Network site selection prediction method, device, equipment, medium and program product

Abstract

The present disclosure provides a network site selection prediction method, which can be applied to the field of artificial intelligence technology or the financial field. The method includes: obtaining signaling data of a target area; storing the signaling data of the target area in a graph database, and constructing a network site selection information association graph; segmenting the network site selection information association graph based on a preset distance granularity, and obtaining m local network site selection information association graph data, where m is an integer greater than or equal to 2; and pre-processing the local network site selection information association graph and inputting it into a pre-trained network site selection prediction model to obtain a network site selection prediction result, wherein the network site selection prediction result includes a business level prediction rating of the target area, and the pre-trained network site selection prediction model is constructed based on a graph convolutional neural network. The present disclosure also provides a network site selection prediction device, equipment, storage medium and program product.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology or the financial field, and specifically to a network site selection prediction method, device, equipment, medium and program product. Background Art

[0002] With the continuous development of social economy, the competition among banks is becoming increasingly fierce. Traditional bank physical outlets are one of the important sales channels and the most intuitive way to showcase the bank's own brand image. The location of bank outlets is directly related to the development and economic benefits of banks. Choosing a scientific and reasonable outlet address has become an important condition for banks to gain advantages in the competition in the financial market, and it is of great significance to the development of banks and the improvement of their core competitiveness. Therefore, it is very important to establish a fast, accurate and automated bank outlet location evaluation and recommendation method to seize the regional financial market and compete for customer resources. There are two main methods for the existing bank outlet location selection methods. The first is the manual location selection method, and the second is to use signaling data to construct a location selection model according to a specified weight formula or by establishing a logistic regression machine learning method.

[0003] In the process of realizing the concept of the present disclosure, the inventors found that there are at least the following problems in the prior art: the first manual site selection method, through manual collection of data such as human traffic and traffic, and then experts use these data to establish a site selection model to select the address of the branch, which takes a long time and the survey results have large errors. This kind of site selection is time-consuming and labor-intensive, inefficient, and may have many unreasonable places, which will not only lead to duplication and waste of resources, but also make the bank's income poor. The second method uses signaling data to construct a site selection model according to a specified weight formula or by establishing a logistic regression machine learning method. Although it solves the problem of high labor costs of traditional site selection to a certain extent. However, the specified weight formula is often set by human experience, and the influence of subjective judgment may cause the site selection results to be inaccurate. By establishing a simple logistic regression machine learning model, it is not possible to fully explore the correlation information between feature data, which will also affect the accuracy of branch site selection. Summary of the invention

[0004] In view of the above problems, the embodiments of the present disclosure provide a method, device, equipment, medium and program product for improving the accuracy and scientificity of network site selection prediction.

[0005] According to the first aspect of the present disclosure, there is provided a network point location prediction method, comprising: obtaining signaling data of a target area; storing the signaling data of the target area into a graph database, and constructing a network point location information association graph; segmenting the network point location information association graph based on a preset distance granularity, and obtaining m local network point location information association graph data, wherein m is an integer greater than or equal to 2; and preprocessing the local network point location information association graph and inputting it into a pre-trained network point location prediction model to obtain a network point location prediction result, wherein the network point location prediction result includes a service level prediction rating of the target area, and the pre-trained network point location prediction model is constructed based on a graph convolutional neural network.

[0006] According to an embodiment of the present disclosure, the signaling data includes at least two of economic data, geographic data, physical condition data, and industry competition data.

[0007] According to an embodiment of the present disclosure, the construction of a network point location information association graph includes: taking the target area as the core node, taking the attributes of the target area as the associated nodes, and storing the target area and the attributes of the target area in a graph database, wherein the graph database contains stored existing network point core nodes and existing network point attribute associated nodes; the attributes correspond to signaling data types; wherein a mapping relationship exists between the i-th core node and m associated nodes, wherein i satisfies 1≤i≤n and i is an integer, n is the total number of nodes in the network point location information association graph, and m is the number of signaling data types.

[0008] According to an embodiment of the present disclosure, the graph database also includes stored service level data of existing outlets, and the preprocessing of the local outlet location information association graph includes: normalizing the signaling data of the ith core node to obtain the feature vector of the ith core node, wherein the ith core node feature vector includes m features, and the m features are obtained based on associated nodes that have a mapping relationship with the ith core node; obtaining an edge feature vector based on the distance between the ith core node and the remaining n-1 core nodes, and obtaining an adjacency matrix of the ith core node based on the edge feature vector; marking the existing outlet core node data as service level data; and obtaining local outlet location information association graph data based on the feature vector of the core node, the adjacency matrix of the core node, and the existing outlet core node data.

[0009] According to an embodiment of the present disclosure, the method of obtaining the prediction results of network point location selection also includes: obtaining the business level prediction ratings of k target areas, k being an integer greater than 1; sorting the business level prediction ratings of the k target areas in descending order, and taking the target area ranked first as the optimal network point prediction location selection.

[0010] According to an embodiment of the present disclosure, the business data includes at least two of business transaction data, customer growth data, and customer flow data, and the business data is used to obtain business level data of existing outlets.

[0011] According to an embodiment of the present disclosure, the network site selection prediction model includes a first graph convolutional network layer, a second graph convolutional network layer, a graph feature averaging layer, a first fully connected layer, a second fully connected layer and a classification network layer.

[0012] According to an embodiment of the present disclosure, the method of obtaining the prediction result of network point location selection includes: inputting the local network point location selection information association graph data into the first graph convolutional network layer to obtain first-order node association feature information; inputting the first-order node association feature information into the second graph convolutional network layer to obtain second-order node association feature information; inputting the second-order node association feature information into the graph feature averaging layer to obtain graph average feature information; inputting the graph average feature information into the first fully connected layer to obtain first global feature information; inputting the first global feature information into the second fully connected layer after nonlinear activation transformation to obtain second global feature information; and inputting the second global feature information into the classification network layer to obtain the target area business level prediction result.

[0013] According to an embodiment of the present disclosure, when pre-training the network point location prediction model, back propagation and gradient descent optimization algorithms are used to update the parameters of the network point location prediction model to be trained until a preset training cutoff condition is reached.

[0014] The second aspect of the present disclosure provides a network point location prediction device, including: a first acquisition module, configured to acquire signaling data of a target area; a first calculation module, configured to store the signaling data of the target area into a graph database, and construct a network point location information association graph; a second calculation module, configured to segment the network point location information association graph based on a preset distance granularity, and acquire m local network point location information association graph data, wherein m is an integer greater than or equal to 2; and a prediction module, configured to pre-process the local network point location information association graph and input it into a pre-trained network point location prediction model to acquire a network point location prediction result, wherein the network point location prediction result includes a predicted rating of a service level of the target area, and the pre-trained network point location prediction model is constructed based on a graph convolutional neural network.

[0015] According to an embodiment of the present disclosure, the first computing module further includes a storage submodule, which is configured to store the target area and the attributes of the target area in a graph database, taking the target area as a core node and the attributes of the target area as an associated node. The graph database includes stored existing network point core nodes and existing network point attribute associated nodes; the attributes correspond to the types of signaling data; the i-th core node has a mapping relationship with m associated nodes, wherein i satisfies 1≤i≤n and i is an integer, n is the total number of nodes in the network point location information association graph, and m is the number of signaling data types.

[0016] According to an embodiment of the present disclosure, the prediction module includes a first convolution submodule, a second convolution submodule, a feature average submodule, a first comprehensive submodule, a second comprehensive submodule and a classification submodule. Among them, the first convolution submodule is configured to input all m local network point location information association graph data into the first graph convolution network layer to obtain first-order node association feature information. The second convolution submodule is configured to input the first-order node association feature information into the second graph convolution network layer to obtain second-order node association feature information. The feature average submodule is configured to input the second-order node association feature information into the graph feature average layer to obtain graph average feature information. The first comprehensive submodule is configured to input the graph average feature information into the first fully connected layer to obtain the first global feature information. The second comprehensive submodule is configured to input the first global feature information into the second fully connected layer after nonlinear activation transformation to obtain the second global feature information. The classification submodule is configured to input the second global feature information into the classification network layer to obtain the network point business level prediction result.

[0017] According to an embodiment of the present disclosure, the graph database also includes the stored service level data of existing outlets, and the prediction module also includes a normalization submodule, a vectorization submodule, a marking submodule, and a data summary submodule. Among them, the normalization submodule is configured to normalize the signaling data of the i-th core node and obtain the feature vector of the i-th core node, wherein the i-th core node feature vector contains m features, and the m features are obtained based on the associated nodes that have a mapping relationship with the i-th core node. The vectorization submodule is configured to obtain the edge feature vector based on the distance between the i-th core node and the remaining n-1 core nodes, and obtain the adjacency matrix of the i-th core node based on the edge feature vector. The marking submodule is configured to mark the existing outlet core node data as service level data. The data summary submodule is configured to obtain the local outlet location information association graph data based on the feature vector of the core node, the adjacency matrix of the core node, and the existing outlet core node data.

[0018] According to an embodiment of the present disclosure, the network point location prediction device further includes a sorting module. The sorting module is configured to obtain the business level prediction ratings of k target areas, where k is an integer greater than 1. The business level prediction ratings of the k target areas are sorted in descending order, and the target area ranked first is used as the optimal network point prediction location.

[0019] The third aspect of the present disclosure provides an electronic device, comprising: one or more processors; a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above-mentioned network site selection prediction method.

[0020] The fourth aspect of the present disclosure further provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the above-mentioned network site selection prediction method.

[0021] The fifth aspect of the present disclosure also provides a computer program product, including a computer program, which implements the above-mentioned network site selection prediction method when executed by a processor.

[0022] The method provided by the embodiments of the present disclosure builds a branch location prediction model through a graph neural network, fully explores the correlation between feature data, overcomes the problems of low efficiency and low accuracy of traditional bank branch location selection, and provides a more scientific and accurate branch location prediction method. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The above contents and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0024] Figure 1 The application scenario diagram of the network site selection prediction method, device, equipment, medium and program product according to the embodiments of the present disclosure is schematically shown.

[0025] Figure 2 The flowchart of the network point location prediction method according to the embodiment of the present disclosure is schematically shown.

[0026] Figure 3 The flowchart of the method for constructing the network point location information association diagram according to the embodiment of the present disclosure is schematically shown.

[0027] Figure 4 The flowchart of the method for preprocessing the local network point location information association graph according to the embodiment of the present disclosure is schematically shown.

[0028] Figure 5 The flowchart of the method for obtaining network point location prediction results according to an embodiment of the present disclosure is schematically shown.

[0029] Figure 6 The flowchart of the method for obtaining the network point location prediction result based on the network point location prediction model according to the embodiment of the present disclosure is schematically shown.

[0030] Figure 7 The structural block diagram of the network point location prediction device according to an embodiment of the present disclosure is schematically shown.

[0031] Figure 8 The structural block diagram of the first computing module 720 according to an embodiment of the present disclosure is schematically shown.

[0032] Fig. 9 The structural block diagram of the prediction module 740 according to an embodiment of the present disclosure is schematically shown.

[0033] Fig.10 The structural block diagram of some other prediction modules 740 according to the embodiment of the present disclosure is schematically shown.

[0034] Fig.11 The structural block diagram of the network site selection prediction device 700 according to some other embodiments of the present disclosure is schematically shown.

[0035] Fig.12 A block diagram of an electronic device suitable for implementing a network site selection prediction method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0036] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0037] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0038] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0039] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0040] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision, disclosure and application of user personal information involved comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

[0041] In the technical solution of the present disclosure, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0042] With the continuous development of social economy, the competition among banks is becoming increasingly fierce. Traditional bank physical outlets are one of the important sales channels and the most intuitive way to display the bank's own brand image. The location of bank outlets is directly related to the development and economic benefits of banks. Choosing a scientific and reasonable outlet address has become an important condition for banks to gain advantages in the competition in the financial market, and is of great significance to the development of banks and the improvement of their core competitiveness. Therefore, it is very important to establish a fast, accurate and automated bank outlet location evaluation and recommendation method for seizing regional financial markets and competing for customer resources. There are two main methods for the existing bank outlet location selection methods. The first is the manual location selection method, and the second is to use signaling data to build a location selection model according to a specified weight formula or by establishing a logistic regression machine learning method. The first manual location selection method manually collects data such as traffic flow and transportation, and then experts use these data to establish a location selection model to select the outlet address. It takes a long time and the survey results have large errors. This location selection is time-consuming, labor-intensive, and inefficient. There may be many unreasonable places, which will not only lead to duplication and waste of resources, but also poor bank returns. The second method is to use signaling data to build a site selection model according to a specified weight formula or by establishing a logistic regression machine learning method. This can mine the relationship between signaling data and the value of banking business, and to a certain extent solve the problem of high labor costs in traditional site selection. However, the specified weight formula is often set based on human experience, and the influence of subjective judgment may lead to inaccurate site selection results. By establishing a simple logistic regression machine learning model, the correlation information between feature data cannot be fully mined, which will also affect the accuracy of branch site selection prediction.

[0043] Graph structured data is a more complex data structure than linear lists and trees. In a linear list, data elements are strung together and have only linear relationships. Each data element has only one direct predecessor and one direct successor. In a tree structure, data elements have a clear hierarchical relationship, and data elements on each layer may be related to multiple elements in the next layer, but can only be related to one element in the previous layer. In a graph structure, the relationship between nodes can be arbitrary, and any two data elements in the graph may be related. Graph convolutional neural network is a variant of convolutional neural network on graph data, which learns the representation of graph data by stacking several first-order spectral filters in front of nonlinear functions. Graph convolutional neural network can naturally integrate the attribute information of the graph for learning and mine the associations between the nodes of each graph. In the process of bank branch site selection, a large amount of unstructured signaling data needs to be collected. These unstructured signaling data are unstructured, disordered and random. The embodiments of the present disclosure find that the branch signaling data, business level and other feature data are stored in the form of a graph structure, and a site selection model is constructed using a graph neural network. This can make full use of the advantage of the graph convolutional neural network in mining features without being restricted to the dependency relationship on the two-dimensional structure, and more abundantly mining the correlation information between feature data to fully mine the correlation relationship between feature information such as the signaling data of each branch, solve the problem of data information islands, and improve the accuracy of branch site selection prediction.

[0044] An embodiment of the present disclosure provides a network point location prediction method, including obtaining signaling data of a target area; storing the signaling data of the target area into a graph database to construct a network point location information association graph; segmenting the network point location information association graph based on a preset distance granularity to obtain m local network point location information association graph data, wherein m is an integer greater than or equal to 2; pre-processing the local network point location information association graph and inputting it into a pre-trained network point location prediction model to obtain a network point location prediction result, wherein the network point location prediction result includes a business level prediction rating of the target area, and the pre-trained network point location prediction model is constructed based on a graph convolutional neural network.

[0045] It should be noted that the network site selection prediction method, device, equipment, medium and program product provided in the embodiments of the present disclosure can be used in the application of artificial intelligence technology in graph neural networks, and can also be used in various fields other than artificial intelligence technology, such as the financial field. The application field of the network site selection prediction method, device, equipment, medium and program product provided in the embodiments of the present disclosure is not limited.

[0046] The above operations for achieving at least one purpose of the present disclosure will be described below in conjunction with the accompanying drawings and their explanatory texts.

[0047] Figure 1The application scenario diagram of the network site selection prediction method, device, equipment, medium and program product according to the embodiments of the present disclosure is schematically shown.

[0048] like Figure 1 As shown, the application scenario 100 according to this embodiment may include a terminal device 101, a terminal device 102, a terminal device 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0049] The user can use the terminal device 101, the terminal device 102, and the terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, the terminal device 102, and the terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0050] Terminal device 101, terminal device 102, and terminal device 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers, etc. Terminal device 101, terminal device 102, and terminal device 103 can send signaling data to server 105 and receive network site selection prediction results.

[0051] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the terminal devices 101, 102, and 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0052] It should be noted that the network site selection prediction method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the network site selection prediction device provided in the embodiment of the present disclosure can generally be set in the server 105. The network site selection prediction method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal device 101, the terminal device 102, the terminal device 103 and / or the server 105. Accordingly, the network site selection prediction device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal device 101, the terminal device 102, the terminal device 103 and / or the server 105.

[0053] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to the implementation requirements.

[0054] The following will be based on Figure 1 The scene described by Figure 2~Figure 6 The network site selection prediction method of the disclosed embodiment is described in detail.

[0055] Figure 2 The flowchart of the network point location prediction method according to the embodiment of the present disclosure is schematically shown.

[0056] like Figure 2 As shown, the network site selection prediction method of this embodiment includes operations S210 to S240. The transaction processing method can be executed by a processor or any electronic device including a processor.

[0057] In operation S210, signaling data of a target area is acquired.

[0058] According to an embodiment of the present disclosure, the target area may be a pre-selected candidate area that may be used as a location for the network point. The target area may be pre-selected based on preliminary research or expert experience. The method of selecting the target area is not within the scope of discussion of the embodiments of the present disclosure. Signaling data is data associated with the service level of the network point, which can be obtained through communication companies, government departments, cooperative data companies and other units, or through preliminary visits and surveys. It should be noted that in an embodiment of the present disclosure, when the signaling data includes user information, the user's consent or authorization may be obtained before obtaining the user information. For example, before operation S210, a request to obtain user information may be issued to the user. When the user agrees or authorizes that the user information can be obtained, the operation S210 is performed.

[0059] In operation S220, the signaling data of the target area is stored in a graph database to construct a network point location information association graph.

[0060] According to an embodiment of the present disclosure, the collected signaling data can be stored in a graph database to construct an association graph of network site selection information. A typical graph database may include Neo4J. The storage process of the graph may include: reading signaling data, field names corresponding to node types, field values ​​corresponding to nodes, establishing associations with target areas or existing network points as core nodes, and node attributes (such as signaling data) as slave nodes, and storing them in a graph database. When storing new node data, query whether the node already exists. If it already exists, the data is not stored repeatedly in the graph database. According to the above processing process, the target area and all existing network point data are stored in the graph data for subsequent use. It can be understood that when using the embodiment of the present disclosure to evaluate the target area to obtain the network site selection prediction results, the graph database contains the attribute data of the stored existing network points, wherein the attribute data of the existing network points includes the signaling data of the existing network points.

[0061] In operation S230, the network point location information association map is segmented based on a preset distance granularity to obtain m local network point location information association map data, where m is an integer greater than or equal to 2.

[0062] According to the embodiments of the present disclosure, in order to fully explore the association relationship between core nodes and the impact of node characteristics on node service levels, the network point location information association map can be segmented based on a preset distance granularity. The preset distance granularity can be set based on expert experience. For example, the network point location information association map can be segmented with a core node as the center and a radius of 5 kilometers to obtain a local network point location information association map, so as to fully explore the impact of other network points and network point attribute data within 5 kilometers on the service level of the network point to be evaluated.

[0063] In operation S240, the local network site selection information association graph is pre-processed and input into a pre-trained network site selection prediction model to obtain a network site selection prediction result, wherein the network site selection prediction result includes a predicted rating of the service level of the target area, and the pre-trained network site selection prediction model is constructed based on a graph convolutional neural network.

[0064] According to an embodiment of the present disclosure, in order to facilitate the network site selection prediction model to read and process data, the original unstructured signaling data can be preprocessed to convert it into a form that can be input by the model. In the processing process of the graph data disclosed in the present disclosure, node vector conversion and adjacency matrix acquisition can be performed. Among them, vector conversion is used to vectorize node features, and the adjacency matrix is ​​used to represent the association between the edges composed of core nodes. The network site selection prediction model is constructed based on a graph convolutional neural network. It should be understood that when the preprocessed local network site selection information association graph is input into the network site selection prediction model, the existing network points in the core node also have associated signaling data, and the existing network points also contain business level labels, so that the target area business prediction rating can be predicted after model processing to obtain the network site selection prediction result. In the embodiment of the present disclosure, the unstructured, disordered, and random target area signaling data is stored as a graph data structure, and the graph convolutional network is used to break through the dependency relationship of the two-dimensional structure, and more richly mine the association information between the feature data. Converting the network site selection prediction into a classification problem and obtaining the target area business forecast rating can enable more scientific and accurate network site selection deployment than traditional methods.

[0065] The embodiments of the present disclosure find that, based on the consideration of data availability, accessibility, effectiveness of branch business level prediction, and data processing efficiency, the signaling data may include at least two of economic data, geographic data, physical condition data, and peer competition data. Among them, the economic data includes signaling data related to the economic development level of the target area or existing branches. The geographic data includes signaling data related to the geographical location of the target area or existing branches. The physical condition data includes signaling data related to the physical attributes, infrastructure conditions, etc. of the target area or existing branches, and the peer competition data includes data of other bank branches that have peer competition relationships with the target area or existing branches. Preferably, the signaling data may include the surrounding economic volume B, the flow of people V, the age structure Y, the per capita income I, the number of surrounding permanent residents R, the number of takeaway orders per person per day T, the number of surrounding elevator communities F, the number of surrounding communities without elevators N, the number of parking spaces in front of the door P, the bus data and subway data L within 300 meters, the geographical location of the target area or outlet G, the newness of the target area or outlet O, the outdoor advertising effect K, the annual rent of the target area or outlet M, the area of ​​the target area or outlet C, the internal facilities of the target area or outlet D, the total salary of employees in the target area or outlet E, the number of surrounding outlets of the same industry H, etc. Among them, the surrounding area can be defined based on a preset geographical range, such as a specified number of blocks; or the geographical range can be defined with the outlet or target area as the center and a preset distance as the radius.

[0066] Figure 3 The flowchart of the method for constructing the network point location information association diagram according to the embodiment of the present disclosure is schematically shown.

[0067] like Figure 3 As shown, the method for constructing a network point location information association graph in this embodiment includes operation S310.

[0068] In operation S310, the target area is taken as the core node, and the attribute of the target area is taken as the associated node, and the target area and the attribute of the target area are stored in a graph database. Wherein, the graph database contains the stored existing network core nodes and existing network attribute associated nodes. The attribute corresponds to the type of signaling data. For example, each type of signaling data is an attribute. Wherein, there is a mapping relationship between the i-th core node and the m associated nodes. Wherein, i satisfies 1≤i≤n and i is an integer, n is the total number of nodes in the network site selection information association graph, and m is the number of signaling data types. Thus, each core node forms a one-to-many association relationship with the signaling data related to it. In an embodiment of the present disclosure, i can traverse values ​​from 1 to n to obtain a network site selection information association graph for expressing the association relationships of all nodes.

[0069] According to an embodiment of the present disclosure, the graph database also includes stored service level data of existing outlets. When applying the outlet site selection prediction model of the embodiment of the present disclosure, the service level of the target area can be predicted by mining the association between the service level data of existing outlets and the signaling data of existing outlets.

[0070] Figure 4 The flowchart of the method for preprocessing the local network point location information association graph according to the embodiment of the present disclosure is schematically shown.

[0071] like Figure 4 As shown, the method for preprocessing the local network point location information association graph in this embodiment includes operations S410 to S440.

[0072] In operation S410, the signaling data of the ith core node is normalized to obtain a feature vector of the ith core node, wherein the feature vector of the ith core node includes m features, and the m features are obtained based on associated nodes that have a mapping relationship with the ith core node.

[0073] In the embodiments of the present disclosure, in order to avoid the problem of difficult convergence caused by excessive numerical differences between signaling data, the preprocessing method may include a step of normalizing the signaling data. Specifically, the signaling data may be normalized using Z-score normalization. As shown in formula (1):

[0074] (1)

[0075] in, is the value of signaling data X, Signaling data The mean of Signaling data The standard deviation of . It should be understood that before normalizing the signaling data, feature engineering can also be used to clean the signaling data, including mean filling and abnormal sample deletion. Assigning meaning to feature values, for example, the geographical location G of the target area or network point can be represented based on longitude and latitude, the newness O of the target area or network point can be represented by a value, the outdoor advertising effect K can be represented by a value, and the internal facilities D of the target area or network point can be represented by counting the number of internal facility types or presetting the number of internal facility types to achieve feature representation.

[0076] In operation S420, an edge feature vector is obtained based on the distance between the ith core node and the remaining n-1 core nodes, and an adjacency matrix of the ith core node is obtained based on the edge feature vector.

[0077] In the embodiment of the present disclosure, in order to explore the relationship between other outlets and the target area that may have an impact on the service level of the target area due to their geographical location, the edge feature vector can be obtained by the distance between the i-th core node and the remaining n-1 core nodes, and the adjacency matrix can be further constructed. Exemplarily, the outlet location information association graph of the embodiment of the present disclosure is an undirected graph G=(V, E). In the undirected graph G, there is a node v x ∈V, edge (v i , v j )∈E, where i,j∈x. The adjacency matrix of the core node can be shown as formula (2):

[0078]

[0079] Formula (2)

[0080] Among them, Aij represents the values ​​of nodes i and j in the adjacency matrix, vi represents node i, vj represents node j, and E represents the edge set. In this way, the influence of the distance between nodes on the node business level can be explored.

[0081] In operation S430, the existing network point core node data is marked as business level data. According to an embodiment of the present disclosure, the business level data of the existing network point in the core node can be marked as a core node value. Thus, the normalized data of each slave node can correspond to the feature vector [B, V, Y... D, E, H] of the core node, and the distance between a core node and another core node can be used as the feature vector [S] of each edge in the graph data, and the feature vector [S] is converted into an adjacency matrix of the core node, and the business level data of each existing network point core node data in each local graph is marked.

[0082] In operation S440, local network point location information association graph data is obtained based on the feature vector of the core node, the adjacency matrix of the core node, and the existing network point core node data. It can be understood that the local network point location information association graph data includes the feature vector of the core node, the adjacency matrix of the core node, and the service level data of the existing network point core node. After being processed by the network point prediction model, the local network point location information association graph data can output the target area service level prediction rating result.

[0083] In some specific embodiments, the business data includes at least two of business transaction data, customer growth data, and customer flow data. The business level of existing outlets can be evaluated based on the business data, and the business level data of the existing outlets can be obtained. For example, the business level of the outlets can be scored on a percentage basis, and 5 points can be used as one level, and the business level of the outlets can be divided into 20 levels. Among them, the scores corresponding to the levels can be flexibly set based on expert experience and business needs.

[0084] According to the embodiments of the present disclosure, the candidate target areas may include multiple ones. Accordingly, after obtaining the service level prediction rating results of the multiple target areas, further screening may be performed to obtain the optimal predicted site selection location of the network point.

[0085] Figure 5 The flowchart of the method for obtaining network point location prediction results according to an embodiment of the present disclosure is schematically shown.

[0086] like Figure 5 As shown, the method for obtaining network site selection prediction results in this embodiment includes operations S510 to S520.

[0087] In operation S510, predicted ratings of business levels of k target areas are obtained, where k is an integer greater than 1.

[0088] In operation S520, the predicted service levels of the k target areas are ranked from high to low, and the first ranked target area is selected as the optimal network point prediction location. It should be understood that k is the total number of candidate target areas.

[0089] In an embodiment of the present disclosure, a network site selection prediction model is constructed based on a graph convolutional neural network, including a first graph convolutional network layer, a second graph convolutional network layer, a graph feature averaging layer, a first fully connected layer, a second fully connected layer and a classification network layer.

[0090] Figure 6 The flowchart of the method for obtaining the network point location prediction result based on the network point location prediction model according to the embodiment of the present disclosure is schematically shown.

[0091] like Figure 6As shown, the method for obtaining a network point location prediction result based on a network point location prediction model in this embodiment includes operations S610 to S660.

[0092] In operation S610, the local network point location information association graph data is input into the first graph convolutional network layer to obtain first-order node association feature information.

[0093] In operation S620, the first-order node association feature information is input into the second graph convolutional network layer to obtain second-order node association feature information.

[0094] In operation S630, the second-order node association feature information is input into a graph feature averaging layer to obtain graph average feature information.

[0095] In operation S640, the graph average feature information is input into a first fully connected layer to obtain first global feature information.

[0096] In operation S650, the first global feature information is input into a second fully connected layer after being transformed by nonlinear activation to obtain second global feature information.

[0097] In operation S660, the second global feature information is input into a classification network layer to obtain a target area service level prediction result.

[0098] In an embodiment of the present disclosure, the local network point location information association graph data is a test set of local network point location information association graph data containing the target area. The model first extracts the first-order and second-order node association feature information through two layers of graph convolutional network layers respectively. Since the number of nodes in different local network point location information association graphs is different, the second-order node association feature information can be averaged based on the graph feature averaging layer. Further, the averaged graph average feature information is used to synthesize the global features using two layers of fully connected layers. Among them, after the first global feature information is obtained through the first layer of fully connected layer, the first global feature information can be nonlinearly activated by using the ReLU function. The second global feature information obtained after processing by the two layers of fully connected layers is input into the classification network layer, and the probability of the target area being predicted as each business level rating is calculated after the second global feature information is linearly transformed using the classifier of the classification layer to obtain the classification result, that is, the target area business level prediction result.

[0099] It should be understood that before applying the model, a portion of data containing only existing network signaling data and service levels can be used to construct and partially segment the sample network site selection information association map, and the segmented local map data samples can be preprocessed to construct a training map data set, and another portion of sample data containing only existing network signaling data and service level data can be input as a verification map data set for pre-training of the network site selection prediction model to be trained. The construction, segmentation and pre-processing methods of the map data are described above and will not be repeated here. The ratio of training map data to verification map data can be preset based on expert experience.

[0100] According to an embodiment of the present disclosure, when pre-training the network site selection prediction model, back propagation and gradient descent optimization algorithms are used to update the parameters of the network site selection prediction model to be trained until a preset training cutoff condition is reached. In some specific embodiments, a Nadam gradient descent optimization algorithm can be used to improve data processing efficiency.

[0101] In some specific embodiments, when training a model, the training graph data set can be input into the model to be trained in batches. The data table conversion process is as follows: For an undirected graph G=(V,E), each node ∈V, each edge ( , )∈E, where i, j∈x. The hidden state of each node is defined as , the neighbor set of each node is N( ), for each node in the graph , the information propagation formula in the model is shown in formula (3):

[0102] (3)

[0103] Where c is the number of graph neural network layers, is the feature of neighbor node u at layer c, is the number of neighbor nodes, is the weight vector on layer c, is the bias term on layer c, and ReLU is the activation function.

[0104] The features of each core node are obtained through the graph neural network layer, and the final task is to classify the graph data. Since the number of core nodes of each local graph in the embodiment of the present disclosure may be different, the graph features read out after graph convolution are averaged using the graph feature averaging layer to obtain the average graph features. , as shown in formula (4):

[0105] (4)

[0106] in, is the graph feature vector of neighbor node u, is the number of nodes.

[0107] The average graph features are transformed through the fully connected layer and then input into the classification network layer for business level classification, as shown in formula (5):

[0108] (5)

[0109] in, To predict the probability of belonging to category (i.e., business level) i, is the number of categories, is the weight vector, is the bias term, is the average graph feature.

[0110] In an embodiment of the present disclosure, the classification network layer may include a Softmax layer. A cross entropy loss function may be used during training. The training cutoff condition may be to set a preset training round or to achieve a preset model accuracy. During the training process, by adjusting the number of graph neural network layers and fully connected layers, and adjusting the distance granularity of the segmentation of the local network site selection information association graph, it was found that using 5Km as the distance granularity for segmentation, the model uses 2 graph neural network layers, 2 fully connected layers, and using ReLU transformation after the first fully connected layer can obtain a higher model accuracy, so as to improve the scientificity and accuracy of the network site selection prediction.

[0111] Based on the above-mentioned network site selection prediction method, the present disclosure also provides a network site selection prediction device. Figure 7 to Figure 11 The device is described in detail.

[0112] Figure 7 The structural block diagram of the network point location prediction device according to an embodiment of the present disclosure is schematically shown.

[0113] like Figure 7 As shown, the network site selection prediction device 700 of this embodiment includes a first acquisition module 710 , a first calculation module 720 , a second calculation module 730 and a prediction module 740 .

[0114] The first acquisition module 710 is configured to acquire signaling data of the target area.

[0115] The first calculation module 720 is configured to store the signaling data of the target area into a graph database and construct a network point location information association graph.

[0116] The second calculation module 730 is configured to segment the network point location information association graph based on a preset distance granularity, and obtain m local network point location information association graph data, where m is an integer greater than or equal to 2.

[0117] The prediction module 740 is configured to pre-process the local network site selection information association graph and input it into a pre-trained network site selection prediction model to obtain a network site selection prediction result. The network site selection prediction result includes a predicted rating of the business level of the target area, and the pre-trained network site selection prediction model is constructed based on a graph convolutional neural network.

[0118] In some specific embodiments, the first computing module 720 may further include a storage submodule 7201 .

[0119] Figure 8 The structural block diagram of the first computing module 720 according to an embodiment of the present disclosure is schematically shown.

[0120] like Figure 8 As shown, the storage submodule 7201 is configured to take the target area as the core node and the attribute of the target area as the associated node, and store the target area and the attribute of the target area in the graph database. The graph database includes the stored existing network core nodes and existing network attribute associated nodes; the attributes correspond to the types of signaling data; the i-th core node has a mapping relationship with m associated nodes, where i satisfies 1≤i≤n and i is an integer, n is the total number of nodes in the network site selection information association graph, and m is the number of signaling data types.

[0121] In some specific embodiments, the prediction module 740 may also include a first convolution submodule 7401, a second convolution submodule 7402, a feature average submodule 7403, a first integration submodule 7404, a second integration submodule 7405 and a classification submodule 7406.

[0122] Fig. 9 The structural block diagram of the prediction module 740 according to an embodiment of the present disclosure is schematically shown.

[0123] like Fig. 9 As shown, the first convolution submodule 7401 is configured to input all m local network site selection information association graph data into the first graph convolutional network layer to obtain first-order node association feature information.

[0124] The second convolution submodule 7402 is configured to input the first-order node association feature information into the second graph convolutional network layer to obtain the second-order node association feature information.

[0125] The feature averaging submodule 7403 is configured to input the second-order node association feature information into the graph feature averaging layer to obtain graph average feature information.

[0126] The first integration submodule 7404 is configured to input the graph average feature information into the first fully connected layer to obtain the first global feature information.

[0127] The second integration submodule 7405 is configured to input the first global feature information into the second fully connected layer after nonlinear activation transformation to obtain second global feature information.

[0128] The classification submodule 7406 is configured to input the second global feature information into the classification network layer to obtain the network point service level prediction result.

[0129] In other specific embodiments, the graph database also includes stored business level data of existing outlets, and the prediction module 740 also includes a normalization submodule 7407, a vectorization submodule 7408, a marking submodule 7409 and a data aggregation submodule 7410.

[0130] Fig.10 The structural block diagram of some other prediction modules 740 according to the embodiment of the present disclosure is schematically shown.

[0131] like Fig.10 As shown, the normalization submodule 7407 is configured to normalize the signaling data of the i-th core node to obtain the feature vector of the i-th core node, wherein the i-th core node feature vector includes m features, and the m features are obtained based on associated nodes that have a mapping relationship with the i-th core node.

[0132] The vectorization submodule 7408 is configured to obtain an edge feature vector based on the distance between the ith core node and the remaining n-1 core nodes, and obtain an adjacency matrix of the ith core node based on the edge feature vector.

[0133] The marking submodule 7409 is configured to mark the existing network core node data as business level data.

[0134] The data aggregation submodule 7410 is configured to acquire local network point location information association graph data based on the feature vector of the core node, the adjacency matrix of the core node, and the existing network point core node data.

[0135] According to some other embodiments of the present disclosure, the network point location prediction device also includes a sorting module 750 .

[0136] Fig.11 The structural block diagram of the network site selection prediction device 700 according to some other embodiments of the present disclosure is schematically shown.

[0137] like Fig.11As shown, the sorting module 750 is configured to obtain the predicted business ratings of k target areas, where k is an integer greater than 1. The predicted business ratings of the k target areas are sorted from high to low, and the first ranked target area is used as the optimal network point prediction location.

[0138] According to an embodiment of the present disclosure, any multiple modules among the first acquisition module 710, the first calculation module 720, the second calculation module 730, the prediction module 740, the sorting module 750, the storage submodule 7201, the first convolution submodule 7401, the second convolution submodule 7402, the feature average submodule 7403, the first comprehensive submodule 7404, the second comprehensive submodule 7405, the classification submodule 7406, the normalization submodule 7407, the vectorization submodule 7408, the labeling submodule 7409 and the data summary submodule 7410 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the first acquisition module 710, the first calculation module 720, the second calculation module 730, the prediction module 740, the sorting module 750, the storage submodule 7201, the first convolution submodule 7401, the second convolution submodule 7402, the feature averaging submodule 7403, the first integration submodule 7404, the second integration submodule 7405, the classification submodule 7406, the normalization submodule 7407, the vectorization submodule 7408, the labeling submodule 7409 and the data aggregation submodule 7410 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or can be implemented in any one of the three implementation methods of software, hardware and firmware or in an appropriate combination of any of them. Alternatively, at least one of the first acquisition module 710, the first calculation module 720, the second calculation module 730, the prediction module 740, the sorting module 750, the storage submodule 7201, the first convolution submodule 7401, the second convolution submodule 7402, the feature averaging submodule 7403, the first integration submodule 7404, the second integration submodule 7405, the classification submodule 7406, the normalization submodule 7407, the vectorization submodule 7408, the labeling submodule 7409 and the data summary submodule 7410 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is executed.

[0139] Fig.12A block diagram of an electronic device suitable for implementing a network site selection prediction method according to an embodiment of the present disclosure is schematically shown.

[0140] like Fig.12 As shown, the electronic device 900 according to an embodiment of the present disclosure includes a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage part 908 to a random access memory (RAM) 903. The processor 901 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a dedicated microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 901 may also include an onboard memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0141] In RAM 903, various programs and data required for the operation of electronic device 900 are stored. Processor 901, ROM 902 and RAM 903 are connected to each other via bus 904. Processor 901 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 902 and / or RAM 903. It should be noted that the program can also be stored in one or more memories other than ROM 902 and RAM 903. Processor 901 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0142] According to an embodiment of the present disclosure, the electronic device 900 may further include an input / output (I / O) interface 905, which is also connected to the bus 904. The electronic device 900 may further include one or more of the following components connected to the I / O interface 905: an input portion 906 including a keyboard, a mouse, etc.; an output portion 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 908 including a hard disk, etc.; and a communication portion 909 including a network interface card such as a LAN card, a modem, etc. The communication portion 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the I / O interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed, so that a computer program read therefrom is installed into the storage portion 908 as needed.

[0143] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0144] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present disclosure, a computer-readable storage medium may include the ROM 902 and / or RAM 903 described above and / or one or more memories other than ROM 902 and RAM 903.

[0145] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the method provided by the embodiment of the present disclosure.

[0146] The above functions defined in the system / device of the embodiment of the present disclosure are performed when the computer program is executed by the processor 901. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0147] In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device, a magnetic storage device, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of a signal on a network medium, and downloaded and installed through the communication part 909, and / or installed from a removable medium 911. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0148] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0149] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0150] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0151] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations and / or combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways without departing from the spirit and teachings of the present disclosure. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0152] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. The scope of the present disclosure is defined by the attached claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A network site selection prediction method, characterized in that: include: Obtain signaling data of the target area; Storing the signaling data of the target area in a graph database to construct a network point location information association graph; Segmenting the network point location information association graph based on a preset distance granularity to obtain m local network point location information association graph data, where m is an integer greater than or equal to 2; and pre-processing the local network point location information association graph and inputting it into a pre-trained network point location prediction model to obtain a network point location prediction result. The network site selection prediction result includes a predicted rating of the service level of the target area, and the pre-trained network site selection prediction model is constructed based on a graph convolutional neural network; Wherein, the construction of the network point location information association map includes: Taking the target area as the core node and the attribute of the target area as the associated node, the target area and the attribute of the target area are stored in a graph database, wherein the graph database contains stored existing network core nodes and existing network attribute associated nodes; the attributes correspond to the types of signaling data; wherein there is a mapping relationship between the i-th core node and the m associated nodes, wherein i satisfies 1≤i≤n and i is an integer, n is the total number of nodes in the network site selection information association graph, and m is the number of signaling data types.

2. The method according to claim 1, wherein: The signaling data includes at least two of economic data, geographic data, physical condition data, and industry competition data.

3. The method according to claim 1, wherein: The graph database also includes stored service level data of existing outlets, and the pre-processing of associating the local outlet location information with the graph includes: Normalizing the signaling data of the ith core node to obtain a feature vector of the ith core node, wherein the feature vector of the ith core node includes m features, and the m features are obtained based on associated nodes that have a mapping relationship with the ith core node; Obtaining an edge feature vector based on the distance between the ith core node and the remaining n-1 core nodes, and obtaining an adjacency matrix of the ith core node based on the edge feature vector; Marking the existing network core node data as business-level data; and Based on the characteristic vector of the core node, the adjacency matrix of the core node, and the core node data of the existing network points, local network point location information association graph data is obtained.

4. The method according to claim 3, wherein: The obtaining of the prediction result of the network site selection also includes: Obtain the predicted business level ratings of k target areas, where k is an integer greater than 1; The predicted service level ratings of the k target areas are sorted in descending order, and the target area ranked first is used as the optimal network point prediction location.

5. The method according to any one of claims 3 to 4, wherein: The business data includes at least two of business transaction data, customer growth data, and customer flow data, and the business data is used to obtain business level data of existing outlets.

6. The method according to claim 1, wherein: The network site selection prediction model includes a first graph convolutional network layer, a second graph convolutional network layer, a graph feature averaging layer, a first fully connected layer, a second fully connected layer and a classification network layer.

7. The method according to claim 3, wherein: The obtaining of the network site selection prediction result comprises: Inputting the local network point location information association graph data into the first graph convolutional network layer to obtain first-order node association feature information; Inputting the first-order node association feature information into the second graph convolutional network layer to obtain second-order node association feature information; Inputting the second-order node association feature information into the graph feature averaging layer to obtain graph average feature information; Inputting the graph average feature information into a first fully connected layer to obtain first global feature information; Inputting the first global feature information into a second fully connected layer after nonlinear activation transformation to obtain second global feature information; and The second global feature information is input into the classification network layer to obtain the target area service level prediction result.

8. The method according to claim 6, wherein: When pre-training the network point location prediction model, back propagation and gradient descent optimization algorithms are used to update the parameters of the network point location prediction model to be trained until a preset training cutoff condition is reached.

9. A network site selection prediction device, characterized in that: include: A first acquisition module, configured to acquire signaling data of a target area; A first calculation module is configured to store the signaling data of the target area into a graph database and construct a network point location information association graph; A second calculation module is configured to segment the network point location information association graph based on a preset distance granularity to obtain m local network point location information association graph data, where m is an integer greater than or equal to 2; as well as A prediction module is configured to pre-process the local network point location information association graph and input it into a pre-trained network point location prediction model to obtain a network point location prediction result, wherein the network point location prediction result includes a predicted rating of a business level of a target area, and the pre-trained network point location prediction model is constructed based on a graph convolutional neural network; Wherein, the construction of the network point location information association map includes: Taking the target area as the core node and the attribute of the target area as the associated node, the target area and the attribute of the target area are stored in a graph database, wherein the graph database contains stored existing network core nodes and existing network attribute associated nodes; the attributes correspond to the types of signaling data; wherein there is a mapping relationship between the i-th core node and the m associated nodes, wherein i satisfies 1≤i≤n and i is an integer, n is the total number of nodes in the network site selection information association graph, and m is the number of signaling data types.

10. An electronic device comprising: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors execute the method according to any one of claims 1 to 8.

11. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the method according to any one of claims 1 to 8.

12. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images