation is embedded into the smart city map calculation model, which may include: semantic features and/or structural features are extracted from the smart city heterogeneous map, and the semantic features and/or structural features are embedded into a map rolling network (GCN) to learn the features of the smart city heterogeneous map, so as to generate a smart city map calculation model.
Traditional graph rolling networks rely mainly on designed model architecture to learn task-specific features, without purposefully extracting the naturally occurring structural and semantic features of the graph data. In this regard, embodiments of the present invention provide a semantic-structural graph convolutional network (SS-GCN) that enhances the analysis process of node classification and analysis of the graph convolutional network by extracting semantic features and structural features from a smart city heterogeneous graph embedded in the graph convolution network.
Extracting semantic features and/or structural features from the smart city heterogeneous map, embedding the semantic features and/or structural features into the map convolutional network to learn features of the smart city heterogeneous map, generating a smart city map calculation model may include:
acquiring characteristics of the nodes;
extracting semantic features and structural features from the smart city heterogeneous map;
Fusing the characteristics, semantic characteristics and structural characteristics of the nodes to obtain fused characteristics;
the fusion features are embedded into a graph convolution network to learn the features of the smart city heterogeneous graph, and a smart city graph calculation model is generated.
The semantic-structure graph convolutional network (SS-GCN) provided by the embodiment of the invention can be expressed by the following formula:
；
wherein, Is characteristic of the ith node,/>For embedded representation of structural features,/>For embedded representation of semantic features,/>For the graph-convolution network model, G is the smart city heterogeneous graph,/>A hidden state representation is generated for the graph convolution network.
As shown in fig. 3, in the process of constructing the smart platform graph computation model, the characteristics of the nodes, the semantic characteristics of the smart platform heterogeneous graph and the structural characteristics of the smart platform heterogeneous graph are embedded, so as to implement the classification processing (y 1、y2 … …) of the nodes by the smart platform graph computation model.
In the semantic-structure graph convolutional network (SS-GCN) provided by the embodiment of the invention, the characteristics, semantic characteristics and structural characteristics of the nodes are fused to obtain the fused characteristics, and the fused characteristics can be obtained through the following calculation:
；
wherein, For the fusion feature,/>For the node characteristics of the ith node,/>As a structural feature of the i-th node,For semantic features of the ith node,/>For matrix connection operations.
The semantic features can be obtained by learning semantic relation expression among entities of the smart city heterogeneous map by adopting a knowledge map embedding algorithm, the structural features can be obtained by learning structural information of the smart city heterogeneous map by adopting a map embedding algorithm through random walk, and then the semantic features are embedded into a map convolution network by fusing semantic-structure, so that the generalization performance of the model can be further improved.
Finally, the probability that the obtained hidden state is mapped into the category through the linear layer is represented, and then the objective function of model optimization is negative log likelihood loss, which can be shown as follows:
；
wherein, Is a negative log likelihood loss,/>For model parameters of a graph convolution network,/>For the model parameter/>The corresponding negative log likelihood loss is the smallest,/>W is a parameter matrix for the number of nodes,/>Is the transposed matrix of the parameter matrix W,/>For the hidden state representation corresponding to the ith node,/>The probability for each category corresponds to the ith node.
The semantic feature embedding method, the structural feature embedding method and the training method of the graph rolling network provided by the embodiment of the invention are respectively described below.
In the smart city data processing method provided by the embodiment of the invention, the knowledge graph embedding algorithm can be adopted to learn the semantic feature embedding. Given a set of triplesWherein E represents the entity set, R represents the relation type set, and knowledge graph embedding aims at finding a mapping/>The representation of entities and relationships in the smart city heterogeneous graph is referred to as a low-dimensional dense vector. This representation enables the smart city heterogeneous map to facilitate the use of downstream tasks.
The knowledge-graph embedding algorithm may employ TransE algorithm, which is optimized with the goal of satisfying the vector representationLearning can be performed using interval-based ordering loss (TRIPLET RANKING loss).
In the smart city data processing method provided by the embodiment of the invention, the semantic features are extracted from the smart city heterogeneous map, including:
Based on the smart city heterogeneous diagram, obtaining triple data consisting of a first node h, a second node t and an association relation r between the first node and the second node;
employing a knowledge-graph embedding algorithm to satisfy the vector representation For optimizing the target, learning to obtain semantic features of the smart city heterogeneous map.
Modeling of the relationship types is included in the TransE algorithm learning process, the relationship types obtained by open relationship extraction are more diversified, and for the case of directly applying extracted data, the embodiment of the invention regards the extracted data as only one relationship type by default, namely, the difference of the relationship types is not considered. For graph data of which the relationship types are normalized, a plurality of relationship types can be directly used. In addition, in the graph data, there may be more one-to-many or many-to-many relationships, and the problem provides a corresponding solution in other methods, such as TransR, transD.
In the embodiment of the invention, a knowledge graph embedding algorithm is adopted so that the vector representation meets the requirementFor optimization purposes, learning to obtain semantic features of the smart city heterogeneous map may include:
representing the triplet data as a random initialization vector;
sampling the random initialization vector to obtain a negative sample;
Learning the negative sample by taking h+t approximately equal to r as an objective function and a semantic loss function to obtain semantic features of the smart city heterogeneous map;
wherein, the semantic loss function is:
；
wherein, Is an interval parameter,/>To take positive number calculate,/>Is a negative sample set,/>For the sampled header entity,/>The tail entity obtained for sampling.
TransE the representation of the node is obtained by an unsupervised learning algorithm. The algorithm learns semantic relationships between nodes, which are not extracted by the original graph convolution network, and the graph convolution network mainly utilizes supervised learning tasks to passively extract neighborhood information from graph data. Finally, the representation is embedded by combining the knowledge graph, so that the generalization capability of the node representation is improved.
In the smart city data processing method provided by the embodiment of the invention, the structure characteristics of the smart city heterogeneous map can be obtained by adopting a map embedding algorithm through random walk learning of the structure information of the smart city heterogeneous map. Extracting structural features from the smart city heterogeneous map may include: the method aims at converting nodes in the smart city heterogeneous map into embedded representation with low dimension and density, and enabling similar nodes in the smart city heterogeneous map to be similar in low dimension space distance, and adopts a map embedding algorithm to learn the map structure embedded representation of the smart city heterogeneous map so as to obtain structural characteristics in the smart city heterogeneous map.
That is, the graph embedding algorithm aims to find a mapping: the nodes in the graph are converted into a low-dimensional dense embedded representation such that similar nodes in the graph are closely spaced in low-dimensional space.
The method for obtaining the structural characteristics of the smart city heterogeneous map by learning the map structure embedded representation of the smart city heterogeneous map by using a map embedding algorithm with the goal of converting the nodes in the smart city heterogeneous map into the embedded representation with low dimension and dense and enabling similar nodes in the smart city heterogeneous map to be similar in low dimension space distance can comprise the following steps:
Aiming at the network neighborhood which can be observed with maximum probability based on the embedding of the central node, adopting a bias random walk model to learn the relation structure of each node;
and calculating according to the vector representation of each node and the vector representation of the relation structure of each node to obtain the structural characteristics in the smart city heterogeneous diagram.
In the embodiment of the invention, a node2vec algorithm is adopted. By using the concept of a continuous-Skip-gram model in the graph data, the goal of this algorithm optimization is to be able to observe the neighborhood of the network with maximum probability based on the embedding of the central node u.
The graph embedding shows that the structure correlation information of the smart city heterogeneous graph is captured through an unsupervised learning algorithm. The graph convolutional network cannot extract the structural correlation of random walk in the supervised learning process. Finally, the generalization of the node features is enhanced by graph embedding the representation.
In the smart city data processing method provided by the embodiment of the present invention, the training method of the smart city map calculation model may include:
Acquiring an initial graph rolling network;
Setting initial parameters for an initial graph rolling network;
training the graph rolling network according to the semantic features and/or the structural features and the features of the nodes to obtain the prediction category of the nodes;
And optimizing model parameters of the graph rolling network by using the actual type of the node and the loss of the predicted type of the node until the type prediction target of the node is met, so as to obtain a smart city graph calculation model.
For the spectrum-based graph neural network research method, the graph data is regarded as undirected, the graph rolling network (GCN) model is a special graph neural network, and the regularized graph Laplacian matrix has better mathematical characteristics and can be used as a robust mathematical representation of the graph data in the heterogeneous graph of the smart city, as follows:
；
Wherein A is an adjacency matrix, and the diagonal matrix D is a degree matrix, and the calculation process is as follows:
；
Wherein the graph Laplace matrix of the robustness representation is a real symmetrical semi-positive definite matrix, and the matrix can be subjected to spectral decomposition according to mathematical properties, and the formula is shown as follows:
；
Wherein U is a matrix formed by eigenvectors, the eigenvectors form a set of orthogonal basis, and Λ is an eigenvalue matrix of L. The calculation process of the graph convolution is shown as follows:
；
Wherein the method comprises the steps of I is an identity matrix. /(I)Is/>The calculation process is the same as the above formula. /(I)Is the hidden state of each layer,/>Is the original input feature X,/>A nonlinear activation function.
The graph convolution network inputs the hidden state vector of the last layer of codes into a linear layer, and maps the hidden state vector into class probability so as to realize node classification tasks. There are also related studies that add residual connection or gating mechanisms, train deep graph rolling networks and model timing information.
In the embodiment of the invention, the category prediction target of the graph roll-up network for learning the characteristics of the smart city heterogeneous graph can be expressed by the following formula:
；/>
wherein, Is a negative log likelihood loss,/>For model parameters of a graph convolution network,/>For the model parameter/>The corresponding negative log likelihood loss is the smallest,/>W is a parameter matrix for the number of nodes,/>Is the transposed matrix of the parameter matrix W,/>For the hidden state representation corresponding to the ith node,/>The probability for each category corresponds to the ith node.
According to the smart city data processing method provided by the embodiment of the invention, the semantic features and the structural features are extracted from the smart city heterogeneous map and embedded into the graph rolling network learning to conduct node classification training, so that the generalization capability of the graph rolling network to node representation is enhanced, the graph rolling network depth analysis can be conducted based on the constructed smart city heterogeneous map, and the smart city data processing method can be widely applied to various intelligent platform scenes.
The fifth embodiment of the present invention will be described below.
Based on the method for constructing the smart city heterogeneous map and the method for constructing the smart city map calculation model provided by the embodiments of the present invention, given the smart city signal sources of multiple sources and multiple scenes, the smart city heterogeneous map can be constructed by adopting the ontology modeling technology, the association relation mining algorithm and the pre-training language model feature extraction method in the knowledge map. Then, based on the smart city heterogeneous diagram, the node characteristics represented by the pre-trained language model, the semantic characteristics learned by the knowledge graph embedding algorithm and the structural characteristics learned by the graph embedding algorithm are embedded into a graph neural network (SS-GCN) for training and prediction, and the whole calculation process can be completed in an artificial intelligent server.
According to the method for constructing the heterogeneous map of the smart city, which is provided by the embodiment of the invention, the heterogeneous map of the cross-scene is constructed through ontology construction, association relation mining algorithm and pre-training language model characteristic representation. The entity association relation mining method provided by the embodiment of the invention can be applied to knowledge extraction scenes of smart cities in various fields, can realize large-scale automatic knowledge graph construction, and can also be inserted into the flow of other natural language processing models.
According to the construction method of the smart city heterogeneous map, different heterogeneous map data can be constructed by connecting information of different nodes, and the visual function of map data is supported based on deep analysis of a smart city map calculation model.
In the smart city data processing method provided by the embodiment of the present invention, in S204, the smart city data processing task of the user equipment is received and smart city data processing is performed based on a smart city map calculation model, which may include:
Receiving a smart city monitoring task sent by user equipment;
Identifying node attributes and monitoring items in the smart city monitoring task;
Calling a smart city diagram calculation model to classify nodes belonging to node attributes in the smart city heterogeneous diagram according to the monitoring items and/or perform node clustering processing to obtain node classification results and/or node clustering results;
and outputting a node classification result and/or a node clustering result.
In some implementations of the embodiment of the invention, in places such as a data center and a monitoring center, the smart city data processing method provided by the embodiment of the invention can be applied to realize the visual display of all the map data in the smart city heterogeneous map and execute the smart city data analysis tasks such as monitoring and early warning. For cross-domain monitoring tasks, the monitoring object is determined by acquiring the monitoring authority of the cross-domain monitoring tasks, and the information of the monitoring object can comprise a node name or a node attribute. The target node for determining the node name can be directly added into graph calculation analysis, and the node for determining the node attribute can be used for performing graph calculation to determine the corresponding node based on the smart city graph calculation model. In the process of executing the monitoring and early warning task, the nodes in the smart city heterogeneous diagram are classified, clustered, linked and the like through the monitoring and early warning threshold value set by the monitoring and early warning task, for example, the nodes exceeding the monitoring and early warning threshold value and the nodes not exceeding the monitoring and early warning threshold value are visually displayed in different forms.
In S204, the steps of receiving the smart city data processing task of the user equipment and performing smart city data processing based on the smart city map calculation model may further include:
receiving a to-be-solved problem sent by user equipment;
Calling a smart city diagram calculation model to search information of related nodes of the problem to be solved and neighborhood information of the related nodes;
Invoking a pre-trained second language model to generate answer information according to the information of the related nodes and the neighborhood information of the related nodes;
and outputting the answer information.
In some implementations of the embodiments of the present invention, a user may solve a problem based on the smart city platform provided by the embodiments of the present invention. In the traditional scheme, the problems generated by users can be solved only in a single field, and the users are required to collect information of all fields for integration, for example, the users need to travel, the users are required to inquire road conditions by themselves to determine a transportation travel mode, inquire the accommodation condition of a destination, inquire the travel policy and the flow of people of tourist attractions, and the like, so that a reasonable travel plan can be formulated. By applying the smart city data processing method provided by the embodiment of the invention, the smart city graph calculation model is utilized to search the nodes and the neighborhood thereof related to the to-be-solved problem sent by the user equipment in a cross-domain manner, the cross-domain node association calculation is performed around the to-be-solved problem, and then the vector representation output by the smart city graph calculation model can be decoded into answer information by combining the second language model, so that the cross-domain search problem of the user is solved. If the user needs to travel, the smart city data processing method provided by the embodiment of the invention can search and obtain the multidimensional information such as traffic conditions, local travel policies, holiday people flow, weather and the like related to travel destinations so as to automatically generate a reasonable travel plan through association relation calculation, thereby realizing intelligent service of the smart city.
Based on the construction method of the smart city map calculation model provided by the embodiment of the invention, the cross-scene modeling of the smart city is realized based on the learning of the map calculation model on the smart city heterogeneous map, the method of extracting node characteristics and dynamic knowledge representation embedded by the map by using the pre-training language model is utilized, and unified dynamic knowledge representation is generated by fusing real-time information of different scenes and structure information in the global. Cross-scene correlation computation and analysis can be performed by a smart city graph computation model that fuses graph convolution networks and pre-trained language models.
By utilizing the smart city data processing method provided by the embodiment of the invention, implicit association modeling of cross-scenes can be realized, and further tasks such as supporting global optimal regulation and control, timely finding out problems, optimizing and scheduling resources and the like can be realized, so that the problems of cross-scene information fusion, analysis, general calculation and the like of various types of smart cities such as smart cities, smart public security, smart travel, smart government affairs, smart traffic, smart farms and the like can be solved, the method can be applied to the requirements of optimizing the intelligent analysis of data center graph data, and the effects of classifying graph data nodes and optimizing classification are solved.
By applying the smart city data processing method provided by the embodiment of the invention, in the initial stage of constructing a smart city, the smart city data processing method provided by the embodiment of the invention is used for supporting the development of business under the condition of insufficient knowledge in the field and carrying out system cold start analysis. The smart city data processing method provided by the embodiment of the invention can further comprise the following steps: and updating the smart city heterogeneous map by using the dynamic information of the nodes and the newly added nodes, and updating the smart city map calculation model by using the updated smart city heterogeneous map.
Various embodiments of the smart city data processing method are described above in detail, and on the basis of the embodiments, the invention also discloses a smart city data processing device, a device and a readable storage medium corresponding to the method.
The sixth embodiment of the present invention will be described.
Fig. 4 is a schematic structural diagram of a smart city data processing device according to an embodiment of the present invention.
As shown in fig. 4, the smart city data processing device provided in the embodiment of the present invention includes:
The acquisition unit 401 is used for acquiring data from information sources in multiple fields corresponding to the smart city and storing the data into the smart city storage system;
A mapping unit 402, configured to extract data of information sources in each domain from the smart city storage system, and construct a smart city heterogeneous map with information source entities as nodes and association relationships between the information source entities as edges;
A modeling unit 403, configured to construct a smart city map computation model based on the smart city heterogeneous map, and extract features in the smart city heterogeneous map, convert the features into a low-dimensional embedded representation, and then embed the feature into the smart city map computation model during the construction process;
The processing unit 404 is configured to deploy an application program interface of the smart city map computing model to the user equipment, so as to receive a smart city data processing task of the user equipment and perform smart city data processing based on the smart city map computing model.
In some implementations, the mapping unit 402 constructs a smart city heterogeneous map with information source entities as nodes and association relations between the information source entities as edges, including:
And constructing a smart city heterogeneous graph by taking the ontology of the information source entity as an entity node, the category of the information source entity as a virtual node and the association relationship between different entity nodes and the association relationship between the entity node and the virtual node as edges.
In some implementations, the mapping unit 402 constructs a smart city heterogeneous map with information source entities as nodes and association relations between the information source entities as edges, including:
Taking two nodes as a node pair, and utilizing at least one of a network document, map software and an artificial intelligence large model to mine the association relation between the two nodes in each node pair;
and constructing a smart city heterogeneous graph by utilizing the information of each node and the association relation between each node.
In some implementations, the mapping unit 402 utilizes the network document to mine the association between two nodes in each node pair, including:
And extracting association relations between two nodes in each node pair from the network document by adopting a multi-strategy Chinese open relation extraction algorithm.
In some implementations, mapping unit 402 utilizes map software to mine the association between two nodes in each node pair, including:
Searching the geographic distance between two nodes in each node pair from map software respectively;
Comparing the geographic distance with a neighbor node distance threshold, and if the geographic distance is greater than or equal to the neighbor node distance threshold, determining that no geographic neighbor relation exists between the two nodes; if the geographic distance is smaller than the neighbor node distance threshold, determining that a geographic neighbor relationship exists between the two nodes.
In some implementations, mapping unit 402 utilizes an artificial intelligence large model to mine associations between two nodes in each node pair, including:
Generating corresponding association relation identification tasks according to the information of two nodes in the node pair;
And inputting the association relation identification task into the artificial intelligent large model, and outputting the association relation of two nodes in the node pair.
In some implementations, the mapping unit 402 uses two nodes as a node pair, and uses at least one of the network document, the map software and the artificial intelligence large model to mine association relations between two nodes in each node pair, including:
Initializing a node interface, a network document interface, a map software interface and an artificial intelligence large model interface;
generating a first association list, a second association list and a third association list;
Receiving information of each node pair from a node interface;
Extracting a first association relation between two nodes in each node pair from the network document according to the information of each node pair, and writing the first association relation into a first association relation list;
Searching the geographic distance between two nodes in each node pair from map software respectively;
comparing the geographic distance with a neighbor node distance threshold, and if the geographic distance is greater than or equal to the neighbor node distance threshold, determining that no geographic neighbor relation exists between the two nodes; if the geographic distance is smaller than the distance threshold value of the neighbor node, determining that a geographic neighbor relation exists between the two nodes;
writing information of two nodes with geographic neighbor relations into a second association relation list;
Generating corresponding association relation identification tasks according to the information of two nodes in the node pair;
Inputting the association relation identification task into the artificial intelligent large model, and outputting a third association relation of two nodes in the node pair;
writing the third association into a third association list;
And outputting the updated first association list, the updated second association list and the updated third association list.
In some implementations, the mapping unit 402 constructs a smart city heterogeneous map with information source entities as nodes and association relations between the information source entities as edges, including:
Extracting characteristics of information source entities as characteristics of nodes by using a pre-trained first language model, and constructing the nodes and edges into a smart city heterogeneous diagram;
The characteristics of the nodes comprise node names, static information of the nodes and dynamic information of the nodes.
In some implementations, the modeling unit 403 builds a smart city map computation model based on the smart city heterogeneous map, and in the building process, extracts features in the smart city heterogeneous map to be converted into a low-dimensional embedded representation and then embeds the feature into the smart city map computation model, including:
Extracting semantic features and/or structural features from the smart city heterogeneous map, embedding the semantic features and/or structural features into the map convolutional network to learn the features of the smart city heterogeneous map, and generating a smart city map calculation model.
In some implementations, the modeling unit 403 extracts semantic features and/or structural features from the smart city heterogeneous map, embeds the semantic features and/or structural features into the map convolutional network to learn the features of the smart city heterogeneous map, generates a smart city map calculation model, including:
acquiring characteristics of the nodes;
extracting semantic features and structural features from the smart city heterogeneous map;
Fusing the characteristics, semantic characteristics and structural characteristics of the nodes to obtain fused characteristics;
the fusion features are embedded into a graph convolution network to learn the features of the smart city heterogeneous graph, and a smart city graph calculation model is generated.
In some implementations, the modeling unit 403 extracts semantic features from the smart city heterogeneous map, including:
Based on the smart city heterogeneous diagram, obtaining triple data consisting of a first node h, a second node t and an association relation r between the first node and the second node;
employing a knowledge-graph embedding algorithm to satisfy the vector representation For optimizing the target, learning to obtain semantic features of the smart city heterogeneous map.
In some implementations, modeling unit 403 employs a knowledge-graph embedding algorithm such that the vector representation satisfiesFor optimizing the target, learning to obtain semantic features of the smart city heterogeneous map comprises the following steps:
representing the triplet data as a random initialization vector;
sampling the random initialization vector to obtain a negative sample;
To be used for Learning the negative samples for the objective function and the semantic loss function to obtain semantic features of the smart city heterogeneous map;
wherein, the semantic loss function is:
；
wherein, Is an interval parameter,/>To take positive number calculate,/>Is a negative sample set,/>For the sampled header entity,/>The tail entity obtained for sampling.
In some implementations, the modeling unit 403 extracts structural features from the smart city heterogeneous map, including:
the method aims at converting nodes in the smart city heterogeneous map into embedded representation with low dimension and density, and enabling similar nodes in the smart city heterogeneous map to be similar in low dimension space distance, and adopts a map embedding algorithm to learn the map structure embedded representation of the smart city heterogeneous map so as to obtain structural characteristics in the smart city heterogeneous map.
In some implementations, the modeling unit 403 learns the graph structure embedded representation of the smart city heterogeneous graph using a graph embedding algorithm with the goal of converting nodes in the smart city heterogeneous graph into a low-dimensional dense embedded representation, and making similar nodes in the smart city heterogeneous graph close in low-dimensional spatial distance, to obtain structural features in the smart city heterogeneous graph, including:
Aiming at the network neighborhood which can be observed with maximum probability based on the embedding of the central node, adopting a bias random walk model to learn the relation structure of each node;
and calculating according to the vector representation of each node and the vector representation of the relation structure of each node to obtain the structural characteristics in the smart city heterogeneous diagram.
In some implementations, the modeling unit 403 performs fusion processing on the feature, the semantic feature, and the structural feature of the node to obtain a fused feature, which can be calculated by the following formula:
；
wherein, For the fusion feature,/>For the node characteristics of the ith node,/>As a structural feature of the i-th node,For semantic features of the ith node,/>For matrix connection operations.
In some implementations, a training method of a smart city map computing model includes:
Acquiring an initial graph rolling network;
Setting initial parameters for an initial graph rolling network;
training the graph rolling network according to the semantic features and/or the structural features and the features of the nodes to obtain the prediction category of the nodes;
And optimizing model parameters of the graph rolling network by using the actual type of the node and the loss of the predicted type of the node until the type prediction target of the node is met, so as to obtain a smart city graph calculation model.
In some implementations, the category prediction target may be represented by the following formula:
；/>
wherein, Is a negative log likelihood loss,/>For model parameters of a graph convolution network,/>For the model parameter/>The corresponding negative log likelihood loss is the smallest,/>W is a parameter matrix for the number of nodes,/>Is the transposed matrix of the parameter matrix W,/>For the hidden state representation corresponding to the ith node,/>The probability for each category corresponds to the ith node.
In some implementations, the processing unit 404 receives the user device's smart city data processing tasks and performs smart city data processing based on a smart city map computing model, including:
Receiving a smart city monitoring task sent by user equipment;
Identifying node attributes and monitoring items in the smart city monitoring task;
Calling a smart city diagram calculation model to classify nodes belonging to node attributes in the smart city heterogeneous diagram according to the monitoring items and/or perform node clustering processing to obtain node classification results and/or node clustering results;
and outputting a node classification result and/or a node clustering result.
In some implementations, the processing unit 404 receives the user device's smart city data processing tasks and performs smart city data processing based on a smart city map computing model, including:
receiving a to-be-solved problem sent by user equipment;
Calling a smart city diagram calculation model to search information of related nodes of the problem to be solved and neighborhood information of the related nodes;
Invoking a pre-trained second language model to generate answer information according to the information of the related nodes and the neighborhood information of the related nodes;
and outputting the answer information.
Since the embodiments of the apparatus portion and the embodiments of the method portion correspond to each other, the embodiments of the apparatus portion are referred to the description of the embodiments of the method portion, and are not repeated herein.
The seventh embodiment of the present invention will be described.
Fig. 5 is a schematic structural diagram of a smart city data processing device according to an embodiment of the present invention.
As shown in fig. 5, the smart city data processing device provided by the embodiment of the present invention includes:
a memory 510 for storing a computer program 511;
A processor 520 for executing a computer program 511, which computer program 511 when executed by the processor 520 implements the steps of the smart city data processing method according to any of the embodiments described above.
Processor 520 may include one or more processing cores, such as a 3-core processor, an 8-core processor, etc., among others. The processor 520 may be implemented in at least one hardware form of a digital signal Processing DSP (DIGITAL SIGNAL Processing), a Field-Programmable gate array FPGA (Field-Programmable GATE ARRAY), or a Programmable logic array PLA (Programmable Logic Array). Processor 520 may also include a main processor, which is a processor for processing data in an awake state, also referred to as central processor CPU (Central Processing Unit), and a coprocessor; a coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 520 may be integrated with an image processor GPU (Graphics Processing Unit), a GPU for use in responsible for rendering and rendering of the content that is to be displayed by the display screen. In some embodiments, the processor 520 may also include an artificial intelligence AI (Artificial Intelligence) processor for processing computing operations related to machine learning.
The memory 510 may include one or more readable storage media, which may be non-transitory. Memory 510 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In this embodiment, the memory 510 is at least used for storing a computer program 511, where the computer program 511 can implement relevant steps in the smart city data processing method disclosed in any of the foregoing embodiments after being loaded and executed by the processor 520. In addition, the resources stored in the memory 510 may further include an operating system 512, data 513, and the like, where the storage manner may be transient storage or permanent storage. The operating system 512 may be Windows. The data 513 may include, but is not limited to, data related to the methods described above.
In some embodiments, the smart city data processing device may further include a display 530, a power supply 540, a communication interface 550, an input-output interface 560, a sensor 570, and a communication bus 580.
It will be appreciated by those skilled in the art that the architecture shown in fig. 5 is not limiting of the smart city data processing device and may include more or fewer components than shown.
The smart city data processing device provided by the embodiment of the invention comprises the memory and the processor, wherein the processor can realize the smart city data processing method when executing the program stored in the memory, and the effects are the same as the above.
The eighth embodiment of the present invention will be described.
The embodiment of the invention also provides a readable storage medium, and the readable storage medium stores a computer program which realizes steps of a smart city data processing method when being executed by a processor.
The readable storage medium may include: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory RAM (Random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The computer program included in the readable storage medium provided in this embodiment can implement the steps of the smart city data processing method as described above when executed by the processor, and the same effects as above.
The method, the device, the equipment and the readable storage medium for processing the smart city data provided by the invention are described in detail. In the description, each embodiment is described in a progressive manner, and each embodiment is mainly described by the differences from other embodiments, so that the same similar parts among the embodiments are mutually referred. The apparatus, device and readable storage medium disclosed in the embodiments are relatively simple to describe, and the relevant points refer to the description of the method section since they correspond to the methods disclosed in the embodiments. It should be noted that it will be apparent to those skilled in the art that the present invention may be modified and practiced without departing from the spirit of the present invention.
It should also be noted that in this specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
Claims (20)
1. A smart city data processing method, comprising:
collecting data from information sources in multiple fields corresponding to the smart city and storing the data into a smart city storage system;
Extracting data of the information sources in each field from the smart city storage system, and constructing a smart city heterogeneous graph taking information source entities as nodes and the association relationship among the information source entities as edges;
constructing a smart city map calculation model based on the smart city heterogeneous map, and extracting features in the smart city heterogeneous map to be converted into a low-dimensional embedded representation and then embedding the low-dimensional embedded representation into the smart city map calculation model in the construction process;
deploying an application program interface of the smart city map calculation model to user equipment to receive a smart city data processing task of the user equipment and processing smart city data based on the smart city map calculation model;
The smart city map calculation model is built based on the smart city heterogeneous map, features in the smart city heterogeneous map are extracted to be converted into low-dimensional embedded representation and then embedded into the smart city map calculation model in the building process, and the smart city map calculation model comprises the following steps:
Acquiring the characteristics of the node;
extracting semantic features and structural features from the smart city heterogeneous map;
fusing the characteristics of the nodes, the semantic characteristics and the structural characteristics to obtain fused characteristics;
And embedding the fusion features into a graph rolling network to learn the features of the smart city heterogeneous graph, and generating the smart city graph calculation model.
2. The smart city data processing method of claim 1, wherein the constructing a smart city heterogeneous map with information source entities as nodes and association relationships between the information source entities as edges comprises:
And constructing the smart city heterogeneous graph by taking the ontology of the information source entity as an entity node, the category of the information source entity as a virtual node, and the association relationship between different entity nodes and the association relationship between the entity node and the virtual node as the edges.
3. The smart city data processing method of claim 1, wherein the constructing a smart city heterogeneous map with information source entities as nodes and association relationships between the information source entities as edges comprises:
Taking two nodes as a node pair, and excavating association relations between the two nodes in each node pair by utilizing at least one of a network document, map software and an artificial intelligence large model;
and constructing the smart city heterogeneous graph by utilizing the information of the nodes and the association relation between the nodes.
4. A smart city data processing method as claimed in claim 3, wherein mining association between two of the nodes in each of the node pairs using the web documents comprises:
And extracting association relations between two nodes in each node pair from the network document by adopting a multi-strategy Chinese open relation extraction algorithm.
5. A smart city data processing method as claimed in claim 3, wherein mining the association between two of the nodes in each of the node pairs using the map software comprises:
searching for a geographic distance between two of the nodes in each of the node pairs from the map software, respectively;
Comparing the geographic distance with a neighbor node distance threshold, and if the geographic distance is greater than or equal to the neighbor node distance threshold, determining that no geographic neighbor relation exists between the two nodes; and if the geographic distance is smaller than the distance threshold value of the neighbor nodes, determining that the geographic neighbor relation exists between the two nodes.
6. A smart city data processing method as claimed in claim 3, wherein mining the association between two of the nodes in each of the node pairs using the artificial intelligence large model comprises:
generating corresponding association relation identification tasks according to the information of the two nodes in the node pair;
inputting the association relation identification task into the artificial intelligent large model, and outputting the association relation of two nodes in the node pair.
7. A smart city data processing method as claimed in claim 3, wherein said mining the association between two of said nodes in each of said node pairs using at least one of web documents, map software and artificial intelligence large models using two of said nodes as a node pair comprises:
Initializing a node interface, a network document interface, a map software interface and an artificial intelligence large model interface;
generating a first association list, a second association list and a third association list;
receiving information of each node pair from the node interface;
extracting first association relations between two nodes in each node pair from the network document according to the information of the node pairs, and writing the first association relations into the first association relation list;
searching for a geographic distance between two of the nodes in each of the node pairs from the map software, respectively;
Comparing the geographic distance with a neighbor node distance threshold, and if the geographic distance is greater than or equal to the neighbor node distance threshold, determining that no geographic neighbor relation exists between the two nodes; if the geographic distance is smaller than the distance threshold value of the neighbor node, determining that the geographic neighbor relation exists between the two nodes;
Writing information of two nodes with the geographic neighbor relation into the second association relation list;
generating corresponding association relation identification tasks according to the information of the two nodes in the node pair;
Inputting the association relation identification task into the artificial intelligent large model, and outputting a third association relation of two nodes in the node pair;
writing the third association into the third association list;
and outputting the updated first association list, the updated second association list and the updated third association list.
8. The smart city data processing method of claim 1, wherein the constructing a smart city heterogeneous map with information source entities as nodes and association relationships between the information source entities as edges comprises:
extracting the characteristics of the information source entity as the characteristics of the nodes by using a pre-trained first language model, and constructing the nodes and the edges into the smart city heterogeneous diagram;
the characteristics of the node comprise a node name, static information of the node and dynamic information of the node.
9. The smart city data processing method of claim 1, wherein extracting the semantic features from the smart city heterogeneous map comprises:
Based on the smart city heterogeneous diagram, obtaining triple data consisting of a first node h, a second node t and an association relation r between the first node and the second node;
employing a knowledge-graph embedding algorithm to satisfy the vector representation For optimization purposes, learning the semantic features of the smart city heterogeneous map.
10. The smart city data processing method of claim 9, wherein the knowledge-graph embedding algorithm is employed such that the vector representation satisfiesFor optimization purposes, learning the semantic features of the smart city heterogeneous map, comprising:
Representing the triplet data as a random initialization vector;
Sampling the random initialization vector to obtain a negative sample;
To be used for Learning the negative sample for an objective function and a semantic loss function to obtain the semantic features of the smart city heterogeneous diagram;
Wherein the semantic loss function is:
；
wherein, Is an interval parameter,/>To take positive number calculate,/>Is a negative sample set,/>For the sampled header entity,/>The tail entity obtained for sampling.
11. The smart city data processing method of claim 1, wherein extracting the structural features from the smart city heterogeneous map comprises:
And aiming at converting the nodes in the smart city heterogeneous map into embedded representations with low dimension and density and enabling similar nodes in the smart city heterogeneous map to be similar in low dimension space distance, learning the map structure embedded representations of the smart city heterogeneous map by adopting a map embedding algorithm, and obtaining the structural characteristics in the smart city heterogeneous map.
12. The smart city data processing method of claim 11, wherein the learning the graph structure embedded representation of the smart city heterogeneous graph with the graph embedding algorithm with the goal of converting the nodes in the smart city heterogeneous graph into a low-dimensional dense embedded representation such that similar nodes in the smart city heterogeneous graph are closely spaced in low-dimensional space comprises:
the method comprises the steps of taking the network neighborhood which can be observed with maximum probability based on the embedding of a central node as a target, and adopting a bias random walk model to learn the relation structure of each node;
and calculating the structural characteristics in the smart city heterogeneous graph according to the vector representation of each node and the vector representation of the relation structure of each node.
13. The smart city data processing method of claim 1, wherein the fusing of the node features, the semantic features, and the structural features results in a fused feature, which is calculated by the following formula:
；
wherein, For the fusion feature,/>For the node characteristics of the ith node,/>For the structural feature of the ith node,/>For the semantic features of the ith node,/>For matrix connection operations.
14. The smart city data processing method of claim 1, wherein the training method of the smart city map calculation model comprises:
acquiring an initial graph rolling network;
Setting initial parameters for the initial graph rolling network;
Training the graph convolution network according to the semantic features and/or the structural features and the features of the nodes to obtain a prediction category of the nodes;
And optimizing model parameters of the graph convolution network by using the actual category of the node and the loss of the predicted category of the node until the category prediction target of the node is met, so as to obtain the smart city graph calculation model.
15. The smart city data processing method of claim 14, wherein the category prediction target is represented by:
；
wherein, Is a negative log likelihood loss,/>For model parameters of the graph convolution network,/>For the model parameter/>The corresponding negative log likelihood loss is the smallest,/>W is a parameter matrix for the number of nodes,/>Is the transposed matrix of the parameter matrix W,/>For the hidden state representation corresponding to the ith node,And (5) the probability of each category corresponding to the ith node.
16. The smart city data processing method of claim 1, wherein the receiving the smart city data processing task of the user device and performing smart city data processing based on the smart city map calculation model comprises:
Receiving a smart city monitoring task sent by the user equipment;
identifying node attributes and monitoring items in the smart city monitoring task;
Invoking the smart city diagram calculation model to perform node classification and/or node clustering processing on the nodes belonging to the node attribute in the smart city heterogeneous diagram according to the monitoring item to obtain a node classification result and/or a node clustering result;
And outputting the node classification result and/or the node clustering result.
17. The smart city data processing method of claim 1, wherein the receiving the smart city data processing task of the user device and performing smart city data processing based on the smart city map calculation model comprises:
receiving a to-be-solved problem sent by the user equipment;
Invoking the smart city diagram calculation model to search information of related nodes of the to-be-solved problem and neighborhood information of the related nodes;
Invoking a pre-trained second language model to generate answer information according to the information of the related nodes and the neighborhood information of the related nodes;
And outputting the answer information.
18. A smart city data processing apparatus, comprising:
The acquisition unit is used for acquiring data from information sources in multiple fields corresponding to the smart city and storing the data into the smart city storage system;
the mapping unit is used for extracting the data of the information sources in each field from the smart city storage system and constructing a smart city heterogeneous map which takes information source entities as nodes and the association relationship between the information source entities as edges;
the modeling unit is used for constructing a smart city map calculation model based on the smart city heterogeneous map, and extracting features in the smart city heterogeneous map to be converted into a low-dimensional embedded representation and then embedded into the smart city map calculation model in the construction process;
The processing unit is used for deploying the application program interface of the smart city map calculation model to the user equipment so as to receive the smart city data processing task of the user equipment and process the smart city data based on the smart city map calculation model;
The smart city map calculation model is built based on the smart city heterogeneous map, features in the smart city heterogeneous map are extracted to be converted into low-dimensional embedded representation and then embedded into the smart city map calculation model in the building process, and the smart city map calculation model comprises the following steps:
Acquiring the characteristics of the node;
extracting semantic features and structural features from the smart city heterogeneous map;
fusing the characteristics of the nodes, the semantic characteristics and the structural characteristics to obtain fused characteristics;
And embedding the fusion features into a graph rolling network to learn the features of the smart city heterogeneous graph, and generating the smart city graph calculation model.
19. A smart city data processing apparatus, comprising:
a memory for storing a computer program;
A processor for executing said computer program, which when executed by said processor implements the steps of the smart city data processing method as claimed in any one of claims 1 to 17.
20. A readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the smart city data processing method as claimed in any of claims 1 to 17.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410078615.XA CN117592006B (en) | 2024-01-19 | 2024-01-19 | Smart city data processing method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410078615.XA CN117592006B (en) | 2024-01-19 | 2024-01-19 | Smart city data processing method, device, equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117592006A CN117592006A (en) | 2024-02-23 |
| CN117592006B true CN117592006B (en) | 2024-04-26 |
Family
ID=89920579
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410078615.XA Active CN117592006B (en) | 2024-01-19 | 2024-01-19 | Smart city data processing method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117592006B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112214483A (en) * | 2019-07-11 | 2021-01-12 | 广联达科技股份有限公司 | Method and device for analyzing, associating, storing and accessing data in urban information model |
| CN112598563A (en) * | 2020-12-29 | 2021-04-02 | 中国科学技术大学 | Smart city data construction method based on knowledge graph |
| WO2023123182A1 (en) * | 2021-12-30 | 2023-07-06 | 深圳晶泰科技有限公司 | Multi-source heterogeneous data processing method and apparatus, computer device and storage medium |
| CN116991916A (en) * | 2023-08-09 | 2023-11-03 | 广东纬昊科技有限公司 | Smart city regional management method and system based on multi-source data analysis |
| CN117272017A (en) * | 2023-08-23 | 2023-12-22 | 北京邮电大学 | Training method for heterogeneous graph data node embedded feature extraction model, embedded feature extraction method, node classification method and device |
-
2024
- 2024-01-19 CN CN202410078615.XA patent/CN117592006B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112214483A (en) * | 2019-07-11 | 2021-01-12 | 广联达科技股份有限公司 | Method and device for analyzing, associating, storing and accessing data in urban information model |
| CN112598563A (en) * | 2020-12-29 | 2021-04-02 | 中国科学技术大学 | Smart city data construction method based on knowledge graph |
| WO2023123182A1 (en) * | 2021-12-30 | 2023-07-06 | 深圳晶泰科技有限公司 | Multi-source heterogeneous data processing method and apparatus, computer device and storage medium |
| CN116991916A (en) * | 2023-08-09 | 2023-11-03 | 广东纬昊科技有限公司 | Smart city regional management method and system based on multi-source data analysis |
| CN117272017A (en) * | 2023-08-23 | 2023-12-22 | 北京邮电大学 | Training method for heterogeneous graph data node embedded feature extraction model, embedded feature extraction method, node classification method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117592006A (en) | 2024-02-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Satellite image spoofing: Creating remote sensing dataset with generative adversarial networks (short paper) | |
| CN111885399B (en) | Content distribution method, device, electronic equipment and storage medium | |
| CN109543690A (en) | Method and apparatus for extracting information | |
| CN109783666A (en) | A kind of image scene map generation method based on iteration fining | |
| CN109033107A (en) | Image search method and device, computer equipment and storage medium | |
| CN111368789B (en) | Image recognition method, device, computer equipment and storage medium | |
| CN111914156A (en) | Cross-modal retrieval method and system for self-adaptive label perception graph convolution network | |
| CN113051927B (en) | Social network emergency detection method based on multi-modal graph convolutional neural network | |
| CN111652329B (en) | Image classification method and device, storage medium and electronic equipment | |
| CN115809327B (en) | Real-time social network rumor detection method based on multimode fusion and topics | |
| Forestier et al. | Coastal image interpretation using background knowledge and semantics | |
| CN111598252A (en) | University computer basic knowledge problem solving method based on deep learning | |
| Wang et al. | Human-instructed deep hierarchical generative learning for automated urban planning | |
| CN113569814A (en) | Unsupervised pedestrian re-identification method based on feature consistency | |
| CN111881997B (en) | Multi-modal small sample learning method based on significance | |
| CN117592006B (en) | Smart city data processing method, device, equipment and readable storage medium | |
| CN116362294B (en) | Neural network searching method and device and readable storage medium | |
| Lüscher et al. | Where is the terraced house? On the use of ontologies for recognition of urban concepts in cartographic databases | |
| Ballatore | Semantic challenges for volunteered geographic information | |
| CN110096506B (en) | Tree cell structure description and storage method for multi-layer requirements | |
| Li et al. | Prediction model of urban street public space art design indicators based on deep convolutional neural network | |
| CN114565730A (en) | Method and device for processing map vector sample and computer readable storage medium | |
| Quintero et al. | An ontology-driven approach for the extraction and description of geographic objects contained in raster spatial data | |
| Chen et al. | Remote sensing image monitoring and recognition technology for the conservation of rare wild animals | |
| Liu et al. | Extracting River Illegal Buildings from UAV Image Based on Deeplabv3+ |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |