Method, device, storage medium and electronic equipment for determining potential customers

Abstract

The application discloses a potential customer determination method and device, a storage medium and electronic equipment, and relates to the field of financial technology. The method comprises the following steps: acquiring customer data of a plurality of customers, and generating a target attribute network according to the customer data; determining an adjacency matrix corresponding to the target attribute network, and performing dimension reduction processing on the adjacency matrix to obtain a reduced matrix; determining a node feature vector of each customer node according to the reduced matrix; processing each node feature vector through a target model to obtain a predicted label of each customer, and determining a target potential customer from the plurality of customers according to the predicted label of each customer. The application solves the technical problem that the accuracy of identifying potential customers is low in the prior art when determining potential customers of a financial product according to artificial experience.

Description

Technical Field

[0001] This invention relates to the field of financial technology, and more specifically, to a method, apparatus, storage medium, and electronic device for identifying potential customers. Background Technology

[0002] With the booming development of financial services, the referral work faced by customer managers at the grassroots branches of financial institutions is becoming increasingly heavy. The current referral method mainly relies on human experience to identify potential customers for financial products and recommend them to them. This method is labor-intensive, time-consuming, and has the problem of low accuracy in identifying potential customers, resulting in low customer conversion rates.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a method, apparatus, storage medium, and electronic device for identifying potential customers, in order to at least solve the technical problem that the accuracy of identifying potential customers for financial products based on human experience in the prior art is low.

[0005] According to one aspect of the present invention, a method for determining potential customers is provided, comprising: acquiring customer data of multiple customers and generating a target attribute network based on the customer data, wherein the target attribute network includes at least customer nodes; determining an adjacency matrix corresponding to the target attribute network and performing dimensionality reduction processing on the adjacency matrix to obtain a dimensionality-reduced matrix; determining a node feature vector for each customer node based on the dimensionality-reduced matrix, wherein the node feature vector is used to characterize the characteristics of the customer node; processing each node feature vector through a target model to obtain a predicted label for each customer; and determining target potential customers from multiple customers based on the predicted labels for each customer, wherein the target model is obtained by training a machine learning model with sample customer data, and the predicted label is used to characterize whether a customer intends to purchase a target financial product.

[0006] Furthermore, the customer data includes at least basic customer information, customer asset information, customer liability information, and customer risk assessment information. The generation of the target attribute network based on the customer data includes: determining each customer node and multiple edges based on the basic customer information, and generating an initial attribute network based on each customer node and multiple edges, where edges are used to represent the relationships between customers; determining the weight value of each edge based on the customer asset information, customer liability information, and customer risk assessment information, and generating the target attribute network based on the weight value of each edge and the initial attribute network.

[0007] Further, the adjacency matrix corresponding to the target attribute network is determined, including: transforming the target attribute network based on the weight value of each edge to obtain the adjacency matrix.

[0008] Furthermore, the target model is used to process the feature vector of each node to obtain the predicted label of each customer. This includes: inputting the feature vector of each node into the target model, and calculating the probability value of each node feature vector belonging to each of the multiple preset labels through the target model to obtain multiple probability values ​​corresponding to each node feature vector; for each node feature vector, determining the maximum probability value from the multiple probability values, and using the target preset label corresponding to the maximum probability value as the predicted label to obtain the predicted label of each customer.

[0009] Furthermore, the target model is generated through the following steps: obtaining sample customer data and generating a sample attribute network based on the sample customer data, wherein the sample customer data contains real labels, which are used to identify whether the sample customer has purchased the target financial product; determining the sample node feature vector of each sample customer node in the sample attribute network, and training the machine learning model based on the sample node feature vector to obtain the target model.

[0010] Furthermore, the machine learning model is trained based on the feature vectors of the sample nodes to obtain the target model, including: inputting the feature vector of each sample node into the machine learning model and outputting the sample prediction label of each sample customer; iteratively training the machine learning model based on the sample prediction label and the true label of each sample customer until the target evaluation index of the machine learning model meets the preset evaluation criteria, thereby obtaining the target model.

[0011] Furthermore, after identifying target potential customers from multiple customers, the method also includes: obtaining product information of the target financial product, and recommending the target financial product to the target potential customer based on the product information.

[0012] According to another aspect of the present invention, a potential customer identification device is also provided, comprising: a first acquisition module, configured to acquire customer data of multiple customers and generate a target attribute network based on the customer data, wherein the target attribute network includes at least customer nodes; a first determination module, configured to determine the adjacency matrix corresponding to the target attribute network, and perform dimensionality reduction processing on the adjacency matrix to obtain a dimensionality-reduced matrix, and determine the node feature vector of each customer node based on the dimensionality-reduced matrix, wherein the node feature vector is used to characterize the features of the customer node; and a second determination module, configured to process the node feature vector of each customer through a target model to obtain a predicted label for each customer, and determine a target potential customer from multiple customers based on the predicted label of each customer, wherein the target model is obtained by training a machine learning model with sample customer data, and the predicted label is used to characterize whether the customer intends to purchase the target financial product.

[0013] According to another aspect of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to execute the above-described method for determining potential customers when it is run.

[0014] According to another aspect of the present invention, an electronic device is also provided, the electronic device including 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 are configured to run the programs, wherein the programs are configured to execute the above-described method for determining potential customers during runtime.

[0015] In this embodiment of the invention, a method based on attribute network representation learning is used to determine potential customers. First, customer data from multiple customers is acquired, and a target attribute network is generated based on this data. Then, the adjacency matrix corresponding to the target attribute network is determined, and the adjacency matrix is ​​reduced in dimensionality to obtain a dimensionality-reduced matrix. Based on the dimensionality-reduced matrix, the node feature vector of each customer node is determined. Then, the target model processes each node feature vector to obtain a predicted label for each customer. Based on the predicted labels of each customer, target potential customers are determined from multiple customers. The target attribute network includes at least customer nodes, the node feature vectors characterize the features of the customer nodes, the target model is trained using sample customer data, and the predicted labels characterize whether a customer intends to purchase the target financial product.

[0016] In the above process, customer nodes are transformed into information-rich and highly discriminative network representation vectors through attribute-based network representation learning methods. Then, the trained machine learning model predicts the label of each customer, i.e., whether the customer is going to purchase the target financial product. Based on the predicted label of each customer, the target potential customer can be identified from multiple customers, saving time and manpower costs, improving the efficiency and accuracy of identifying potential customers, and thus effectively improving the customer conversion rate.

[0017] Therefore, the technical solution of this invention achieves the goal of effectively improving customer conversion rate, thereby improving the efficiency and accuracy of identifying potential customers. It also solves the technical problem that the accuracy of identifying potential customers for financial products based on human experience in the prior art is low. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0019] Figure 1 This is a flowchart of an optional method for determining potential customers according to an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of an optional process for mining potential customers according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of an optional potential customer identification device according to an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of an optional electronic device according to an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] It should be noted that all relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this invention are information and data authorized by the user or fully authorized by all parties. For example, this system has an interface with the relevant user or organization. Before obtaining relevant information, it needs to send an acquisition request to the aforementioned user or organization through the interface, and obtain the relevant information after receiving consent from the aforementioned user or organization.

[0026] Example 1

[0027] According to an embodiment of the present invention, an embodiment of a method for identifying potential customers is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0028] Figure 1 This is a flowchart of an optional method for determining potential customers according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0029] Step S101: Obtain customer data from multiple customers and generate a target attribute network based on the customer data, wherein the target attribute network includes at least customer nodes.

[0030] In the above steps, customer data of multiple customers can be obtained through application systems, processors, electronic devices, etc. Optionally, customer data of multiple customers can be obtained through a potential customer identification system. The customer data includes basic customer information (e.g., the customer's industry, customer age range, etc.) and additional customer information with application value (e.g., customer asset information, customer liability information, and customer risk assessment information, etc.).

[0031] Optionally, by integrating basic customer information and additional customer information, a customer network (i.e., a target attribute network) containing weight information can be constructed. For example, a customer attribute network can be constructed using basic customer information, treating each customer as a node (i.e., a customer node), the relationship between customers as an edge, and the weight of the edge between each customer node and other customer nodes can be determined based on the additional customer information to obtain the target attribute network.

[0032] Step S102: Determine the adjacency matrix corresponding to the target attribute network, and perform dimensionality reduction processing on the adjacency matrix to obtain the dimensionality-reduced matrix. Based on the dimensionality-reduced matrix, determine the node feature vector of each customer node, where the node feature vector is used to characterize the features of the customer node.

[0033] Optionally, after obtaining the target attribute network, the network data is transformed into an adjacency matrix, and the adjacency matrix is ​​subjected to dimensionality reduction processing. For example, the adjacency matrix is ​​subjected to dimensionality reduction processing by the Singular Value Decomposition (SVD) method to obtain the dimensionality-reduced matrix. From the dimensionality-reduced matrix, the node feature vector of each customer node can be obtained, that is, the feature vector representation of each customer node in the network is obtained.

[0034] Step S103: Process the feature vector of each node through the target model to obtain the predicted label of each customer, and determine the target potential customer from multiple customers based on the predicted label of each customer. The target model is obtained by training a machine learning model with sample customer data, and the predicted label is used to characterize whether the customer is going to purchase the target financial product.

[0035] Optionally, the feature vector of each node is input into a trained machine learning model (i.e., the target model). The model can predict the label corresponding to each customer node and obtain the predicted label of each customer. For example, by using a trained logistic regression model, it can predict whether each customer is going to purchase the target financial product and obtain the label of whether they will buy or not. Thus, based on the predicted label of each customer, the target potential customer can be identified from multiple customers, and the accurate identification of potential customers can be achieved.

[0036] Based on the scheme defined in steps S101 to S103 above, it can be understood that in this embodiment of the invention, the method of determining potential customers using attribute network representation learning first obtains customer data from multiple customers, generates a target attribute network based on the customer data, then determines the adjacency matrix corresponding to the target attribute network, and performs dimensionality reduction processing on the adjacency matrix to obtain a dimensionality-reduced matrix. Based on the dimensionality-reduced matrix, the node feature vector of each customer node is determined. Then, the target model processes each node feature vector to obtain the predicted label of each customer, and based on the predicted label of each customer, the target potential customer is determined from multiple customers. The target attribute network includes at least customer nodes, the node feature vector is used to characterize the characteristics of the customer nodes, the target model is obtained by training a machine learning model using sample customer data, and the predicted label is used to characterize whether the customer intends to purchase the target financial product.

[0037] It is noteworthy that in the above process, customer nodes are transformed into information-rich and highly discriminative network representation vectors through attribute-based network representation learning methods. Then, the trained machine learning model predicts the label of each customer, i.e., whether the customer intends to purchase the target financial product. Based on the predicted label of each customer, the target potential customer can be identified from multiple customers, saving time and manpower costs, improving the efficiency and accuracy of identifying potential customers, and thus effectively improving the customer conversion rate.

[0038] Therefore, the technical solution of this invention achieves the goal of effectively improving customer conversion rate, thereby improving the efficiency and accuracy of identifying potential customers. It also solves the technical problem that the accuracy of identifying potential customers for financial products based on human experience in the prior art is low.

[0039] In one optional embodiment, the customer data includes at least customer basic information, customer asset information, customer liability information, and customer risk assessment information. The generation of the target attribute network based on the customer data includes: determining each customer node and multiple edges based on the customer basic information, and generating an initial attribute network based on each customer node and multiple edges, wherein the edges are used to represent the relationships between customers; determining the weight value of each edge based on the customer asset information, customer liability information, and customer risk assessment information, and generating the target attribute network based on the weight value of each edge and the initial attribute network.

[0040] Optionally, basic customer information may include the customer's industry, age group, etc.; customer asset information may include the customer's asset level; customer liability information may include the customer's debt level; and customer risk assessment information may include the customer's risk assessment level. Optionally, in generating the target attribute network based on customer data, each customer node and multiple edges are first determined based on the basic customer information, and an initial attribute network is generated based on each customer node and multiple edges. For example, customer data from a certain region can be abstracted into a network, with each customer considered as a node and the relationships between customers as edges (e.g., constructing edges between customer nodes with the same account opening location), constructing a network G = (V, E), where V is the set of nodes, E is the set of edges, and edge e = (V... i V j )∈E represents V i To V j An edge between them.

[0041] Optionally, the weight value of each edge can be determined based on customer asset information, customer liability information, and customer risk assessment information, and a target attribute network can be generated based on the weight value of each edge and the initial attribute network. For example, the average weight can be calculated based on three indicators: the asset level, liability level, and risk assessment level of the customer node. The weight value of the calculated average weight can be used as the weight value of each edge to obtain a customer network containing weight information (i.e., the target attribute network).

[0042] It should be noted that in the above process, the massive amount of customer information and the additional customer information with application value are integrated. That is, on the basis of constructing a customer attribute network using basic customer information, the additional customer information is introduced as the weight of the edge between each customer node and other customer nodes. This makes the vector representation of each node in the network richer and more distinguishable, thereby effectively improving the accuracy of identifying potential customers.

[0043] In one optional embodiment, determining the adjacency matrix corresponding to the target attribute network includes: transforming the target attribute network based on the weight value of each edge to obtain the adjacency matrix.

[0044] Optionally, in determining the adjacency matrix corresponding to the target attribute network, the target attribute network is transformed based on the weight value of each edge to obtain the adjacency matrix. Specifically, the network data G = (V, E) is transformed into the adjacency matrix A ∈ R. |V|×|V| , where if (V i V j If )∈E, then A ij Not equal to 0, otherwise A ij =0. After calculating the average weight, the average weight is used as the value of the adjacency matrix. Since the proximity relationship between each customer node and other customer nodes is different, the weight value of each unit in the adjacency matrix is ​​different. Based on this, the dimensionality of the adjacency matrix can be reduced by the SVD decomposition method, thereby obtaining the feature vector representation of each customer node in the network.

[0045] It should be noted that by transforming the customer attribute network into an adjacency matrix, the dimensionality of the adjacency matrix can be reduced, thereby obtaining a distinctive node vector representation, which provides a data foundation for subsequently identifying target potential customers from multiple customers.

[0046] In one optional embodiment, the target model is used to process the feature vector of each node to obtain the predicted label of each customer. This includes: inputting the feature vector of each node into the target model, and calculating the probability value of each node feature vector belonging to each of the multiple preset labels through the target model to obtain multiple probability values ​​corresponding to each node feature vector; for each node feature vector, determining the maximum probability value from the multiple probability values, and using the target preset label corresponding to the maximum probability value as the predicted label to obtain the predicted label of each customer.

[0047] Optionally, in the process of processing the feature vector of each node through the target model to obtain the predicted label of each customer, the feature vector of each node is first input into the target model, and the probability value of each node feature vector belonging to each of the multiple preset labels is calculated by the target model. Multiple probability values ​​corresponding to each node feature vector can be obtained. For example, the vector representation of each customer node generated is used as the customer node feature for label prediction. The vector representation of the network customer node is input into the target model (e.g., logistic regression model), and the probability value of customer node A belonging to preset label 1 (e.g., the "will buy" label) is calculated to be 0.9, and the probability value of customer node A belonging to preset label 2 (e.g., the "will not buy" label) is 0.1.

[0048] Optionally, for each node feature vector, the maximum probability value is determined from multiple probability values, and the target preset label corresponding to the maximum probability value is used as the predicted label to obtain the predicted label for each customer. For example, for customer node A, the probability value of customer node A belonging to preset label 1 is the highest, so preset label 1 is the target preset label, that is, the predicted label of customer node A is preset label 1.

[0049] It should be noted that by using a trained machine learning model to predict each customer's label, i.e., to predict whether the customer is ready to purchase the target financial product, a data foundation is provided for identifying potential customers from multiple customers, saving time and manpower costs and improving the efficiency of identifying potential customers.

[0050] In one optional embodiment, the target model is generated through the following steps: obtaining sample customer data and generating a sample attribute network based on the sample customer data, wherein the sample customer data contains real labels, which are used to identify whether the sample customer has purchased the target financial product; determining the sample node feature vector of each sample customer node in the sample attribute network, and training the machine learning model based on the sample node feature vector to obtain the target model.

[0051] Optionally, the target model can be obtained by training the machine learning model using sample customer data. For example, sample customer data (including basic customer information and additional customer information) can be obtained, and the sample customer nodes that have purchased the target financial product can be labeled using expert manual annotation. These labels serve as the true labels for each sample customer node. Each sample customer is considered as a node, and the relationships between sample customers are considered as edges. The weights of the edges are determined based on the additional customer information to construct a sample attribute network.

[0052] Furthermore, by determining the feature vector of each sample customer node in the sample attribute network and training the machine learning model based on these feature vectors, the target model can be obtained. For example, the sample attribute network can be transformed into an adjacency matrix, and the adjacency matrix can be reduced in dimensionality to obtain the feature vector representation of each sample customer node in the network. This feature vector is then used as the sample customer feature input to train the machine learning model, thus obtaining the target model.

[0053] In one optional embodiment, training a machine learning model based on the feature vectors of sample nodes to obtain a target model includes: inputting the feature vector of each sample node into the machine learning model and outputting the sample prediction label of each sample customer; iteratively training the machine learning model based on the sample prediction label and the true label of each sample customer until the target evaluation index of the machine learning model meets the preset evaluation criteria, thereby obtaining the target model.

[0054] Optionally, in the process of training the machine learning model based on the feature vectors of sample nodes to obtain the target model, the feature vector of each sample node is first input into the machine learning model, and the model outputs the predicted label of each sample customer. Then, based on the predicted label and the true label of each sample customer, the machine learning model is iteratively trained until the target evaluation metric of the machine learning model meets the preset evaluation criteria, thus obtaining the target model. For example, the accuracy of the model is measured by comparing whether the predicted label output by the model is consistent with the true label. Generally speaking, the larger the F1 score, the better the accuracy of the model classification. When the preset evaluation threshold is reached, the model training is considered complete, and the target model is obtained.

[0055] It should be noted that a well-trained machine learning model can predict each customer's tags, that is, predict whether the customer is going to purchase the target financial product. Based on the predicted tags of each customer, target potential customers can be identified from multiple customers, saving time and manpower costs, improving the efficiency and accuracy of identifying potential customers, and effectively improving customer conversion rates.

[0056] In one alternative embodiment, after identifying target potential customers from multiple customers, product information of the target financial product is obtained, and the target financial product is recommended to the target potential customers based on the product information.

[0057] Optionally, after identifying target potential customers from multiple customers, product information of the financial products to be recommended can be obtained based on the current recommendation task. Based on the product information, online and offline recommendation activities can be carried out on the massive number of potential customers identified, that is, recommending target financial products to target potential customers.

[0058] Figure 2 This is a schematic diagram of an optional process for identifying potential customers according to an embodiment of the present invention, such as... Figure 2 As shown, customer data is acquired and cleaned. A network is constructed based on the customer's basic information and labeled. The network is then transformed into a matrix vector. The weights of the node's assets, liabilities, and risk information are calculated and used as values ​​in the matrix vector. Dimensionality is then reduced using the SVD decomposition method to obtain the feature vector representation of the node. This vector representation is then used as the feature input of the regression model for multi-label classification to obtain the label of the unknown node.

[0059] It should be noted that, in this embodiment of the invention, the method of mining potential customers based on attribute network representation learning integrates massive amounts of customer information and additional customer information with application value. That is, based on constructing a customer attribute network using basic customer information, additional customer information is introduced as the weight of the edges between each customer node and other customer nodes. The customer attribute network is then transformed into an adjacency matrix, and the dimensionality of the adjacency matrix is ​​reduced using the SVD decomposition method to obtain a discriminative node vector representation. Then, a trained machine learning model is used to predict the label of each customer. This enables the prediction of the label of unknown nodes using a small number of manually labeled known nodes, saving time and manpower costs, improving the efficiency and accuracy of identifying potential customers, and thus effectively improving the customer conversion rate. This provides strong support for grassroots branch business personnel to accurately recommend financial products to customers.

[0060] Therefore, the technical solution of this invention achieves the goal of effectively improving customer conversion rate, thereby improving the efficiency and accuracy of identifying potential customers. It also solves the technical problem that the accuracy of identifying potential customers for financial products based on human experience in the prior art is low.

[0061] Example 2

[0062] According to an embodiment of the present invention, an embodiment of a potential customer identification device is provided, wherein, Figure 3 This is a schematic diagram of an optional potential customer identification device according to an embodiment of the present invention, such as... Figure 3 As shown, the device includes: a first acquisition module 301, used to acquire customer data of multiple customers and generate a target attribute network based on the customer data, wherein the target attribute network includes at least customer nodes; a first determination module 302, used to determine the adjacency matrix corresponding to the target attribute network, and perform dimensionality reduction processing on the adjacency matrix to obtain a dimensionality-reduced matrix, and determine the node feature vector of each customer node based on the dimensionality-reduced matrix, wherein the node feature vector is used to characterize the features of the customer node; and a second determination module 303, used to process the feature vector of each node through a target model to obtain a predicted label for each customer, and determine target potential customers from multiple customers based on the predicted label of each customer, wherein the target model is obtained by training a machine learning model with sample customer data, and the predicted label is used to characterize whether the customer intends to purchase the target financial product.

[0063] It should be noted that the first acquisition module 301, the first determination module 302 and the second determination module 303 mentioned above correspond to steps S101 to S103 in the above embodiments. The three modules and the corresponding steps implement the same examples and application scenarios, but are not limited to the content disclosed in the above embodiment 1.

[0064] Optionally, the customer data includes at least customer basic information, customer asset information, customer liability information, and customer risk assessment information. The first acquisition module includes: a first generation module, used to determine each customer node and multiple edges based on the customer basic information, and to generate an initial attribute network based on each customer node and multiple edges, wherein the edges are used to represent the relationship between customers; and a second generation module, used to determine the weight value of each edge based on the customer asset information, customer liability information, and customer risk assessment information, and to generate a target attribute network based on the weight value of each edge and the initial attribute network.

[0065] Optionally, the first determining module includes: a first processing module, used to transform the target attribute network based on the weight value of each edge to obtain an adjacency matrix.

[0066] Optionally, the second determining module includes: a calculation module, used to input each node feature vector into the target model, and calculate the probability value of each node feature vector belonging to each of the multiple preset labels through the target model, so as to obtain multiple probability values ​​corresponding to each node feature vector; and a third determining module, used to determine the maximum probability value from the multiple probability values ​​for each node feature vector, and use the target preset label corresponding to the maximum probability value as the prediction label, so as to obtain the prediction label for each customer.

[0067] Optionally, the potential customer identification device further includes the following modules for generating a target model through the following steps: a second acquisition module for acquiring sample customer data and generating a sample attribute network based on the sample customer data, wherein the sample customer data contains real labels, which are used to identify whether the sample customer has purchased the target financial product; and a fourth determination module for determining the sample node feature vector of each sample customer node in the sample attribute network and training the machine learning model based on the sample node feature vector to obtain the target model.

[0068] Optionally, the fourth determining module includes: a second processing module, used to input the feature vector of each sample node into the machine learning model and output the sample prediction label of each sample customer; and a training module, used to iteratively train the machine learning model based on the sample prediction label and the true label of each sample customer until the target evaluation index of the machine learning model meets the preset evaluation criteria, thereby obtaining the target model.

[0069] Optionally, the device for identifying potential customers further includes a recommendation module, which, after identifying target potential customers from multiple customers, obtains product information of the target financial product and recommends the target financial product to the target potential customer based on the product information.

[0070] Example 3

[0071] According to another aspect of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, wherein the computer program is configured to execute the above-described method for determining potential customers when it is run.

[0072] Example 4

[0073] According to another aspect of the present invention, an electronic device is also provided, wherein, Figure 4 This is a schematic diagram of an optional electronic device according to an embodiment of the present invention, such as... Figure 4 As shown, the electronic device includes one or more processors; a memory for storing one or more programs, which, when executed by the one or more processors, enable the one or more processors to run the programs, wherein the programs are configured to execute the aforementioned method for determining potential customers. When the processor executes the program, it performs the following steps: acquiring customer data from multiple customers and generating a target attribute network based on the customer data, wherein the target attribute network includes at least customer nodes; determining the adjacency matrix corresponding to the target attribute network and performing dimensionality reduction on the adjacency matrix to obtain a dimensionality-reduced matrix; determining the node feature vector of each customer node based on the dimensionality-reduced matrix, wherein the node feature vector is used to characterize the features of the customer node; processing each node feature vector through a target model to obtain a predicted label for each customer; and determining target potential customers from multiple customers based on the predicted labels of each customer, wherein the target model is obtained by training a machine learning model using sample customer data, and the predicted labels are used to characterize whether a customer intends to purchase the target financial product.

[0074] Optionally, the processor, when executing the program, also implements the following steps: the customer data includes at least customer basic information, customer asset information, customer liability information, and customer risk assessment information, wherein generating a target attribute network based on the customer data includes: determining each customer node and multiple edges based on the customer basic information, and generating an initial attribute network based on each customer node and multiple edges, wherein the edges are used to represent the relationship between customers; determining the weight value of each edge based on the customer asset information, customer liability information, and customer risk assessment information, and generating a target attribute network based on the weight value of each edge and the initial attribute network.

[0075] Optionally, when the processor executes the program, it also performs the following steps: determining the adjacency matrix corresponding to the target attribute network, including: transforming the target attribute network based on the weight value of each edge to obtain the adjacency matrix.

[0076] Optionally, the processor also performs the following steps when executing the program: processing the feature vector of each node through the target model to obtain the predicted label of each customer, including: inputting the feature vector of each node into the target model, and calculating the probability value of each node feature vector belonging to each of the multiple preset labels through the target model to obtain multiple probability values ​​corresponding to each node feature vector; for each node feature vector, determining the maximum probability value from the multiple probability values, and using the target preset label corresponding to the maximum probability value as the predicted label to obtain the predicted label of each customer.

[0077] Optionally, the processor also performs the following steps when executing the program: generating a target model through the following steps: acquiring sample customer data and generating a sample attribute network based on the sample customer data, wherein the sample customer data contains real labels, which are used to identify whether the sample customer has purchased the target financial product; determining the sample node feature vector of each sample customer node in the sample attribute network, and training the machine learning model based on the sample node feature vector to obtain the target model.

[0078] Optionally, the processor may also perform the following steps when executing the program: training the machine learning model based on the feature vectors of the sample nodes to obtain the target model, including: inputting the feature vector of each sample node into the machine learning model and outputting the sample prediction label of each sample customer; iteratively training the machine learning model based on the sample prediction label and the true label of each sample customer until the target evaluation index of the machine learning model meets the preset evaluation criteria, thereby obtaining the target model.

[0079] Optionally, the processor may also perform the following steps when executing the program: after identifying target potential customers from multiple customers, obtain product information of the target financial product, and recommend the target financial product to the target potential customers based on the product information.

[0080] The devices mentioned in this article can be servers, PCs, tablets, mobile phones, etc.

[0081] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0082] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0083] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0084] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0085] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0086] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0087] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying potential customers, characterized in that, include: Acquire customer data from multiple customers and generate a target attribute network based on the customer data, wherein the target attribute network includes at least customer nodes; The adjacency matrix corresponding to the target attribute network is determined, and the adjacency matrix is ​​reduced in dimension to obtain a dimension-reduced matrix. Based on the dimension-reduced matrix, the node feature vector of each customer node is determined, wherein the node feature vector is used to characterize the features of the customer node. The target model processes the feature vector of each node to obtain the predicted label of each customer, and the target potential customer is determined from the multiple customers based on the predicted label of each customer. The target model is obtained by training a machine learning model with sample customer data, and the predicted label is used to characterize whether the customer is going to purchase the target financial product. Specifically, the target model processes the feature vector of each node to obtain the predicted label for each customer, including: Each node feature vector is input into the target model, and the probability value of each node feature vector belonging to each of the multiple preset labels is calculated through the target model to obtain multiple probability values ​​corresponding to each node feature vector. For each node feature vector, the maximum probability value is determined from the plurality of probability values, and the target preset label corresponding to the maximum probability value is used as the predicted label to obtain the predicted label for each customer; The customer data includes at least basic customer information, customer asset information, customer liability information, and customer risk assessment information. The generation of a target attribute network based on the customer data includes: Each customer node and multiple edges are determined based on the customer's basic information, and an initial attribute network is generated based on each customer node and the multiple edges, wherein the edges are used to represent the relationship between customers; The weight value of each edge is determined based on the customer asset information, the customer liability information, and the customer risk assessment information, and the target attribute network is generated based on the weight value of each edge and the initial attribute network. Determining the adjacency matrix corresponding to the target attribute network includes: Based on the weight value of each edge, the target attribute network is transformed to obtain the adjacency matrix, wherein the value of the adjacency matrix is ​​the weight value of each edge.

2. The method according to claim 1, characterized in that, The target model is generated through the following steps: The sample customer data is obtained, and a sample attribute network is generated based on the sample customer data, wherein the sample customer data includes real labels, and the real labels are used to identify whether the sample customer has purchased the target financial product; The sample node feature vector of each sample client node in the sample attribute network is determined, and the machine learning model is trained based on the sample node feature vector to obtain the target model.

3. The method according to claim 2, characterized in that, The machine learning model is trained based on the feature vectors of the sample nodes to obtain the target model, including: The feature vector of each sample node is input into the machine learning model, and the predicted label of each sample customer is output. Based on the predicted labels and the true labels of each sample customer, the machine learning model is iteratively trained until the target evaluation index of the machine learning model meets the preset evaluation criteria, thus obtaining the target model.

4. The method according to claim 1, characterized in that, After identifying target potential customers from the plurality of customers, the method further includes: Obtain product information of the target financial product, and recommend the target financial product to the target potential customer based on the product information.

5. A device for identifying potential customers, characterized in that, include: The first acquisition module is used to acquire customer data from multiple customers and generate a target attribute network based on the customer data, wherein the target attribute network includes at least customer nodes; The first determining module is used to determine the adjacency matrix corresponding to the target attribute network, and to perform dimensionality reduction processing on the adjacency matrix to obtain a dimensionality-reduced matrix. Based on the dimensionality-reduced matrix, the node feature vector of each customer node is determined, wherein the node feature vector is used to characterize the features of the customer node. The second determining module is used to process the feature vector of each node through the target model to obtain the predicted label of each customer, and to determine the target potential customer from the multiple customers based on the predicted label of each customer. The target model is obtained by training a machine learning model with sample customer data, and the predicted label is used to characterize whether the customer is going to purchase the target financial product. The second determining module includes: a calculation module, used to input each node feature vector into the target model, and calculate the probability value of each node feature vector belonging to each of the multiple preset labels through the target model, so as to obtain multiple probability values ​​corresponding to each node feature vector; and a third determining module, used to determine the maximum probability value from the multiple probability values ​​for each node feature vector, and use the target preset label corresponding to the maximum probability value as the prediction label, so as to obtain the prediction label for each customer. The customer data includes at least customer basic information, customer asset information, customer liability information, and customer risk assessment information. The first acquisition module includes: a first generation module, used to determine each customer node and multiple edges based on the customer basic information, and to generate an initial attribute network based on each customer node and multiple edges, wherein the edges are used to represent the relationship between customers; and a second generation module, used to determine the weight value of each edge based on the customer asset information, customer liability information, and customer risk assessment information, and to generate a target attribute network based on the weight value of each edge and the initial attribute network. The first determining module includes: a first processing module, used to transform the target attribute network based on the weight value of each edge to obtain an adjacency matrix, wherein the value of the adjacency matrix is ​​the weight value of each edge.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute, when run, the method for determining potential customers as described in any one of claims 1 to 4.

7. An electronic device, characterized in that, The electronic device includes one or more processors; A memory for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to be configured to run the programs, wherein the programs are configured to execute the method for determining potential customers as described in any one of claims 1 to 4.

Images