Method for classifying urban functional areas based on place2vec and graph convolutional neural network
By combining Place2vec and graph convolutional neural networks, an undirected graph of POIs is constructed and network structure features are learned, which solves the problem of inaccurate representation of urban functional area features and achieves more accurate functional area classification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- QINGDAO UNIV OF TECH
- Filing Date
- 2023-07-13
- Publication Date
- 2026-06-23
AI Technical Summary
Existing technologies do not accurately represent the characteristics of urban functional zones, especially the classification methods of urban functional zones based on POI data, which fail to fully consider the influence of the network topology of POIs on urban functional types.
This paper adopts a combination of Place2vec and graph convolutional neural network. By constructing an undirected graph of POIs, the Place2vec model is used to generate the initialization vectors of POI nodes. The network structure features of POI nodes are learned in the graph convolutional neural network. By integrating the local spatial association features and network topology features of POIs, the classification of urban functional areas can be achieved.
It improves the accuracy of urban functional zone classification, enabling more accurate identification of functional zone types, such as distinguishing between residential and commercial areas, thus enhancing classification accuracy.
Smart Images
Figure CN116956151B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of geographic information technology, specifically relating to a method for classifying urban functional areas based on Place2vec and graph convolutional neural networks. Background Technology
[0002] Urban functional zones are the basic units of urban management and planning. The classification of urban functional zones is not only of great significance to urban management and planning, but also provides convenience for human life, work, and transportation. However, the high complexity of urban systems presents a significant challenge to the classification of urban functional zones. On the one hand, the spatial structure of urban functional zones depends not only on government decisions but is also influenced by various factors such as people's daily lives and the market economy. On the other hand, the rapid urbanization process in recent years has led to increasingly complex spatial structures of urban functional zones. Therefore, research on methods for classifying urban functional zones has become a current research hotspot.
[0003] Urban functional zone classification methods based on POI data are an important current research direction. Currently, research on urban functional zone classification methods mainly relies on high spatial resolution remote sensing imagery, POIs (Points of Interest, points representing geographic entities on electronic maps), and socially-sensory big data (such as data from social media, public transport card payments, mobile communications, and shared bicycles). Compared to high spatial resolution remote sensing imagery and socially-sensory big data, POI data is object-based data that can be directly used to characterize geographic entities with socio-economic attributes. It also features easy accessibility, comprehensive data coverage, and high data integrity, thus attracting widespread attention from scholars, and research on POI-based urban functional zone classification methods has made some progress.
[0004] Currently, some scholars have used models such as Word2vec, Place2vec, and Zone2vec to generate vector representations of POI types and functional zones based on the local spatial relationships of POIs, and combined them with machine learning algorithms to construct urban functional zone classification methods. However, these methods do not accurately express the characteristics of urban functional zones. Summary of the Invention
[0005] This invention provides a classification method for urban functional zones based on Place2vec and graph convolutional neural networks, aiming to solve the problem of insufficient accuracy in the representation of urban functional zone features in existing technologies.
[0006] To achieve the above objectives, this invention provides a method for classifying urban functional zones based on Place2vec and graph convolutional neural networks, the method comprising:
[0007] Construct an undirected graph of POIs based on POI data for urban functional areas;
[0008] Based on the preset Plane2vec model, each POI node in the POI undirected graph and its connected POI nodes are taken as input, and the initialization vector of the POI type of the POI node is output.
[0009] Based on the POI undirected graph, construct the network topology of the graph convolutional neural network;
[0010] The initialization vector of the POI type is used as the initialization vector of the POI node in the network topology of the graph convolutional neural network. The network structure features of the POI node are learned to obtain the probability value of the functional type of the functional area.
[0011] Preferably, in the urban functional area classification method based on Plane2vec and graph convolutional neural networks, the step of using the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network, learning the network structure features of the POI node, and obtaining the probability value of the functional type of the functional area includes:
[0012] The initialization vector of the POI type is used as the initialization vector of the POI node in the network topology of the graph convolutional neural network;
[0013] In each layer of the graph convolutional neural network network topology, information is aggregated for POI nodes and their connected nodes, and information is transferred between layers in the graph convolutional neural network network topology.
[0014] In the graph convolutional neural network topology, all POI nodes are pointed to a preset virtual node, and all POI node information is propagated to the virtual node.
[0015] Based on the characteristics of the virtual nodes, a probability vector for the urban functional area type is determined.
[0016] Preferably, in the urban functional area classification method based on Plane2vec and graph convolutional neural networks, the step of aggregating information about POI nodes and their connected nodes in each layer of the graph convolutional neural network network topology, and transferring information between layers in the graph convolutional neural network network topology, includes:
[0017] In the network topology of each layer of the graph convolutional neural network, information is aggregated for POI nodes and their connected nodes;
[0018] The aggregation result is passed as a new feature value of the POI node to the network topology of the next layer of the graph convolutional neural network.
[0019] Preferably, in the urban functional area classification method based on Plane2vec and graph convolutional neural networks, the calculation process for the step of aggregating information about POI nodes and their connected nodes in each layer of the graph convolutional neural network network topology, and transferring information between layers in the graph convolutional neural network network topology, is as follows:
[0020]
[0021] in,
[0022] H (l) The feature vector representing the POI node at layer l;
[0023] H(0) is the initialization vector X of the POI node;
[0024] W (l) The parameters are the convolution parameters of the l-th layer;
[0025] A is the POI adjacency matrix;
[0026] I is the identity matrix;
[0027] It is the degree matrix based on the POI network structure;
[0028] σ(·) is a non-linear activation operation.
[0029] Preferably, in the urban functional area classification method based on Plane2vec and graph convolutional neural networks, the step of constructing an undirected graph of POIs based on the POI data of urban functional areas includes:
[0030] The POIs of urban functional areas are used as nodes in the POI undirected graph, and the neighbors of the current POI node are generated based on the spatial distance of the POIs.
[0031] Connect the current POI node with its neighbors to form an undirected POI graph.
[0032] Preferably, in the urban functional area classification method based on Plane2vec and graph convolutional neural networks, the step of using the POIs of urban functional areas as nodes in an undirected POI graph and generating the neighbors of the current POI node based on the spatial distance of the POIs includes:
[0033] Use the POIs of urban functional areas as nodes in an undirected POI graph;
[0034] The KNN algorithm selects a preset number of POI nodes that are closest to the current POI node as the current POI node's neighbors.
[0035] Preferably, in the urban functional area classification method based on Platform2vec and graph convolutional neural networks, the step of taking each POI node in the undirected POI graph and its connected POI points as input, and outputting the initialization vector of the POI type of the POI node according to the preset Platform2vec model, includes:
[0036] Based on the preset Plane2vec model, urban functional areas are treated as documents, POI types are treated as words, and each POI node in the undirected POI graph and its connected POI points are taken as input, and the initialization vector of the POI type of the POI node is output.
[0037] To achieve the above objectives, the present invention also provides an urban functional area classification device based on Plane2vec and graph convolutional neural networks, comprising:
[0038] The building unit is used to construct an undirected graph of POIs based on the POI data of urban functional areas;
[0039] The initial unit is used to take each POI node in the POI undirected graph and the POI points connected to it as input according to the preset Plan2vec model, and output the initialization vector of the POI type of the POI node.
[0040] The generation unit is used to construct the network topology of the graph convolutional neural network based on the POI undirected graph;
[0041] The output unit is used to take the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network, learn the network structure features of the POI node, and obtain the probability value of the functional type of the functional area.
[0042] To achieve the above objectives, the present invention also provides an electronic device, comprising:
[0043] At least one processor; and,
[0044] A memory communicatively connected to the at least one processor; wherein,
[0045] The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the above-described urban functional area classification method based on Plane2vec and graph convolutional neural networks.
[0046] To achieve the above objectives, the present invention also provides a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the above-described urban functional area classification method based on Plane2vec and graph convolutional neural networks.
[0047] The technical solution provided by this invention has at least the following advantages:
[0048] The urban functional area classification method based on Platform2vec and graph convolutional neural networks provided by this invention first constructs an undirected graph of POIs based on POI data of urban functional areas. Then, according to a preset Platform2vec model, each POI node in the undirected graph and its connected POIs are used as input to output the initialization vector of the POI type of the POI node. Next, a network topology of a graph convolutional neural network is constructed based on the undirected POI graph. Finally, the initialization vector of the POI type is used as the initialization vector of the POI node in the network topology of the graph convolutional neural network to learn the network structure features of the POI nodes and obtain the probability value of the functional type of the functional area. In this way, the urban functional area classification method established by combining Platform2vec and graph convolutional neural network models can achieve the classification of urban functional areas by integrating the local spatial correlation features of POIs and the network topology features, and the expression accuracy is more precise. Attached Figure Description
[0049] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0050] Figure 1 This is a schematic diagram of functional area A;
[0051] Figure 2 This is a schematic diagram of functional area B;
[0052] Figure 3 This is a schematic diagram of one embodiment of the urban functional area classification method based on Plane2vec and graph convolutional neural network of the present invention;
[0053] Figure 4 This is a schematic diagram of another embodiment of the urban functional area classification method based on Plane2vec and graph convolutional neural network of the present invention;
[0054] Figure 5This is a schematic diagram of an embodiment of the urban functional area classification device based on Plane2vec and graph convolutional neural network of the present invention;
[0055] Figure 6 This is a schematic diagram of an embodiment of an electronic device.
[0056] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0057] In this embodiment of the invention, the term "and / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the preceding and following associated objects have an "or" relationship.
[0058] It should be noted that the terms "first," "second," etc., in the specification, claims, and drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.
[0059] In this embodiment of the invention, the term "multiple" refers to two or more, and other quantifiers are similar.
[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the embodiments of the present invention to facilitate a better understanding of the invention. However, the technical solutions claimed in the present invention can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the following embodiments is for ease of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined with and referenced by each other without contradiction.
[0061] Currently, some scholars have used models such as Word2vec, Plane2vec, and Zone2vec to generate vector representations of POI types and functional zones based on the local spatial relationships of POIs, and combined them with machine learning algorithms to construct urban functional zone classification methods. However, these methods are not accurate enough.
[0062] like Figure 1 and Figure 2 As shown, Figure 1 This diagram illustrates functional area A. Please refer to [link / reference]. Figure 1 middle, Figure 1 It includes 6 building numbers (residential), 7 convenience stores, and 3 companies; Figure 2 This diagram illustrates functional area B. Figure 2 This includes 7 building numbers (residential), 7 convenience stores, and 3 companies. Assuming classification is done using existing technological methods... Figure 1 and Figure 2 The two images are similar in quantity and type (the number and type of building numbers, convenience stores, and companies are roughly the same). If we disregard the POI network structure features and only consider the local spatial relationship features of POIs, then the two images have the same functional type. However, upon closer analysis... Figure 1 It should actually be considered as a combination of residential and commercial use. Figure 2 The buildings and companies are mixed together, representing a relatively mature residential function.
[0063] Compared to urban functional area classification methods based on POI statistical features and POI semantic features, methods based on POI spatial association features consider the local spatial associations of POIs, thus improving the accuracy of functional area representation. However, these methods still do not consider the influence of POI network topology on urban functional types, limiting the accuracy of urban functional area feature representation.
[0064] To address the aforementioned issues, this implementation method relates to a city functional area classification method based on Place2vec and graph convolutional neural networks. This method can be applied to electronic devices, such as desktop computers, tablets, laptops, and other electronic devices with data processing capabilities, without any specific limitations.
[0065] The following describes the implementation details of the urban functional area classification method based on Place2vec and graph convolutional neural network according to the first embodiment of the present invention. The following implementation details are provided for ease of understanding only and are not necessary for implementing this solution.
[0066] The specific process of this implementation method is as follows: Figure 3 As shown, it specifically includes:
[0067] Step S100: Construct an undirected graph of POIs based on the POI data of urban functional areas;
[0068] It should be understood that a POI (Point of Interest) is a point on an electronic map that represents a geographic entity. POI data is object-based data that can be directly used to characterize geographic entities with socioeconomic attributes, and it is characterized by easy accessibility, comprehensive data coverage, and high data integrity. POIs, representing geographic entities, are fundamental elements constituting urban functional zones, and their spatial structural characteristics are a significant factor leading to differences in urban functional zone types. From a sociological perspective, POIs, as carriers of urban social and economic functions, form an invisible social network based on their spatial relationships. The local spatial relationship characteristics and network structure characteristics of POIs are both indispensable parts of their spatial structural features. POIs, representing geographic entities, are fundamental elements constituting urban functional zones, and their spatial structural characteristics are a significant factor leading to differences in urban functional zone types.
[0069] It should be noted that the POI data within urban functional areas are discretely distributed points. Therefore, it is necessary to construct a POI network structure for each functional area, representing each functional area as a POI graph. This invention uses POIs as nodes, then constructs edges between nodes based on the spatial distance between POIs, and then constructs an adjacency matrix of POI nodes to store the connection relationships between nodes. Specifically, step S100 includes:
[0070] Step S110: Use the POIs of urban functional areas as nodes in the POI undirected graph;
[0071] Step S120: Select a preset number of POI nodes that are closest to the current POI node as the neighbors of the current POI node according to the KNN algorithm.
[0072] It should be understood that by taking the POIs of urban functional areas as nodes in an undirected graph of POIs, generating the neighbors of POIs based on their spatial distance, and establishing undirected edges between the current POI and its neighbors, an undirected graph can be formed.
[0073] There are two types of neighbor generation methods. The first uses the KNN algorithm to select the K nearest POIs as its neighbors. The second uses the current POI as the center, sets a search radius, and performs buffer analysis; POIs within the buffer are considered neighbors of the current POI. Since the spatial distribution density of different types of POIs varies significantly, this embodiment uses the first method to generate POI neighbors and establish edges between nodes. The value of K can be determined based on accuracy requirements, or an initial value can be set, such as K = 10.
[0074] In practice, the 10 nearest neighbors of the current POI node are connected to the current POI node, and so on, traversing all I PO nodes to form an undirected POI graph.
[0075] Step S200: Based on the preset Plane2vec model, take each POI node in the POI undirected graph and the POI points connected to it as input, and output the initialization vector of the POI type of the POI node.
[0076] It should be understood that, based on the local spatial relationships of POIs, a Plane2vec model is used to generate vector representations of POI types. Functional areas are treated as documents, and POI types as words. A corpus is generated based on the local spatial relationships of POIs, and then input into a Skip-Gram or CBOW model for training to generate embedded representations of POI types. The classification system of POIs in various electronic maps generally includes three levels of types. To ensure the richness of nodes and network structure, this embodiment uses the three-level POI type as the POI type (i.e., the node in the network). Another commonly used POI type embedding method is Plane2vec. In this embodiment, a Plane2vec model is used to embed POI types and use them as the initialization vectors for nodes.
[0077] In specific implementation, step S200 includes:
[0078] Step S210: Based on the preset Plane2vec model, the urban functional area is regarded as a document, the POI type is regarded as a word, each POI node in the POI undirected graph and the POI points connected to it are taken as input, and the initialization vector of the POI type of the POI node is output.
[0079] Step S300: Construct the network topology of a graph convolutional neural network based on the POI undirected graph;
[0080] It should be understood that, based on the POI undirected graph, the network topology of the graph convolutional neural network is constructed, that is, by using the POI undirected graph constructed in step S100 as the network topology of the graph convolutional neural network (GCN), and the POI adjacency matrix constructed in this step is denoted as A.
[0081] Step S400: Use the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network, learn the network structure features of the POI node, and obtain the probability value of the functional type of the functional area.
[0082] It should be noted that step S400 specifically includes:
[0083] Use the POI type vector generated in the previous step as the initialization vector of the POI node in the GCN; aggregate information of the POI node and its neighboring nodes in each layer of GCN, and pass information between GCN layers; add virtual nodes after the GCN layer, and propagate all node information to the virtual nodes to express the features of the graph, that is, the functional type probability value of the functional area.
[0084] The urban functional area classification method based on Platform2vec and graph convolutional neural networks provided by this invention first constructs an undirected graph of POIs based on POI data of urban functional areas. Then, according to a preset Platform2vec model, each POI node in the undirected graph and its connected POIs are used as input to output the initialization vector of the POI type of the POI node. Next, a network topology of a graph convolutional neural network is constructed based on the undirected POI graph. Finally, the initialization vector of the POI type is used as the initialization vector of the POI node in the network topology of the graph convolutional neural network to learn the network structure features of the POI nodes and obtain the probability value of the functional type of the functional area. In this way, the urban functional area classification method established by combining Platform2vec and graph convolutional neural network models can achieve the classification of urban functional areas by integrating the local spatial correlation features of POIs and the network topology features, and the expression accuracy is more precise.
[0085] Further, please refer to Figure 1 and Figure 2 The urban functional area classification method based on Plane2vec and graph convolutional neural networks provided in this invention is adopted. Figure 1 Considering a combination of residential and commercial uses, Figure 2 For a more mature residential function, this expression is more accurate.
[0086] like Figure 4 As shown, in the second embodiment of the urban functional area classification method based on Plane2vec and graph convolutional neural network provided by the present invention, step S400 includes:
[0087] Step S410: Use the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network;
[0088] Step S420: In each layer of the graph convolutional neural network network topology, information is aggregated for POI nodes and their connected nodes, and information is transferred between layers in the graph convolutional neural network network topology.
[0089] In specific implementation, step S420 includes:
[0090] Step S421: In the network topology of each layer of the graph convolutional neural network, information aggregation is performed on the POI nodes and their connected nodes.
[0091] Step S422: The aggregation result is passed as the new feature value of the POI node to the network topology of the next layer graph convolutional neural network.
[0092] The specific calculation process in step S420 includes:
[0093]
[0094] in,
[0095] H (l) The feature vector representing the POI node at layer l;
[0096] H(0) is the initialization vector X of the POI node;
[0097] W (l) The parameters are the convolution parameters of the l-th layer;
[0098] A is the POI adjacency matrix;
[0099] I is the identity matrix;
[0100] It is the degree matrix based on the POI network structure;
[0101] σ(·) is a non-linear activation operation.
[0102] Step S430: Point all POI nodes in the network topology of the graph convolutional neural network to a preset virtual node, and propagate all POI node information to the virtual node.
[0103] It should be understood that the positions of the preset virtual nodes can be pre-set or randomly generated; no specific restrictions are imposed here. All POI nodes in the graph are pointed to virtual nodes, and then a convolution operation is performed through a GCN layer to propagate all POI node information to the virtual nodes. The features of the virtual nodes are used to represent the features of the graph, which is the probability vector of the functional area type.
[0104] Step S440: Determine the probability vector of the urban functional area type based on the characteristics of the virtual node.
[0105] like Figure 5As shown, this invention provides a city functional area classification device based on Place2vec and graph convolutional neural networks. This device includes a construction unit 501, an initialization unit 502, a generation unit 503, and an output unit 504.
[0106] Construction unit 501 is used to construct an undirected graph of POIs based on the POI data of urban functional areas;
[0107] It should be understood that a POI (Point of Interest) is a point on an electronic map that represents a geographic entity. POI data is object-based data that can be directly used to characterize geographic entities with socioeconomic attributes, and it is characterized by easy accessibility, comprehensive data coverage, and high data integrity. POIs, representing geographic entities, are fundamental elements constituting urban functional zones, and their spatial structural characteristics are a significant factor leading to differences in urban functional zone types. From a sociological perspective, POIs, as carriers of urban social and economic functions, form an invisible social network based on their spatial relationships. The local spatial relationship characteristics and network structure characteristics of POIs are both indispensable parts of their spatial structural features. POIs, representing geographic entities, are fundamental elements constituting urban functional zones, and their spatial structural characteristics are a significant factor leading to differences in urban functional zone types.
[0108] It should be noted that the POI data within urban functional areas are discretely distributed points. Therefore, it is necessary to construct a POI network structure for each functional area, representing each functional area as a POI graph. This invention uses POIs as nodes, then constructs edges between nodes based on the spatial distance between POIs, and then constructs an adjacency matrix of POI nodes to store the connection relationships between nodes.
[0109] The initial unit 502 is used to take each POI node in the POI undirected graph and the POI points connected to it as input according to the preset Place2vec model, and output the initialization vector of the POI type of the POI node.
[0110] It should be understood that, based on the local spatial relationships of POIs, the Place2vec model is used to generate vector representations of POI types. Functional areas are treated as documents, and POI types as words. A corpus is generated based on the local spatial relationships of POIs, and then input into a Skip-Gram or CBOW model for training to generate embedded representations of POI types. The classification system of POIs in various electronic maps generally includes three levels of types. To ensure the richness of nodes and network structure, this embodiment uses the three-level POI type as the POI type (i.e., the node in the network). Another commonly used POI type embedding method is Place2vec. In this embodiment, the Place2vec model is used to embed POI types and uses them as the initialization vectors for nodes.
[0111] The generation unit 503 is used to construct the network topology of the graph convolutional neural network based on the POI undirected graph;
[0112] The output unit 504 is used to use the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network, learn the network structure features of the POI node, and obtain the probability value of the functional type of the functional area.
[0113] In practice, the POI type vector generated in the previous step is used as the initialization vector of the POI node in the GCN; information aggregation is performed on the POI node and its neighboring nodes in each layer of GCN, and information is transferred between GCN layers; virtual nodes are added after the GCN layer, and all node information is propagated to the virtual nodes to express the features of the graph, that is, the functional type probability value of the functional area.
[0114] The urban functional area classification device based on Platform2vec and graph convolutional neural networks provided by this invention first constructs an undirected graph of POIs based on POI data of urban functional areas. Then, according to a preset Platform2vec model, each POI node in the undirected graph and its connected POIs are taken as input, and the initialization vector of the POI type of the POI node is output. Next, a network topology of a graph convolutional neural network is constructed based on the undirected graph of POIs. Finally, the initialization vector of the POI type is used as the initialization vector of the POI node in the network topology of the graph convolutional neural network to learn the network structure features of the POI nodes and obtain the probability value of the functional type of the functional area. In this way, the urban functional area classification method established by combining Platform2vec and graph convolutional neural network models can achieve the classification of urban functional areas by integrating the local spatial correlation features and network topology features of POIs, and the expression accuracy is more precise.
[0115] To achieve the above objectives, the present invention also provides an electronic device, such as... Figure 6As shown, the electronic device includes at least one processor 601; and a memory 602 communicatively connected to the at least one processor 601; wherein the memory 602 stores instructions executable by the at least one processor 601, the instructions being executed by the at least one processor 601 to enable the at least one processor 601 to execute an urban functional area classification method based on Plane2vec and graph convolutional neural networks.
[0116] The memory 602 and processor 601 are connected via a bus, which may include any number of interconnecting buses and bridges. The bus connects various circuits of one or more processors 601 and memory 602 together. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 601 is transmitted over a wireless medium via an antenna, which further receives data and transmits it to processor 601.
[0117] Processor 601 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory 602 can be used to store data used by processor 601 during operation.
[0118] To achieve the above objectives, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described urban functional area classification method based on Plane2vec and graph convolutional neural networks.
[0119] That is, those skilled in the art will understand that all or part of the steps in the methods described above can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0120] Obviously, the embodiments described above are merely some, not all, embodiments of the present invention. Based on the embodiments of the present invention, those skilled in the art can make other variations or modifications without creative effort, and all such variations or modifications should fall within the scope of protection of the present invention.
Claims
1. A method for classifying urban functional zones based on Place2vec and graph convolutional neural networks, characterized in that, include: Based on the POI data of urban functional areas, construct an undirected POI graph, including using the POIs of urban functional areas as nodes of the undirected POI graph, and generating the neighbors of the current POI node based on the spatial distance of the POIs. Connect the current POI node with its neighbors to form an undirected POI graph; Based on the preset Place2vec model, each POI node in the POI undirected graph and its connected POI nodes are taken as input, and the initialization vector of the POI type of the POI node is output. Based on the POI undirected graph, construct the network topology of the graph convolutional neural network; Using the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network, learning the network structure features of the POI node, and obtaining the probability value of the functional type of the functional area, includes: using the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network. In each layer of the graph convolutional neural network network topology, information is aggregated for POI nodes and their connected nodes, and information is transferred between layers in the graph convolutional neural network network topology; all POI nodes in the graph convolutional neural network network topology are pointed to a preset virtual node, and all POI node information is propagated to the virtual node; based on the characteristics of the virtual node, the probability vector of the urban functional area type is determined.
2. The urban functional area classification method based on Place2vec and graph convolutional neural networks as described in claim 1, characterized in that, The steps of aggregating information about Point of Interest (POI) nodes and their connected nodes in each layer of the graph convolutional neural network topology, and transferring information between layers in the graph convolutional neural network topology, include: In the network topology of each layer of the graph convolutional neural network, information is aggregated for POI nodes and their connected nodes; The aggregation result is passed as a new feature value of the POI node to the network topology of the next layer of the graph convolutional neural network.
3. The urban functional area classification method based on Place2vec and graph convolutional neural networks as described in claim 2, characterized in that, The calculation process for aggregating information about POI nodes and their connected nodes in each layer of the graph convolutional neural network topology, and for transferring information between layers in the graph convolutional neural network topology, is as follows: ；（1） in, H (l) The feature vector representing the POI node at layer l; H(0) is the initialization vector X of the POI node; W (l) The parameters are the convolution parameters of the l-th layer; A is the POI adjacency matrix; I is the identity matrix; It is the degree matrix based on the POI network structure; σ(·) is a non-linear activation operation.
4. The urban functional area classification method based on Place2vec and graph convolutional neural networks as described in claim 1, characterized in that, The step of using POIs of urban functional areas as nodes in an undirected POI graph and generating neighbors for the current POI node based on the spatial distance of the POIs includes: Use the POIs of urban functional areas as nodes in an undirected POI graph; The KNN algorithm selects a preset number of POI nodes that are closest to the current POI node as the current POI node's neighbors.
5. The urban functional area classification method based on Place2vec and graph convolutional neural networks as described in claim 1, characterized in that, The step of taking each POI node and its connected POI nodes in the undirected POI graph as input, and outputting the initialization vector of the POI type of the POI node according to the preset Place2vec model, includes: Based on the preset Place2vec model, urban functional areas are regarded as documents and POI types are regarded as words. Each POI node in the undirected graph of POI and its connected POI points are taken as input, and the initialization vector of the POI type of the POI node is output.
6. A city functional area classification device based on Place2vec and graph convolutional neural networks, characterized in that, include: The building unit is used to construct an undirected graph of POIs based on the POI data of urban functional areas; The initial unit is used to take each POI node and its connected POI points in the POI undirected graph as input according to the preset Place2vec model, and output the initialization vector of the POI type of the POI node. The generation unit is used to construct the network topology of the graph convolutional neural network based on the POI undirected graph; The output unit is used to use the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network, learn the network structure features of the POI node, and obtain the probability value of the functional type of the functional area, including: using the initialization vector of the POI type as the initialization vector of the POI node in the network topology of the graph convolutional neural network. In each layer of the graph convolutional neural network network topology, information is aggregated for POI nodes and their connected nodes, and information is transferred between layers in the graph convolutional neural network network topology; all POI nodes in the graph convolutional neural network network topology are pointed to a preset virtual node, and all POI node information is propagated to the virtual node; based on the characteristics of the virtual node, the probability vector of the urban functional area type is determined.
7. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the urban functional area classification method based on Place2vec and graph convolutional neural networks as described in any one of claims 1 to 5.
8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the urban functional area classification method based on Place2vec and graph convolutional neural network as described in any one of claims 1 to 5.