Customer mining method, device and equipment and storage medium

Abstract

The application discloses a customer mining method and device, equipment and a storage medium, and belongs to the technical field of information processing. The application determines a new customer conforming to a received query label; calculates a value score of the new customer based on a preset value score of an inventory customer and a first correlation between the inventory customer and the new customer; objectively evaluates the value score of the new customer, without manually analyzing whether the new customer is valuable, and without comparing the inventory customer with the new customer manually; determines a target new customer based on the value score of the new customer; takes the target new customer as a query result and returns the query result, so that a user can obtain the target new customer by checking the query result. Therefore, the application can improve the mining efficiency of customers.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information processing, in particular to a customer mining method and device, equipment and storage medium. BACKGROUND

[0002] At present, when customer mining is needed, the new customers found are compared one by one with the inventory customers in the enterprise, and it is manually analyzed whether there are valuable new customers.

[0003] This customer mining method may lead to that the value of the customers finally mined is not particularly high due to the subjectivity of manual work, and the time cost is also high when manually comparing new customers with inventory customers one by one, resulting in low efficiency of customer mining. SUMMARY

[0004] The main purpose of the present application is to provide a customer mining method, device, equipment and storage medium, aiming at solving the technical problem of low efficiency of customer mining.

[0005] To achieve the above purpose, the present application provides a customer mining method, which comprises the following steps:

[0006] Determine the new customers meeting the received query label;

[0007] Calculate the value score of the new customers based on the preset value score of the inventory customers and the first association relationship between the inventory customers and the new customers;

[0008] Determine the target new customers based on the value score of the new customers;

[0009] Take the target new customers as the query result and return the query result.

[0010] Optionally, the step of calculating the value score of the new customers based on the preset value score of the inventory customers and the first association relationship between the inventory customers and the new customers comprises:

[0011] Determine a plurality of first customer type labels of the inventory customers based on the mapping relationship between the preset relationship and the label;

[0012] Calculate the value score of the new customers based on the plurality of first customer type labels, the preset value score and the first association relationship between the inventory customers and the new customers through a preset relationship score model, wherein the preset relationship score model is used to represent a second association relationship between the value score of the inventory customers and the new customers.

[0013] Optionally, before the step of calculating the value score of the new customer based on the plurality of first customer type labels, the preset value score, and the first association relationship between the inventory customers and the new customer, the method further comprises:

[0014] obtaining a second customer type label of a historical inventory customer, a historical value score, and a third association relationship between the historical inventory customer and a historical new customer;

[0015] constructing a customer relationship graph based on the second customer type label, the historical value score, and the third association relationship;

[0016] iteratively training a preset graph neural network model based on the customer relationship graph and the first score label of the historical new customer to obtain a relationship score model meeting an accuracy condition.

[0017] Optionally, the step of constructing the customer relationship graph based on the second customer type label, the historical value score, and the third association relationship comprises:

[0018] regarding the historical inventory customer and the historical new customer as nodes;

[0019] regarding the third association relationship between the historical inventory customer and the historical new customer as a directed edge;

[0020] encapsulating features for the nodes based on the second customer type label and the historical value score;

[0021] constructing the customer relationship graph based on the directed edge and the nodes after the encapsulation of features;

[0022] The step of iteratively training a preset graph neural network model based on the customer relationship graph and the first score label of the historical new customer to obtain a relationship score model meeting an accuracy condition comprises:

[0023] obtaining weight features corresponding to different types of third association relationships;

[0024] iteratively training a preset graph convolutional neural network model based on the customer relationship graph, the first score label of the historical new customer, and the weight features to obtain a relationship score model meeting an accuracy condition.

[0025] Optionally, before the step of calculating the value score of the new customer based on the preset value score of the inventory customer and the first association relationship between the inventory customer and the new customer, the method further comprises:

[0026] The value score of the inventory customer is calculated based on the third customer type label of the inventory customer and a preset value score model, wherein the preset value score model is used to represent a fourth correlation between the third customer type label and the value score.

[0027] Optionally, before the step of calculating the value score of the inventory customer based on the third customer type label of the inventory customer and a preset value score model, the method further comprises:

[0028] obtaining a third customer type label of a historical inventory customer;

[0029] determining a second score label of the historical inventory customer based on the third customer type label of the historical inventory customer;

[0030] iteratively training a preset decision tree model to be trained based on the third customer type label of the historical inventory customer and the second score label, to obtain a target value score model meeting an accuracy condition.

[0031] Optionally, the third customer type label comprises at least one of a listed company, a group customer, a deposit size, a loan size, and a tax amount, and the preset decision tree model to be trained comprises at least one of an ID3 model, a C4.5 model, and a CART model;

[0032] After the step of obtaining the third customer type label of the historical inventory customer, the method further comprises:

[0033] combining features of third customer type labels of the same type based on the type of the third customer type label, to obtain a fourth customer type label;

[0034] determining a second score label of the historical inventory customer based on the fourth customer type label of the historical inventory customer;

[0035] iteratively training at least one of the ID3 model, the C4.5 model, and the CART model based on the fourth customer type label of the historical inventory customer and the second score label, to obtain a plurality of value score models meeting the accuracy condition;

[0036] verifying the plurality of value score models based on a preset verification set, to obtain a verification result, and screening the target value score model based on the verification result.

[0037] In addition, to achieve the above-mentioned purposes, the application further provides a customer mining device, which comprises:

[0038] The inventory customer determination module is configured to determine a new customer that meets the received query label;

[0039] The score calculation module is configured to calculate a value score of the new customer based on a preset value score of an inventory customer and a first association relationship between the inventory customer and the new customer.

[0040] The new customer determination module is configured to determine a target new customer based on the value score of the new customer.

[0041] The result returning module is configured to return the target new customer as a query result.

[0042] In addition, to achieve the above object, the present application further provides a device, which comprises a memory, a processor, and a customer mining program stored in the memory and executable on the processor, and the customer mining program is configured to implement the steps of the customer mining method.

[0043] In addition, to achieve the above object, the present application further provides a computer readable storage medium, which stores a customer mining program, and the customer mining program is executable on a processor to implement the steps of the customer mining method.

[0044] The present application can determine a new customer that meets the received query label, calculate a value score of the new customer based on a preset value score of an inventory customer and a first association relationship between the inventory customer and the new customer, objectively evaluate the value score of the new customer without manual analysis of whether the new customer is valuable or comparison between the inventory customer and the new customer, determine a target new customer based on the value score of the new customer, return the target new customer as a query result, and enable a user to obtain the target new customer by checking the query result. Therefore, the present application can improve the mining efficiency of customers. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The figure is a flowchart of a first embodiment of the customer mining method of the present application;

[0046] Figure 2 The figure is a first scenario diagram of the first embodiment of the customer mining method of the present application;

[0047] Figure 3 The figure is a second scenario diagram of the first embodiment of the customer mining method of the present application;

[0048] Figure 4 The figure is a third scenario diagram of the first embodiment of the customer mining method of the present application;

[0049] Figure 5A flowchart of a second embodiment of a method for mining customers for a client of the application;

[0050] Figure 6 A structural block diagram of a first embodiment of a device for mining customers for a client of the application;

[0051] Figure 7 A device structure diagram of a hardware operating environment involved in an embodiment of the application.

[0052] The implementation, functional features and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein are merely intended to explain the application and not to limit the application.

[0054] Referring to Figure 1 , Figure 1 A flowchart of a first embodiment of a method for mining customers for a client of the application.

[0055] In the first embodiment, the method for mining customers comprises the following steps:

[0056] S10: determining new customers meeting the received query label.

[0057] It should be noted that the execution subject of the method of the embodiment is a customer mining device, which can belong to a customer mining device. The customer mining device can be an application program or a customer mining platform, etc., and the application does not make specific limitations.

[0058] It can be understood that the user can perform customer mining queries through an application program or a customer mining platform, etc. Specifically, as shown in Figure 2 , the user inputs or selects a query label on the user interface, and the user query instruction analysis system of the application program or the customer mining platform performs analysis based on the enterprise association database, and returns a valuable new customer list required by the user.

[0059] Among them, the query label can be extracted based on the user input query content, or can be determined based on the user selected keyword. Specifically, the query label can correspond to the customer type label. If the query label is determined, the new customers meeting the customer type label can be determined based on the correspondence between the query label and the customer type label, that is, the new customers meeting the received query label are determined.

[0060] For example, as shown in Figure 3 , the associated inventory customers (customer type) are part of the query label, and the query label can also include the associated approach (relationship type), the establishment date of the new customer, the province and region to which the new customer belongs, etc.

[0061] Specifically, the customer types include total war, sub-war, value account, credit account, listed company, state-owned enterprise / central enterprise, top 100, management team, and operation team, etc.; the relationship types include legal association, senior management association, shareholder association, investment / subsidiary association, and common investment, etc.; the establishment date of the new customer can be selected from the recent 3 days, the recent 7 days, the recent 30 days, the recent 90 days, etc.

[0062] For example, if a user queries from the existing new customers (assuming 10,000) based on the query tags “listed company, shareholder association, recent 30 days, A city”, there may be 100 new customers that meet the received query tags, the relationship of these 100 new customers with the inventory customers includes shareholder association, and the associated inventory customers are listed companies, and the new customers are enterprises established in A city in the recent 30 days.

[0063] Specifically, as shown in Figure 4 The data source of the new customer can be the new registered customer data of multiple suppliers obtained through Kafka, data interface, etc., combined with the data provided by the customer business handling and user feedback, and obtained through multi-source cross verification. Through multi-source data integration, cleaning, cross verification, the new registered customer data obtained can be checked according to program sampling, user feedback, operation and maintenance, etc. to improve the coverage and accuracy of the new registered customer data.

[0064] S20: calculating the value score of the new customer based on the preset value score of the inventory customer and the first association relationship between the inventory customer and the new customer;

[0065] Specifically, the preset value score of the inventory customer can be obtained by the business personnel, and can also be determined based on the customer type of the inventory customer; that is, the value of the inventory customer is known, and it can be understood that if the new customer has a higher association with the inventory customer with a higher value, the value score of the new customer can also be higher, but the specific value of the value score of the new customer still needs to be objectively calculated.

[0066] The calculation of the value score of the new customer can be: determining the similarity between the new customer and the inventory customer based on the preset value score of the inventory customer, the first association relationship between the inventory customer and the new customer, and the customer type of the new customer, calculating the value score of the new customer based on the similarity and the preset value score of the associated inventory customer, for example, if the inventory customer is 90 points, the similarity is 90%, and the new customer scores 81 points.

[0067] To more accurately and more quickly calculate the value score of the new customer, the calculating of the value score of the new customer can also be: determining a plurality of first customer type labels of the inventory customers based on a preset mapping relationship between relationships and labels; calculating the value score of the new customer based on the plurality of first customer type labels, a preset value score, and a first association relationship between the inventory customers and the new customer, through a preset relationship score model.

[0068] The preset relationship score model is used to represent a second association relationship between the inventory customers and the value score of the new customer.

[0069] It should be noted that since only the preset value score of the inventory customers and the first association relationship between the inventory customers and the new customer are used to calculate the value score of the new customer, the characteristics of the inventory customers are considered less, that is, the preset value score can indirectly represent the characteristics of the inventory customers, therefore, in order to improve the accuracy of the calculated value score of the new customer, the embodiment also needs to consider the customer type label, because the customer type label can more directly represent the characteristics of the inventory customers.

[0070] Specifically, the customer type label can be obtained by analyzing and classifying the data of the inventory customers, for example, the customer type label can include strategic customers, listed companies, state-owned companies, tax top 100, industries, etc. The customer type label can also include labels defined by business personnel, such as high-quality customers, branch loyal customers, etc.

[0071] Further, since there are a plurality of customer type labels in total, but different customer type labels have different degrees of association with different inventory customers, therefore, before the step of calculating the value score of the new customer, a plurality of first customer type labels of the inventory customers need to be determined based on a preset mapping relationship between relationships and labels; wherein the first customer type label is at least one of the plurality of customer type labels.

[0072] It should also be noted that the association relationship between the inventory customers and the new customer includes legal association, senior management association, shareholder association, investment / subsidiary, common investment, supplier, distributor, etc. It can also include high-quality customer associations defined by business personnel, such as Figure 4As shown, by analyzing the inventory customer (enterprise) data and the newly registered customer (enterprise) data, the association relationship between the inventory customers and the newly registered customers is classified, a data structure that can be stored in a graph database is formed, and saved into the graph database. When used, a user query system data model is obtained, and the user inputs a new customer query instruction in the application interface. The query instruction includes three parts: the customer type and the preset value score of the inventory customer, and the type of the association relationship. Among them, the customer type of the inventory customer and the type of the association relationship are optional. If not specified, it degenerates into a general new customer information search system, and the query result is only the customer type and the pre-calculated value score of the new customer itself. The first association relationship is at least one of the above-mentioned multiple association relationships.

[0073] Therefore, based on the multiple first customer type labels, the preset value score, and the first association relationship between the inventory customer and the new customer, a more accurate value score of the new customer can be calculated through a preset relationship score model.

[0074] Specifically, before the step of calculating the value score of the new customer based on the multiple first customer type labels, the preset value score, and the first association relationship between the inventory customer and the new customer through the preset relationship score model, a relationship score model needs to be obtained.

[0075] The above-mentioned relationship score model can be obtained by the following way: obtaining the second customer type label, the historical value score of the historical inventory customer, and the third association relationship between the historical inventory customer and the historical new customer; based on the second customer type label, the historical value score, and the third association relationship, constructing a customer relationship graph; based on the customer relationship graph and the first score label of the historical new customer, iteratively training a preset to-be-trained graph neural network model to obtain a relationship score model that meets the accuracy condition.

[0076] Specifically, the historical inventory customer is an inventory customer existing in the enterprise database at a historical moment corresponding to the user query moment. The historical new customer is a new customer compared with the historical inventory customer at the historical moment. The second customer type label of the historical inventory customer is also at least one of the above-mentioned multiple customer type labels. The calculation method of the historical value score is similar to that of the above-mentioned preset value score, which will not be repeated here. The first association relationship is at least one of the above-mentioned multiple association relationships.

[0077] Specifically, the relationship scoring model can be a graph neural network model, and specifically, the graph neural network model can be a GCN (Graph Convolutional Network) or a GAT (Graph Attention Network).

[0078] Specifically, the graph neural network model can learn graph structure data, better extract and explore features in the graph structure data, and better extract features for the inventory customers and new customers with complex association relationships in the embodiment. Therefore, using the graph neural network model can make the scoring result more accurate.

[0079] Using the graph neural network model requires first constructing a customer relationship graph. Specifically, the customer relationship graph can be constructed based on the second customer type label, the historical value score, and the third association relationship. The customer relationship graph includes the third association relationship, the features of the historical inventory customers, and the value (historical value score) of the historical inventory customers. Based on the customer relationship graph and the first score label (historical new customer value score) of the historical new customers, the pre-set graph neural network model is iteratively trained to better extract features, thereby obtaining a relationship scoring model that meets the accuracy condition.

[0080] Specifically, the implementation of constructing the customer relationship graph based on the second customer type label, the historical value score, and the third association relationship can be: regarding the historical inventory customers and the historical new customers as nodes; regarding the third association relationship between the historical inventory customers and the historical new customers as directed edges; encapsulating features for the nodes based on the second customer type label and the historical value score; and constructing the customer relationship graph based on the directed edges and the nodes after encapsulating the features.

[0081] The constructed customer relationship graph can better express the third association relationship between the historical inventory customers and the historical new customers and the influence of the features of the historical inventory customers on the historical new customers.

[0082] Since the customer type label of the inventory customers is in a fixed range and is relatively homogeneous, for example, multiple inventory customers can have labels such as whether to be listed or how much deposit, the GCN (Graph Convolutional Network) is more suitable for graph data with relatively homogeneous nodes and edges than the GAT (Graph Attention Network), can learn the aggregation representation of local neighbors, the content represented by the GCN is relatively simple, the calculation rate is relatively fast, and the hyperparameters are more stable. Therefore, the embodiment uses the GCN as the to-be-trained model.

[0083] The specific implementation of the above iteration training of the preset to-be-trained graph neural network model based on the customer relationship graph and the first score label of the historical new customer to obtain a relationship scoring model meeting the accuracy condition can be: obtaining a weight feature corresponding to different types of third association relationships; and iteration training of the preset to-be-trained graph convolutional neural network model based on the customer relationship graph, the first score label of the historical new customer, and the weight feature to obtain a relationship scoring model meeting the accuracy condition.

[0084] Since there are various association relationships between inventory customers and new customers, some association relationships are relatively important or have a high degree of closeness, and some association relationships are relatively unimportant or have a low degree of closeness. Therefore, different association relationships have different influence weights on the scoring of new customers. For example, the influence weight of shareholder association on the scoring of new customers is higher than that of common investment, a higher weight is set for shareholder association, that is, an influence factor of the scoring of new customers is increased, so that the scoring of new customers calculated by the relationship scoring model is more in line with the actual value of the new customers.

[0085] S30: determining a target new customer based on the value score of the new customer;

[0086] Based on the value score of the new customer calculated above, it can be determined which new customers have the value of mining and which customers do not have the value of mining at the moment. Specifically, the new customers with a value score higher than a preset score can be determined as target new customers, and the value scores of the new customers can be sorted from high to low, and the top n new customers can be determined as target new customers.

[0087] S40: taking the target new customer as a query result and returning the query result.

[0088] After obtaining the target new customer, the target new customer is taken as a query result and the query result is returned. The target new customers can be sorted according to the value scores corresponding to the target new customers from high to low and displayed on a page of the query result, so that the user can intuitively see which new customers can be preferentially mined, thereby providing valuable reference for the user.

[0089] Specifically, a list of target new customers meeting the condition is returned, and the list items of the list include basic business information and value scores of the target new customers, associated inventory customers, types of association relationships, preset value scores of the inventory customers, and system recommended marketing clues and strategies. If the customer types of the inventory customers and the types of the association relationships are not specified in the query label, the associated customers and association relationships are not included in the returned result, and only the basic business information of the new customers is included.

[0090] The customer mining platform can also have a recommendation subscription module. Based on the customer types and preset value scores of existing customers selected by the user, as well as the customer types and value scores of new customers, the customer mining platform will push the calculated target new customers that meet the conditions set by the user to the user.

[0091] This application identifies new customers matching received query tags; calculates the new customer's value score based on the preset value scores of existing customers and the first association relationship between the existing customers and the new customer; objectively evaluates the new customer's value score without manual analysis of whether the new customer is valuable or manual comparison of existing customers with new customers. Furthermore, based on the customer relationship graph and the first rating tags of historical new customers, a preset graph neural network model to be trained is iteratively trained to obtain a relationship scoring model that meets the accuracy requirements. Based on the relationship scoring model, the new customer's value score is calculated, improving the accuracy and computational efficiency of the new customer's value score. Finally, based on the new customer's value score, target new customers are identified; these target new customers are used as query results and returned, allowing users to obtain target new customers by viewing the query results. Therefore, this application can improve customer acquisition efficiency. By comprehensively considering the customer type tags of existing customers and the preset value scores, the accuracy of the new customer's value score is improved, effectively increasing the success rate of customer acquisition.

[0092] like Figure 5 As shown, based on the first embodiment, a second embodiment of the cell handover method of this application is proposed. In this embodiment, before step S20, the method further includes the following steps:

[0093] A1: Based on the third customer type tag of the existing customers, the value score of the existing customers is calculated through a preset value scoring model, wherein the preset value scoring model is used to characterize the fourth association relationship between the third customer type tag and the value score.

[0094] It should be noted that the calculation of the value score of the existing customers can be achieved by: calculating the value score of the existing customers based on the third customer type tag of the existing customers through a preset value scoring model.

[0095] Specifically, the preset value scoring model can be an item-based collaborative filtering algorithm or a decision tree model. Since a decision tree model only needs to be built once and can be used repeatedly, the maximum number of calculations for each prediction does not exceed the depth of the decision tree model. This avoids redundant calculations, improves efficiency, and allows for rapid decision generation. This embodiment chooses to calculate the value score of the existing customers using a decision tree model.

[0096] Specifically, before the step of calculating the value score of the inventory customer based on the third customer type label of the inventory customer through a preset value score model, the third customer type label of a historical inventory customer can be obtained first; then the second score label of the historical inventory customer is determined based on the third customer type label of the historical inventory customer; finally, the preset decision tree model to be trained is iteratively trained based on the third customer type label of the historical inventory customer and the second score label, to obtain a target value score model meeting the accuracy condition.

[0097] It can be understood that for 10,000 inventory customers, their third customer type labels (at least one of the above customer type labels) are obtained, and the third customer type label data is preprocessed, the preprocessing process including: checking the third customer type label, processing missing values, abnormal values, etc., and onehot encoding the third customer type label.

[0098] Based on the preprocessed third customer type label, the second score label of the historical inventory customer is determined, specifically, the second score label is the value score of the historical inventory customer. Based on the third customer type label (training set) of the historical inventory customer and the second score label (training label corresponding to the training set), the preset decision tree model to be trained is iteratively trained (in the iterative training process, the depth of the decision tree, the minimum number of leaf nodes, and other parameters are adjusted constantly to avoid overfitting and select the optimal parameters), to obtain a target value score model meeting the accuracy condition.

[0099] It should be noted that the third customer type label includes at least one of the following: listed company, group customer, deposit size, loan size, and tax amount, and the preset decision tree model to be trained includes at least one of the following: ID3 model, C4.5 model, and CART model.

[0100] Specifically, after the step of obtaining the third customer type label of the historical inventory customer, the third customer type labels of the same type can be combined based on the type of the third customer type label to obtain a fourth customer type label; the second score label of the historical inventory customer is determined based on the fourth customer type label of the historical inventory customer; at least one of the ID3 model, the C4.5 model, and the CART model is iteratively trained based on the fourth customer type label of the historical inventory customer and the second score label, to obtain a plurality of value score models meeting the accuracy condition; the plurality of value score models are verified based on a preset verification set to obtain verification results, and the target value score model is selected based on the verification results.

[0101] Specifically, based on the type of the third customer type label, the third customer type labels of the same type are combined to construct a more predictive feature, for example, a legal relationship and registered capital are combined into a new feature combination, or a more general third customer type label can be split into two new customer type labels.

[0102] Based on the fourth customer type label of the historical inventory customer, a second score label of the historical inventory customer is determined; based on the fourth customer type label of the historical inventory customer and the second score label, at least one of the ID3 model, the C4.5 model and the CART model is iteratively trained (in the training process, useless features (cooperation amount in the cooperation directory of the dealer has less influence) can be removed), to obtain a plurality of value scoring models that meet the accuracy condition; based on a preset validation set (part of the fourth customer type label), the plurality of value scoring models are verified respectively to obtain a verification result, and based on the verification result, the target value scoring model is selected.

[0103] Specifically, the effects of different decision tree models can be compared based on evaluation indicators such as accuracy, recall rate, F1 value, etc., and a decision tree model with the best effect is selected. That is, based on the fourth customer type label of the inventory customer, the value score of the inventory customer is calculated through the target value scoring model.

[0104] In this embodiment, the value score of the inventory customer is calculated by the decision tree model, which can make the value score of the inventory customer more objective, thereby further improving the accuracy of the value score of the new customer.

[0105] In addition, the embodiment of the application also provides a customer mining device, which refers to Figure 6 , the customer mining device comprises:

[0106] The inventory customer determination module 10 is configured to determine a new customer that meets the received query label.

[0107] The score calculation module 20 is configured to calculate the value score of the new customer based on the preset value score of the inventory customer and the first association relationship between the inventory customer and the new customer.

[0108] The new customer determination module 30 is configured to determine a target new customer based on the value score of the new customer.

[0109] The result returning module 40 is configured to return the target new customer as a query result.

[0110] The embodiment can determine a new customer meeting a received query label, calculate a value score of the new customer based on a preset value score of an inventory customer and a first correlation between the inventory customer and the new customer, objectively evaluate the value score of the new customer without manually analyzing whether the new customer is valuable or comparing the inventory customer with the new customer, determine a target new customer based on the value score of the new customer, return the target new customer as a query result, and enable a user to obtain the target new customer by viewing the query result. Therefore, the application can improve customer mining efficiency.

[0111] It should be noted that each module in the device can be used to implement each step in the above method, and achieve the corresponding technical effects. The present embodiment will not be described here.

[0112] Referring to Figure 7 , Figure 7 The device is a structure diagram of a hardware running environment involved in an embodiment of the application.

[0113] As Figure 7 shown, the device can include a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to realize the connection and communication between the components. The user interface 1003 can include a display screen (Display) and an input unit such as a keyboard (Keyboard). The optional user interface 1003 can also include a standard wired interface, a wireless interface. The network interface 1004 can optionally include a standard wired interface, a wireless interface (such as a WI-FI interface). The memory 1005 can be a high-speed RAM memory or a stable memory (non-volatile memory), such as a magnetic disk memory. The memory 1005 can also be an independent storage device from the aforementioned processor 1001.

[0114] Those skilled in the art can understand that Figure 7 the structure shown in the above embodiment does not constitute a limitation on the device, and can include more or fewer components than the diagram, or combine certain components, or different component arrangements.

[0115] As Figure 7 shown, the memory 1005 as a computer storage medium can include an operating system, a network communication module, a user interface module, and a customer mining program.

[0116] In Figure 7In the illustrated device, the network interface 1004 is mainly used for data communication with an external network; the user interface 1003 is mainly used for receiving user input instructions; the device calls the customer mining program stored in the memory 1005 through the processor 1001, and performs the following operations:

[0117] Determine new customers that meet the received query label;

[0118] Calculate the value score of the new customer based on the preset value score of the inventory customer and the first association relationship between the inventory customer and the new customer;

[0119] Determine the target new customer based on the value score of the new customer;

[0120] The target new customer is taken as a query result, and the query result is returned.

[0121] Further, the processor 1001 can call the customer mining program stored in the memory 1005, and further perform the following operations:

[0122] Determine a plurality of first customer type labels of the inventory customer based on a preset mapping relationship between relationships and labels;

[0123] Calculate the value score of the new customer based on the plurality of first customer type labels, the preset value score, and the first association relationship between the inventory customer and the new customer through a preset relationship score model, wherein the preset relationship score model is used to represent a second association relationship between the value score of the inventory customer and the new customer.

[0124] Further, the processor 1001 can call the customer mining program stored in the memory 1005, and further perform the following operations:

[0125] Obtain the second customer type label, the historical value score of the historical inventory customer, and the third association relationship between the historical inventory customer and the historical new customer;

[0126] Construct a customer relationship graph based on the second customer type label, the historical value score, and the third association relationship;

[0127] Iteratively train a preset to-be-trained graph neural network model based on the customer relationship graph and the first score label of the historical new customer, to obtain a relationship score model that meets the accuracy condition.

[0128] Further, the processor 1001 can call the customer mining program stored in the memory 1005, and further perform the following operations:

[0129] Take the historical inventory customer and the historical new customer as nodes;

[0130] a third association relationship between the historical inventory customer and the historical new customer as a directed edge;

[0131] packaging features for the node based on the second customer type label and the historical value score;

[0132] constructing a customer relationship graph based on the directed edge and the node after packaging the features;

[0133] The step of iteratively training a preset graph neural network model based on the customer relationship graph and the first score label of the historical new customer to obtain a relationship scoring model that meets the accuracy condition comprises:

[0134] obtaining weight features corresponding to different types of third association relationships;

[0135] iteratively training a preset graph convolutional neural network model based on the customer relationship graph, the first score label of the historical new customer, and the weight features to obtain a relationship scoring model that meets the accuracy condition.

[0136] Further, the processor 1001 can invoke the customer mining program stored in the memory 1005 and further perform the following operations:

[0137] calculating a value score of the inventory customer based on a third customer type label of the inventory customer through a preset value scoring model, wherein the preset value scoring model is used to represent a fourth association relationship between the third customer type label and the value score.

[0138] Further, the processor 1001 can invoke the customer mining program stored in the memory 1005 and further perform the following operations:

[0139] obtaining a third customer type label of a historical inventory customer;

[0140] determining a second score label of the historical inventory customer based on the third customer type label of the historical inventory customer;

[0141] iteratively training a preset decision tree model based on the third customer type label of the historical inventory customer and the second score label to obtain a target value scoring model that meets the accuracy condition.

[0142] Further, the third customer type label includes at least one of a listed company, a group customer, a deposit size, a loan size, and a tax amount, and the preset decision tree model to be trained includes at least one of an ID3 model, a C4.5 model, and a CART model;

[0143] The processor 1001 can invoke the customer mining program stored in the memory 1005, and further perform the following operations:

[0144] According to the type of the third customer type label, the third customer type labels of the same type are combined to obtain a fourth customer type label;

[0145] Based on the fourth customer type label of the historical inventory customer, a second score label of the historical inventory customer is determined;

[0146] Based on the fourth customer type label of the historical inventory customer and the second score label, at least one of the ID3 model, the C4.5 model and the CART model is iteratively trained respectively, and a plurality of value score models satisfying the accuracy condition are obtained correspondingly;

[0147] Based on a preset verification set, the plurality of value score models are verified respectively to obtain a verification result, and based on the verification result, the target value score model is screened.

[0148] The embodiment determines the new customer meeting the received query label, calculates the value score of the new customer based on the preset value score of the inventory customer and the first association relationship between the inventory customer and the new customer, objectively evaluates the value score of the new customer, does not need to manually analyze whether the new customer is valuable, and does not need to manually compare the inventory customer and the new customer, determines the target new customer based on the value score of the new customer, takes the target new customer as the query result, and returns the query result, so that the user can obtain the target new customer by viewing the query result. Therefore, the application can improve the mining efficiency of the customer.

[0149] In addition, the application embodiment further provides a computer readable storage medium, and the computer readable storage medium stores a customer mining program. When the processor executes the customer mining program, the following operations are implemented:

[0150] Determine the new customer meeting the received query label;

[0151] Calculate the value score of the new customer based on the preset value score of the inventory customer and the first association relationship between the inventory customer and the new customer;

[0152] Determine the target new customer based on the value score of the new customer;

[0153] Take the target new customer as the query result, and return the query result.

[0154] The embodiment can improve the mining efficiency of the customer.

[0155] It should be noted that the computer readable storage medium described above can also implement each step in the above method when executed by the processor, and the corresponding technical effects are achieved, which will not be described herein.

[0156] It should be noted that in this document, the term "comprising" or "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or other elements inherent to such a process, method, article or system. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0157] The above embodiment numbers of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0158] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the methods described in various embodiments of the present application.

[0159] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation based on the content of the specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A client mining method characterized by comprising: The customer mining method comprises the following steps: determine new customers meeting the received query tags; calculate a value score of the new customers based on a preset value score of inventory customers and a first association relationship between the inventory customers and the new customers; determine target new customers based on the value score of the new customers; return the target new customers as query results; The step of calculating the value score of the new customers based on the preset value score of inventory customers and the first association relationship between the inventory customers and the new customers comprises: determine a plurality of first customer type tags of the inventory customers based on a preset mapping relationship between relationships and tags; calculate the value score of the new customers through a preset relationship score model based on the plurality of first customer type tags, a preset value score and the first association relationship between the inventory customers and the new customers, wherein the preset relationship score model is used to represent a second association relationship between the value score of the inventory customers and new customers; Before the step of calculating the value score of the new customers through a preset relationship score model based on the plurality of first customer type tags, a preset value score and the first association relationship between the inventory customers and the new customers, the method further comprises: obtain second customer type tags, historical value scores of historical inventory customers and third association relationships between the historical inventory customers and historical new customers; treat the historical inventory customers and the historical new customers as nodes; treat the third association relationships between the historical inventory customers and the historical new customers as directed edges; encapsulate features for the nodes based on the second customer type tags and the historical value scores; construct a customer relationship graph based on the directed edges and the nodes after the encapsulation of the features; obtain weight features corresponding to different types of third association relationships; iteratively train a preset to-be-trained graph convolutional neural network model based on the customer relationship graph, first score tags of the historical new customers and the weight features to obtain a relationship score model meeting an accuracy condition.

2. The client mining method of claim 1, wherein, Before the step of calculating the value score of the new customers based on the preset value score of inventory customers and the first association relationship between the inventory customers and the new customers, the method further comprises: calculate the value score of the inventory customers through a preset value score model based on third customer type tags of the inventory customers, wherein the preset value score model is used to represent a fourth association relationship between the third customer type tags and the value score.

3. The client mining method of claim 2, wherein, Before the step of calculating the value score of the inventory customers through a preset value score model based on the third customer type tags of the inventory customers, the method further comprises: obtain third customer type tags of historical inventory customers; determine second score tags of the historical inventory customers based on the third customer type tags of the historical inventory customers; Iterative training is performed on the preset decision tree model to be trained based on the third customer type label of the historical inventory customer and the second score label, and a target value score model meeting an accuracy condition is obtained.

4. The client mining method of claim 3, wherein, The third customer type label includes at least one of a listed company, a group customer, a deposit size, a loan size, and a tax amount, and the preset decision tree model to be trained includes at least one of an ID3 model, a C4.5 model, and a CART model. After the step of obtaining the third customer type label of the historical inventory customer, the method further includes: Based on the type of the third customer type label, the third customer type labels of the same type are combined to obtain a fourth customer type label; Based on the fourth customer type label of the historical inventory customer, a second score label of the historical inventory customer is determined; Based on the fourth customer type label of the historical inventory customer and the second score label, at least one of the ID3 model, the C4.5 model, and the CART model is iteratively trained, and a plurality of value score models meeting the accuracy condition are correspondingly obtained; Based on a preset verification set, the plurality of value score models are verified to obtain a verification result, and based on the verification result, the target value score model is screened.

5. A client mining device for implementing the client mining method according to any one of claims 1 to 4, characterized by, The customer mining device includes: An inventory customer determination module configured to determine a new customer meeting a received query label; A score calculation module configured to calculate a value score of the new customer based on a preset value score of an inventory customer and a first association relationship between the inventory customer and the new customer; A new customer determination module configured to determine a target new customer based on the value score of the new customer; A result return module configured to return the target new customer as a query result.

6. An apparatus, comprising: The device includes a memory, a processor, and a customer mining program stored on the memory and executable on the processor, and the customer mining program is configured to implement the steps of the customer mining method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a customer mining program, and the customer mining program is executed by the processor to implement the steps of the customer mining method according to any one of claims 1 to 4.

Images