Potential customer mining method and device, electronic equipment and readable storage medium

Abstract

The application provides a potential customer mining method and device, electronic equipment and readable storage medium, wherein the method comprises the following steps: generating a first training sample set according to historical product purchase records and a customer information set, training an initial mining model based on the first training sample set, obtaining a first intermediate model, and obtaining a second training sample set based on the prediction result of the first intermediate model and product clustering information. The first intermediate model is trained based on the second training sample set to obtain a second intermediate model, and a third training sample set is obtained based on the prediction result of the second intermediate model. The second intermediate model is trained based on the third training sample set to obtain a target potential customer mining model, and then the third positive sample set is input into the target potential customer mining model for prediction to obtain potential customer information. The application trains the model based on the training sample for multiple times, finally obtains the target potential customer mining model, and thus the potential customer information is obtained, and the result accuracy is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of potential customer mining, in particular to a potential customer mining method and device, an electronic device and a readable storage medium. BACKGROUND

[0002] With the continuous development of computer technology, more and more enterprises begin to use computer technology to mine potential customers of the enterprise.

[0003] At present, there are two ways to mine potential customers of an enterprise. One way is not to distinguish specific products, but to use a general intelligent model to mine general potential customers and apply it to all products. Another way is to divide customer groups by business rules and recommend products according to the divided customer groups. For example, a certain product is divided into several themes, and several customers who meet the rules are manually assigned to each theme as the customer group of the theme.

[0004] However, the current method of mining potential customers has the problem of accuracy. SUMMARY

[0005] The present application aims to solve the problem of insufficient accuracy of mining results in the prior art by providing a potential customer mining method, device, electronic device and readable storage medium.

[0006] To achieve the above object, the technical scheme adopted by the embodiments of the present application is as follows:

[0007] In a first aspect, an embodiment of the present application provides a potential customer mining method, comprising:

[0008] generating a first training sample set according to historical product purchase records and a customer information set, the first training sample set including a first positive sample set and a first negative sample set, the first positive sample set including customer information of customers who have purchased any product, and the first negative sample set including customer information of customers who have not purchased any product;

[0009] training an initial mining model based on the first training sample set to obtain a first intermediate model, and obtaining a second training sample set based on the prediction result of the first intermediate model and product clustering information, the second training sample set including a second positive sample set and a second negative sample set, the second positive sample set including customer information of customers who have purchased each product category, and the second negative sample set including customer information of customers who have not purchased each product category;

[0010] training the first intermediate model based on the second training sample set to obtain a second intermediate model, and obtaining a third training sample set based on a prediction result of the second intermediate model, the third training sample set including a third positive sample set and a third negative sample set, the third positive sample set including customer information and product information of customers who have purchased each product, and the third negative sample set including customers who have not purchased each product and product information;

[0011] training the second intermediate model based on the third training sample set to obtain a target potential customer mining model;

[0012] inputting the third positive sample set into the target potential customer mining model for prediction to obtain potential customer information.

[0013] Optionally, the first training sample set is generated according to historical product purchase records and a customer information set, and includes:

[0014] determining purchased customers according to the historical product purchase records;

[0015] reading customer information of the purchased customers from the customer information set according to an identifier of the purchased customers, and adding the customer information of the purchased customers to the first positive sample set;

[0016] reading customer information of customers who do not belong to the purchased customers from the customer information set to obtain candidate non-purchased customer information;

[0017] performing under-sampling processing on the candidate non-purchased customer information to obtain the first negative sample set.

[0018] Optionally, the second training sample set is obtained based on a prediction result of the first intermediate model and product clustering information, and includes:

[0019] inputting each customer information in the first positive sample set into the first intermediate model to obtain a prediction result of the first intermediate model, the prediction result of the first intermediate model being used to indicate a probability of each customer purchasing any product;

[0020] determining a plurality of target customer information in the customer information set according to the prediction result of the first intermediate model;

[0021] clustering products in a product set to obtain a plurality of product categories;

[0022] obtaining the second training sample set according to the plurality of target customer information and the plurality of product categories.

[0023] Optionally, the plurality of target customer information in the customer information set is determined according to the prediction result of the first intermediate model, and includes:

[0024] ranking each customer information in the customer information set according to the probability indicated by the prediction result, to obtain a customer information sequence;

[0025] screening a first preset number of customer information in the customer information sequence as the plurality of target customer information according to the order of the customer information in the customer information sequence.

[0026] Optionally, the second training sample set is obtained according to the plurality of target customer information and the plurality of product categories, and the obtaining comprises:

[0027] determining at least one product category corresponding to each customer according to each target customer information;

[0028] generating the second positive sample set according to each target customer information and each product category corresponding to each customer;

[0029] reading negative customer information not belonging to the target customer information from the customer information set;

[0030] generating the second negative sample set according to the negative customer information.

[0031] Optionally, the third training sample set is obtained based on the prediction result of the second intermediate model, and the obtaining comprises:

[0032] inputting each customer information in the second positive sample set into the second intermediate model to obtain a prediction result of the second intermediate model, the prediction result of the second intermediate model being used to indicate a probability of each customer purchasing any product in each product category;

[0033] determining a second preset number of candidate customer information according to the prediction result of the second intermediate model;

[0034] obtaining the third training sample set according to each candidate customer information.

[0035] Optionally, the third training sample set is obtained according to each candidate customer information, and the obtaining comprises:

[0036] determining purchased products corresponding to each candidate customer information according to each candidate customer information and the historical product purchase record;

[0037] adding each candidate customer information and the purchased products corresponding to each candidate customer information as a positive sample to the third positive sample set;

[0038] reading customer information not belonging to the candidate customer information from the customer information set, and generating the third negative sample set according to the customer information not belonging to the candidate customer information.

[0039] In a second aspect, another embodiment of the present application provides a potential customer mining device, comprising:

[0040] a generating module configured to generate a first training sample set according to historical product purchase records and a customer information set, the first training sample set comprising a first positive sample set and a first negative sample set, the first positive sample set comprising customer information of customers who have purchased any product, and the first negative sample set comprising customer information of customers who have not purchased any product;

[0041] a first training module configured to train an initial mining model based on the first training sample set to obtain a first intermediate model, and obtain a second training sample set based on a prediction result of the first intermediate model and product cluster information, the second training sample set comprising a second positive sample set and a second negative sample set, the second positive sample set comprising customer information of customers who have purchased each product category, and the second negative sample set comprising customer information of customers who have not purchased each product category;

[0042] a second training module configured to train the first intermediate model based on the second training sample set to obtain a second intermediate model, and obtain a third training sample set based on a prediction result of the second intermediate model, the third training sample set comprising a third positive sample set and a third negative sample set, the third positive sample set comprising customer information and product information of customers who have purchased each product, and the third negative sample set comprising customer information and product information of customers who have not purchased each product;

[0043] a third training module configured to train the second intermediate model based on the third training sample set to obtain a target potential customer mining model;

[0044] a prediction module configured to input the third positive sample set into the target potential customer mining model for prediction to obtain potential customer information.

[0045] Optionally, the generating module is specifically configured to:

[0046] determine purchased customers according to the historical product purchase records;

[0047] read customer information of the purchased customers from the customer information set, and add the customer information of the purchased customers to the first positive sample set;

[0048] read customer information of customers who do not belong to the purchased customers from the customer information set to obtain candidate non-purchased customer information;

[0049] perform under-sampling processing on the candidate non-purchased customer information to obtain the first negative sample set.

[0050] Optionally, the first training module is specifically configured to:

[0051] input each customer information in the first positive sample set into the first intermediate model to obtain a prediction result of the first intermediate model, the prediction result of the first intermediate model being used to indicate a probability of each customer purchasing any product;

[0052] determine a plurality of target customer information in the customer information set according to the prediction result of the first intermediate model;

[0053] cluster products in the product set to obtain a plurality of product classes;

[0054] obtain the second training sample set according to the plurality of target customer information and the plurality of product classes.

[0055] Optionally, the first training module is specifically configured to:

[0056] sort each customer information in the customer information set according to the probability indicated by the prediction result to obtain a customer information sequence;

[0057] filter a first preset number of customer information in the customer information sequence as the plurality of target customer information according to an order of customer information in the customer information sequence.

[0058] Optionally, the first training module is specifically configured to:

[0059] determine at least one product class corresponding to each customer according to each target customer information;

[0060] generate the second positive sample set according to each target customer information and each product class corresponding to each customer;

[0061] read negative customer information not belonging to the target customer information from the customer information set;

[0062] generate the second negative sample set according to the negative customer information.

[0063] Optionally, the second training module is specifically configured to:

[0064] input each customer information in the second positive sample set into the second intermediate model to obtain a prediction result of the second intermediate model, the prediction result of the second intermediate model being used to indicate a probability of each customer purchasing any product in each product class;

[0065] determine a second preset number of candidate customer information according to the prediction result of the second intermediate model;

[0066] Optionally, obtain the third training sample set according to each candidate customer information.

[0067] The second training module is specifically configured to:

[0068] According to the candidate customer information and the historical product purchase record, determine the purchased product corresponding to each candidate customer information;

[0069] Add each candidate customer information and the purchased product corresponding to each candidate customer information to the third positive sample set as a positive sample;

[0070] Read customer information not belonging to the candidate customer information from the customer information set, and generate the third negative sample set according to the customer information not belonging to the candidate customer information.

[0071] In a third aspect, another embodiment of the present application provides an electronic device, comprising a processor, a storage medium and a bus, the storage medium stores machine readable instructions executable by the processor, when the electronic device is running, the processor and the storage medium communicate through the bus, and the processor executes the machine readable instructions to perform the steps of the method of the first aspect.

[0072] In a fourth aspect, another embodiment of the present application provides a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program is executed by a processor to perform the steps of the method of the first aspect.

[0073] The beneficial effects of the present application are: according to the historical product purchase record and the customer information set, a first training sample set is generated, an initial mining model is trained based on the first training sample set to obtain a first intermediate model, and a prediction result of the first intermediate model is obtained. Since the first positive sample set in the first training sample set includes customer information that has purchased any product, and the first negative sample set includes customer information that has not purchased any product, the prediction result of the first intermediate model can represent customer information in the customer information set that satisfies the condition of having purchased any product, thereby performing preliminary potential customer mining.

[0074] Based on the prediction result of the first intermediate model and the product clustering information, a second training sample set is obtained, the first intermediate model is trained based on the second training sample set to obtain a second intermediate model, and a prediction result of the second intermediate model is obtained. Since the second positive sample set in the second training sample set includes customer information that has purchased each product category, and the second negative sample set includes customer information that has not purchased each product category, the prediction result of the second intermediate model can represent customer information that has purchased each product category, thereby further mining potential customers in each product category based on preliminary potential customer mining.

[0075] Based on the prediction result of the second intermediate model, a third training sample set is obtained, the second intermediate model is trained based on the third training sample set, and a target potential customer mining model is obtained. Because the positive sample set in the third training sample set includes customer information and product information of customers who have purchased each product, and the negative sample set in the third training sample set includes customers who have not purchased each product and product information, the final prediction result of the target potential customer mining model can represent the information of potential customers of each product, so that further mining of potential customers of each product category is performed to obtain potential customer information of each product.

[0076] The target potential customer mining model is obtained through the initial mining model, the first intermediate model and the second intermediate model. Training by using the first sample set and prediction by using the trained first intermediate model can be regarded as overall rough classification. Generating the second sample set based on the prediction result of the first intermediate model and performing prediction by using the second intermediate model can be regarded as large-class rough classification. Generating the third sample set based on the prediction result of the second intermediate model and performing prediction by using the target potential customer model can be regarded as product fine classification. Through the three progressive steps of overall rough classification, large-class rough classification and product fine classification, the customer group range can be ladder-like reduced, so that the calculation resources are reduced to the greatest extent on the premise of not losing prediction accuracy, thereby realizing obtaining potential customer information of each product, improving accuracy and reducing the calculation burden. BRIEF DESCRIPTION OF DRAWINGS

[0077] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0078] Figure 1 is an application scenario diagram of a potential customer mining method provided by an embodiment of the present application;

[0079] Figure 2 is a flow diagram of a potential customer mining method provided by an embodiment of the present application;

[0080] Figure 3 is a flow diagram of a method for obtaining a first training sample set provided by an embodiment of the present application;

[0081] Figure 4 is a flow diagram of a method for obtaining a second training sample set provided by an embodiment of the present application;

[0082] Figure 5 is a flow diagram of a method for obtaining a plurality of target customer information provided by an embodiment of the present application;

[0083] Figure 6 is a specific flowchart of a method for obtaining a second training sample set provided by an embodiment of the present application;

[0084] Figure 7 is a flowchart of a method for obtaining a third training sample set provided by an embodiment of the present application;

[0085] Figure 8 is a specific flowchart of a method for obtaining a third training sample set provided by an embodiment of the present application;

[0086] Figure 9 is a structural schematic diagram of a potential customer mining device provided by an embodiment of the present application;

[0087] Figure 10 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0088] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of description and illustration, and are not used to limit the protection scope of the present application. In addition, it should be understood that the schematic drawings are not drawn according to the actual proportions. The flowcharts show the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowcharts can not be implemented in sequence, and the steps without logical context relationship can be reversed in sequence or implemented simultaneously. In addition, one or more other operations can be added to the flowcharts or one or more operations can be removed from the flowcharts under the guidance of the content of the present application by those skilled in the art.

[0089] In addition, the described embodiments are only some of the embodiments of the present application, not all the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0090] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0091] At present, there are two ways to mine potential customers of an enterprise. One way is to mine general potential customers by using a general intelligent model without distinguishing specific products, and apply the general potential customers to all products. The other way is to divide customers into groups according to business rules, and recommend products to the customers in the groups. For example, a product is divided into several themes, and several customers who meet the rules are manually assigned to each theme as the customer group of the theme. However, both methods have the problem of insufficient accuracy.

[0092] In order to improve the accuracy, a general targeted potential customer mining model can be established to mine potential customers for each product. However, this method requires a huge computing cost. For example, potential customer mining needs to be performed for ten million customers and thousands of products, and hundreds of thousands of calculations are required for each prediction. Limited computing resources may make this method unfeasible.

[0093] Based on the above problems, the present application can generate a first training sample set according to historical product purchase records and a customer information set, train an initial mining model based on the first training sample, obtain a first intermediate model, obtain a second training sample set based on the prediction result of the first intermediate model and product clustering information, train the first intermediate model based on the second training sample set, obtain a second intermediate model, obtain a third training sample set based on the prediction result of the second intermediate model, train the second intermediate model based on the third training sample set, obtain a target potential customer mining model, and then input the third positive sample set into the target potential customer mining model for prediction to obtain potential customer information. Through the above method, the target potential customer mining model is obtained by training the model multiple times according to different samples, and the potential customer information is obtained, which can improve the accuracy while reducing the computing difficulty.

[0094] Next, the application scenario of the potential customer mining method in the present application is introduced. Referring to Figure 1 , a database can be used to store customer information and product information. When mining potential customers, the server can obtain customer information and product information from the database to generate a first training sample set. Then the server trains an initial mining model based on the first training sample to obtain a first intermediate model. Based on the prediction result of the first intermediate model and product clustering information obtained from the database, a second training sample set is obtained. Then the server trains the first intermediate model based on the second training sample set to obtain a second intermediate model, and obtains a third training sample set based on the prediction result of the second intermediate model. The second intermediate model is trained based on the third training sample set to obtain a target potential customer mining model, and then the third positive sample set is input into the target potential customer mining model for prediction to obtain potential customer information.

[0095] The potential customer mining method is introduced as follows: Figure 2 ​

[0096] S201, generating a first training sample set according to historical product purchase records and a customer information set, the first training sample set including a first positive sample set and a first negative sample set, the first positive sample set including customer information of customers who have purchased any product, and the first negative sample set including customer information of customers who have not purchased any product.

[0097] Optionally, the customer information set can include a large amount of customer information, and the customer information can include customer basic information, customer asset information, transaction and position information, and market information.

[0098] For example, the customer basic information can include customer gender, customer age, province, city, affiliated branch, business department, account opening time, customer type, and customer level, etc., and the customer basic information is used to represent the basic attributes of the customer.

[0099] The customer asset information can include customer net assets, asset balance, product market value, on-site product market value, off-site product market value, and holding market value of different types of products, and the customer asset information is used to represent the purchasing power and product preferences of the customer.

[0100] The transaction and position information can include information of products, stocks, and funds purchased and held by the customer in the past. For example, the product can be an ETF product, and the transaction and position information includes the number of times of purchasing ETF products, average amount, average risk level, standard deviation, number of times of purchasing each product type ETF, number of times of purchasing each asset type ETF, average yield, average yield of similar funds, product Sharpe ratio average, product maximum drawdown rate average, product alpha index average, product beta index average, and product unit net value average of the customer in the past six months. The transaction and position information is used to represent the product preferences and risk preferences of the customer.

[0101] It is worth mentioning that the customer information can also include customer stock, fund, and customer position data, and other related information.

[0102] Optionally, according to the historical product purchase records, customer information of all customers who have purchased any product in the customer information set is taken as the first positive sample set, and customer information of all customers who have not purchased any product in the historical product purchase records is taken as the first negative sample set.

[0103] S202, training the initial mining model based on the first training sample set to obtain a first intermediate model, and obtaining a second training sample set based on a prediction result of the first intermediate model and product clustering information, the second training sample set including a second positive sample set and a second negative sample set, the second positive sample set including customer information of customers who have purchased each product category, and the second negative sample set including customer information of customers who have not purchased each product category.

[0104] Optionally, before training the initial mining model based on the first training sample set, the customer information in the first training sample set can also be processed through feature engineering. For example, feature engineering can include missing value filling and one-hot encoding of discrete variables. The missing value filling can fill the missing part of the customer information, ensuring the stability of the training result. The one-hot encoding of discrete variables can numerically process the discrete customer information, ensuring the accuracy of the training result.

[0105] Optionally, the initial mining model can be a random forest model. By using the high accuracy and parallelism of the random forest model, the first training sample set is input into the random forest model for training to obtain the first intermediate model. The first intermediate model can obtain the probability of each customer purchasing any product, sort the probabilities from high to low, and the prediction result of the first intermediate model can indicate that the top pre-set number of customers are screened to obtain the second training sample set. For example, the prediction result of the first intermediate model can indicate 500,000 customers.

[0106] Optionally, the first intermediate model can be tuned by a random grid to change the depth, maximum number of features, and minimum number of leaf nodes of the first intermediate model, thereby reducing the computational complexity of the first intermediate model.

[0107] Optionally, the second training sample set can be obtained based on the prediction result of the first intermediate model and the product clustering information, and in combination with the customer information.

[0108] Optionally, the product clustering information can be obtained by classifying a plurality of products. Specifically, since some products in the plurality of products have strong similarity, the products with strong similarity can be classified into a product category, and the products in the product category have certain commonality.

[0109] S203, training the first intermediate model based on the second training sample set to obtain a second intermediate model, and obtaining a third training sample set based on a prediction result of the second intermediate model, the third training sample set including a third positive sample set and a third negative sample set, the third positive sample set including customer information and product information of customers who have purchased each product, and the third negative sample set including customer information and product information of customers who have not purchased each product.

[0110] Optionally, the first intermediate model can be used as the second intermediate model after the first intermediate model is trained, and the second intermediate model can also be a random forest model. The second intermediate model can obtain the probability of each customer purchasing any product in each product category, sort the probabilities from high to low, obtain the prediction result of the second intermediate model according to the probability of each customer purchasing each product category, and determine the customer information of the potential customers of each product in each product category according to the prediction result.

[0111] Optionally, the third training sample set is determined based on the prediction result of the second intermediate model, the third training sample set includes a third positive sample set and a third negative sample set, the third positive sample set includes customer information and product information of customers who have purchased each product, and it is worth noting that the customer information and product information of each product in the third positive sample set can refer to customer-product pairs with transaction records within a preset time. For example, the preset time can be three months. The third negative sample set includes customer-product pairs composed of each customer in the customer information set and any product that has not been purchased.

[0112] S204, training the second intermediate model based on the third training sample set to obtain a target potential customer mining model.

[0113] Optionally, the new customer-product pair can be obtained by calculating the inner product of the multiple customer information and the corresponding product information in the prediction result of the second intermediate model, and the new customer-product pair can represent the relationship between the customer and the product. The customer-product pair can be used as information in the third training sample set to train the second intermediate model and improve the accuracy of the model output result.

[0114] Optionally, the target potential customer mining model can also be a random forest model.

[0115] S205, inputting the third positive sample set into the target potential customer mining model to obtain potential customer information.

[0116] Optionally, before inputting the third positive sample set into the target potential customer mining model for prediction, the customer information in the third positive sample set can be screened to retain product-related features and remove product-unrelated features. Then, the screened third positive sample set is input into the target potential customer mining model.

[0117] Optionally, the target potential customer mining model can obtain the probability of each customer purchasing each product, sort the probabilities from high to low, determine the prediction result of the target potential customer mining model according to the top preset number of customers in the sorting, and use the customer information in the above prediction result as the potential customer information. For example, the customer information of the top 10,000 customers in the sorting can be used as the customer information included in the prediction result of the target potential customer mining model.

[0118] In this embodiment, a first training sample set is generated according to historical product purchase records and a customer information set, an initial mining model is trained based on the first training sample set to obtain a first intermediate model, and a prediction result of the first intermediate model is obtained. Since the first positive sample set in the first training sample set includes customer information of customers who have purchased any product, and the first negative sample set includes customer information of customers who have not purchased any product, the prediction result of the first intermediate model can represent customer information in the customer information set that meets the condition of having purchased any product, thereby performing preliminary potential customer mining.

[0119] Based on the prediction result of the first intermediate model and product clustering information, a second training sample set is obtained, the first intermediate model is trained based on the second training sample set to obtain a second intermediate model, and a prediction result of the second intermediate model is obtained. Since the second positive sample set in the second training sample set includes customer information of customers who have purchased each product category, and the second negative sample set includes customer information of customers who have not purchased each product category, the prediction result of the second intermediate model can represent customer information of customers who have purchased each product category, thereby further mining potential customers in each product category based on preliminary potential customer mining.

[0120] Based on the prediction result of the second intermediate model, a third training sample set is obtained, the second intermediate model is trained based on the third training sample set to obtain a target potential customer mining model. Since the positive sample set in the third training sample set includes customer information and product information of customers who have purchased each product, and the third negative sample set includes customer information and product information of customers who have not purchased each product, the final prediction result of the target potential customer mining model can represent potential customer information of each product, thereby further mining potential customers in each product category to obtain potential customer information of each product.

[0121] The target potential customer mining model is trained by the initial mining model, the first intermediate model, and the second intermediate model. Training using the first sample set and predicting using the trained first intermediate model can be regarded as overall rough sorting. Generating the second sample set based on the prediction result of the first intermediate model and predicting using the second intermediate model can be regarded as large category rough sorting. Generating the third sample set based on the prediction result of the second intermediate model and predicting using the target potential customer model can be regarded as product fine sorting. Through the three progressive steps of overall rough sorting, large category rough sorting, and product fine sorting, the customer group range can be ladder-like reduced, thereby reducing the calculation resources to the greatest extent on the premise of basically not losing prediction accuracy, thereby realizing obtaining potential customer information of each product, improving accuracy, and reducing the calculation burden.

[0122] The following refers to Figure 3The first training sample set is generated according to the historical product purchase record and the customer information set in step S201, and will be described in detail as follows:

[0123] S301, determining a purchased customer according to the historical product purchase record.

[0124] Optionally, the historical product purchase record can be obtained from a database. Specifically, if a customer has purchased any product, the historical product purchase record stores the corresponding identification of the customer and the product, and if the customer has not purchased any product, the historical product purchase record does not store the corresponding identification of the customer and the product, or the product item corresponding to the customer is empty.

[0125] S302, reading the customer information of the purchased customer from the customer information set according to the identification of the purchased customer, and adding the customer information of the purchased customer to the first positive sample set.

[0126] Optionally, the customer information of the purchased customer is read from the customer information set according to the corresponding identification of the customer and the product in the historical product purchase record. The customer information set contains information of all customers in the database. For example, if the potential customers of various products of enterprise A are mined, the customer information set includes the customer information of all customers of enterprise A.

[0127] S303, reading the customer information of a customer who does not belong to the purchased customer from the customer information set to obtain candidate non-purchased customer information.

[0128] Optionally, the customer information in which the corresponding identification of the customer and the product is not stored in the historical product purchase record, or the product item corresponding to the customer is empty, is selected to obtain the candidate non-purchased customer information.

[0129] S304, performing undersampling processing on the candidate non-purchased customer information to obtain the first negative sample set.

[0130] Optionally, since the customer information of customers who have not purchased any product may be much more than the customer information of customers who have purchased any product, the candidate non-purchased customer information may be much more than the first positive sample set. At this time, undersampling processing needs to be performed on the candidate non-purchased customer information to obtain the first negative sample set. Specifically, undersampling processing can realize class distribution of multiple sets of classification data by eliminating part of the data. For example, the first positive sample set includes 100 customer information, and the candidate non-purchased customer information is 10000 customer information. Therefore, undersampling processing needs to be performed on the candidate non-purchased customer information to make the candidate non-purchased customer information 1000, so as to ensure that the ratio of the first positive sample set and the first negative sample set is reasonable and the calculation workload is reduced.

[0131] In this embodiment, by determining the purchased customers according to the historical product purchase records, reading the customer information of the purchased customers from the customer information set according to the identification of the purchased customers, and adding the customer information of the purchased customers to the first positive sample set, the customer information of the non-purchased customers is read from the customer information set to obtain the candidate non-purchased customer information, and the under-sampling processing is performed on the candidate non-purchased customer information to obtain the first negative sample set, so as to ensure that the proportion of the first positive sample set and the first negative sample set is reasonable, and the calculation workload is reduced.

[0132] After introducing how to obtain the first positive sample set and the first negative sample set, refer to Figure 4 The second training sample set is obtained based on the prediction result of the first intermediate model and the product clustering information in the above step S202, and is introduced as follows:

[0133] S401, input each customer information in the first positive sample set into the first intermediate model to obtain the prediction result of the first intermediate model, and the prediction result of the first intermediate model is used to indicate the probability of each customer purchasing any product.

[0134] Optionally, each customer information in the first positive sample set can include the customer information that has purchased any product.

[0135] Optionally, the prediction result of the first intermediate model can indicate the probability of each customer purchasing any product. For example, the probability of A customer purchasing any product is 90%, the probability of B customer purchasing any product is 80%, the probability of C customer purchasing any product is 40%, and the probability of D customer purchasing any product is 60%.

[0136] S402, determine a plurality of target customer information in the customer information set according to the prediction result of the first intermediate model.

[0137] Optionally, according to the prediction result of the first intermediate model, a preset number of customers with high probability of purchasing any product are regarded as potential customers who may purchase products. For example, referring to the example in the above step S401, if 2 potential customers are needed, the target customers can be A customer and B customer.

[0138] Optionally, the higher the preset number is, the more complex the calculation is, and the lower the preset number is, the easier the calculation is.

[0139] S403, clustering the products in the product set to obtain a plurality of product classes.

[0140] Optionally, the products are clustered according to the similarity of each product in the product set to obtain a plurality of product classes.

[0141] As a possible implementation, the products can be clustered using a K-means clustering method. The K-means clustering method finds a partition scheme of K clusters according to the internal relations in the product information, so that the loss function corresponding to the clustering result is minimized. The clustering result includes all products in a product class.

[0142] For example, the K-means clustering method can divide the products according to the following features in the product information: product risk level, asset type, average yield (day / month / year), standard deviation of yield in the past year, average yield of the same type in the past year, average Sharpe ratio in the past year, average unit net value in the past year, management fee rate, issuance time, listing time, etc.

[0143] S404, obtaining a second training sample set according to the plurality of target customer information and the plurality of product classes.

[0144] Optionally, according to the customer information of the customers who are likely to purchase the products and the plurality of product classes, the customer information of the customers who have purchased any product in each product class is distinguished as a second positive sample set of the second training sample set. The customer information of the customers who have not purchased the products is taken as a second negative sample set.

[0145] Optionally, whether a customer has purchased a product class can be determined according to the historical transaction records of the products of the customer in the customer information, and the customer information of the customers who have purchased each product class is taken as the second positive sample set, and the customer information of the customers who have not purchased each product class is taken as the second negative sample set.

[0146] For example, A customer has purchased a product class a, but has not purchased product classes b and c, B customer has purchased product classes a and c, but has not purchased product class b, and C customer has purchased product class c, but has not purchased product classes a and b. Therefore, the second positive sample set includes: A customer and B customer in product class a, no customer in product class b, and B customer and C customer in product class c. The second negative sample set can include: C customer in product class a, A customer, B customer and C customer in product class b, and A customer in product class c.

[0147] It is worth noting that the customer information in the second negative sample set can be much more than the customer information in the second positive sample set. Therefore, the second negative sample set can be under-sampled to keep the second negative sample set balanced with the second positive sample set.

[0148] In this embodiment, by inputting each customer information in the first positive sample set into the first intermediate model, the prediction result of the first intermediate model is obtained, a plurality of target customer information in the customer information set is determined according to the prediction result of the first intermediate model, and the products in the product set are clustered to obtain a plurality of product categories. Then, according to the plurality of target customer information and the plurality of product categories, a second training sample set is obtained, further mining of potential customers who are likely to purchase any product is realized, potential customers who are likely to purchase each product category are obtained, and thus the accuracy is improved while the computational burden is reduced.

[0149] The following refers to Figure 5 The step S402 of determining a plurality of target customer information in the customer information set according to the prediction result of the first intermediate model is described in detail as follows:

[0150] S501, according to the probability indicated by the prediction result, each customer information in the customer information set is sorted to obtain a customer information sequence.

[0151] Optionally, according to the probability of each customer purchasing any product indicated by the prediction result of the first intermediate model, the customer information of each customer in the customer information set is sorted to obtain a customer information sequence.

[0152] Optionally, the customers in the obtained customer information sequence can be sorted from high to low according to the probability indicated by the prediction result, and the customers in the front of the customer information sequence are more likely to purchase products.

[0153] S502, according to the order of the customer information in the customer information sequence, the first preset number of customer information in the customer information sequence is screened as a plurality of target customer information.

[0154] Optionally, the greater the first preset number, the more target customer information and the more potential customers preliminarily screened, but the higher the model calculation amount. The smaller the first preset number, the less target customer information and the less potential customers preliminarily screened, and the smaller the model calculation amount.

[0155] For example, referring to the example in the step S401, the first preset number is set to 3, and the target customers can be A customer, B customer and D customer.

[0156] In this embodiment, first, each customer information in the customer information set is sorted according to the probability indicated by the prediction result to obtain a customer information sequence. Then, according to the order of the customer information in the customer information sequence, the first preset number of customer information in the customer information sequence is screened as a plurality of target customer information. The target customer information with small quantity is screened from the customer information set with large quantity, so that the target customer has a great probability of purchasing products, and the burden of subsequent calculation is reduced, which ensures the accuracy of screening.

[0157] Next, refer to Figure 6 In the above step S404, the second training sample set is obtained according to the plurality of target customer information and the plurality of product categories, and the following is introduced:

[0158] S601, according to each target customer information, determine at least one product category corresponding to each customer.

[0159] Optionally, according to the historical product transaction record of each target customer information, at least one product class to which the customer has historically purchased is determined.

[0160] S602, according to each target customer information and each product category corresponding to each customer, generate a second positive sample set.

[0161] Optionally, in the second positive sample set, the same customer may be included in the plurality of product categories, or none of the customers may be included, in which case the product category is empty.

[0162] S603, read negative customer information from the customer information set which does not belong to the target customer information.

[0163] Optionally, based on the historical product transaction record in the customer information set, the customer information which has not purchased the product category can be selected as the negative customer information, or the customer information in the customer information set which does not belong to the target customer information can be selected as the negative customer information.

[0164] S604, generate a second negative sample set according to the negative customer information.

[0165] Optionally, all negative customer information is collected to obtain a second negative sample set. In the second negative sample set, the same customer may be included in the plurality of product categories, or none of the customers may be included, in which case the product category is empty.

[0166] In this embodiment, by determining at least one product category corresponding to each customer according to each target customer information, generating a second positive sample set according to each target customer information and each product category corresponding to each customer, the second positive sample set including customer information that has purchased each product category, and then reading negative customer information from the customer information set which does not belong to the target customer information, generating a second negative sample set according to the negative customer information, the second negative sample set including customer information that has not purchased each product category, thereby determining the potential customer range corresponding to each product category, laying a foundation for further potential customer mining of each product category, and improving the prediction accuracy.

[0167] Next, refer to Figure 7 The third training sample set is obtained based on the prediction result of the second intermediate model in the above step S203, and the following is explained:

[0168] S701, inputting each customer information in the second positive sample set into the second intermediate model to obtain a prediction result of the second intermediate model, the prediction result of the second intermediate model being used to indicate a probability of each customer purchasing any product in each product category.

[0169] Optionally, the each customer information in the second positive sample set can include customer information satisfying having purchased any product in each product category.

[0170] Optionally, the prediction result of the second intermediate model can indicate a probability of each customer purchasing each product category.

[0171] Illustratively, the probability of E customer purchasing a product category a is 20%, the probability of F customer purchasing a product is 35%, and the probability of G customer purchasing a product is 26%.

[0172] S702, determining a second preset number of candidate customer information according to the prediction result of the second intermediate model.

[0173] Optionally, according to the prediction result of the second intermediate model, the probabilities indicated by the prediction result of the second intermediate model are sorted, in each product category, a second preset number of customers with high probability are taken as potential customers, and customer information of the potential customers is taken as the candidate customer information.

[0174] Illustratively, if the second preset number is 2, the potential customers in the product category a are F customer and G customer.

[0175] S703, obtaining a third training sample set according to each candidate customer information.

[0176] Optionally, according to the historical transaction record in each candidate customer information, it is determined whether there is a record of purchasing a product in each product category in each candidate customer information, thereby obtaining the third training sample set.

[0177] In the embodiment, by inputting each customer information in the second positive sample set into the second intermediate model, a prediction result of the second intermediate model is obtained, the prediction result of the second intermediate model is used to indicate a probability of each customer purchasing any product in each product category, a second preset number of candidate customer information is determined according to the prediction result of the second intermediate model, and a third training sample set is obtained according to each candidate customer information, thereby determining potential customers of each product category according to the probability of each customer purchasing each product category, and further improving the accuracy of potential customer mining.

[0178] The following refers to Figure 8 The following describes the step S703 of obtaining a third training sample set according to each candidate customer information:

[0179] S801, determining a product that has been purchased corresponding to each candidate customer information according to each candidate customer information and historical product purchase record.

[0180] Optionally, the candidate customer information includes customer information of customers who have possibly purchased any product in each product category.

[0181] S802, add each candidate customer information and the purchased product corresponding to each candidate customer information to the third positive sample set as a positive sample.

[0182] Optionally, each candidate customer information and the purchased product corresponding to each candidate customer information are added to the third positive sample set as a customer-product pair, and candidate customer information and the purchased product in each product category are added to the third positive sample set as a customer-product pair.

[0183] For example, the candidate customers corresponding to a product category include P customer, Q customer and I customer, and the a product category includes X product, Y product and Z product. The P customer has purchased the X product, the Q customer has purchased the Y product, and the I customer has purchased the Z product. Then, three positive samples, i.e., P customer-X product, Q customer-Y product and I customer-Z product, can be generated. The three positive samples are added to the positive sample set of the a product category, and the positive sample sets of all product categories are added to the third positive sample set.

[0184] S803, read customer information not belonging to the candidate customer information from the customer information set, and generate the third negative sample set according to the customer information not belonging to the candidate customer information.

[0185] Optionally, one product not purchased by each customer not belonging to the candidate customer information in the customer information set can be randomly selected to form a customer-product pair as a negative sample, thereby generating the third negative sample set.

[0186] It is worth noting that, since the negative samples in the customer information set can be much more than the positive samples, the negative samples can be undersampled, and the third negative sample set can be generated according to the processed negative samples.

[0187] In this embodiment, the purchased product corresponding to each candidate customer information is determined according to each candidate customer information and the historical product purchase record, each candidate customer information and the purchased product corresponding to each candidate customer information are added to the third positive sample set as a positive sample, then the customer information not belonging to the candidate customer information is read from the customer information set, and the third negative sample set is generated according to the customer information not belonging to the candidate customer information, thereby obtaining the third training sample set, laying a foundation for training the second intermediate model, and thereby obtaining the potential customers corresponding to each product.

[0188] Based on the same inventive concept, the application also provides a potential customer mining device corresponding to the potential customer mining method. Since the device solves problems in the same principle as the potential customer mining method, the implementation of the device can be referred to the implementation of the method, and the repeated parts will not be described here.

[0189] Referring to Figure 9 FIG. 1 shows a schematic diagram of a potential customer mining device provided by the application, which includes a generation module 901, a first training module 902, a second training module 903, a third training module 904, and a prediction module 905; wherein:

[0190] The generation module 901 is configured to generate a first training sample set according to historical product purchase records and a customer information set, wherein the first training sample set includes a first positive sample set and a first negative sample set, the first positive sample set includes customer information of customers who have purchased any product, and the first negative sample set includes customer information of customers who have not purchased any product.

[0191] The first training module 902 is configured to train an initial mining model based on the first training sample set to obtain a first intermediate model, and obtain a second training sample set based on the prediction result of the first intermediate model and product clustering information, wherein the second training sample set includes a second positive sample set and a second negative sample set, the second positive sample set includes customer information of customers who have purchased each product category, and the second negative sample set includes customer information of customers who have not purchased each product category.

[0192] The second training module 903 is configured to train the first intermediate model based on the second training sample set to obtain a second intermediate model, and obtain a third training sample set based on the prediction result of the second intermediate model, wherein the third training sample set includes a third positive sample set and a third negative sample set, the third positive sample set includes customer information and product information of customers who have purchased each product, and the third negative sample set includes customer information and product information of customers who have not purchased each product.

[0193] The third training module 904 is configured to train the second intermediate model based on the third training sample set to obtain a target potential customer mining model.

[0194] The prediction module 905 is configured to input the third positive sample set into the target potential customer mining model for prediction to obtain potential customer information.

[0195] The generation module 901 is specifically configured to:

[0196] Determine customers who have purchased according to the historical product purchase records.

[0197] reading, from the customer information set, customer information of the purchased customer according to the identification of the purchased customer, and adding the customer information of the purchased customer to the first positive sample set;

[0198] reading, from the customer information set, customer information of a non-purchased customer, to obtain candidate non-purchased customer information;

[0199] performing undersampling processing on the candidate non-purchased customer information to obtain the first negative sample set.

[0200] The first training module 902 is specifically configured to:

[0201] inputting each customer information in the first positive sample set into the first intermediate model to obtain a prediction result of the first intermediate model, the prediction result of the first intermediate model being used to indicate a probability of each customer purchasing any product;

[0202] determining a plurality of target customer information in the customer information set according to the prediction result of the first intermediate model;

[0203] performing clustering on products in a product set to obtain a plurality of product classes;

[0204] obtaining the second training sample set according to the plurality of target customer information and the plurality of product classes.

[0205] The first training module 902 is specifically configured to:

[0206] sorting each customer information in the customer information set according to the probability indicated by the prediction result to obtain a customer information sequence;

[0207] filtering a first preset number of customer information in the customer information sequence as the plurality of target customer information according to an order of the customer information in the customer information sequence.

[0208] The first training module 902 is specifically configured to:

[0209] determining at least one product class corresponding to each customer according to each target customer information;

[0210] generating the second positive sample set according to each target customer information and each product class corresponding to each customer;

[0211] reading, from the customer information set, negative customer information that does not belong to the target customer information;

[0212] generating the second negative sample set according to the negative customer information.

[0213] The second training module 903 is specifically configured to:

[0214] The customer information from the second positive sample set is input into the second intermediate model to obtain the prediction result of the second intermediate model. The prediction result of the second intermediate model is used to indicate the probability of each customer purchasing any product in each product category.

[0215] The second preset number of candidate customer information is determined based on the prediction results of the second intermediate model;

[0216] The third training sample set is obtained based on the information of each candidate customer.

[0217] The second training module 903 is specifically used for:

[0218] Based on the information of each candidate customer and the historical product purchase records, determine the products purchased corresponding to each candidate customer.

[0219] Each candidate customer's information and the corresponding purchased products are added as a positive sample to the third positive sample set.

[0220] Read customer information that does not belong to the candidate customer information from the customer information set, and generate the third negative sample set based on the customer information that does not belong to the candidate customer information.

[0221] The processing flow of each module in the device and the interaction flow between each module can be referred to the relevant descriptions in the above method embodiments, and will not be detailed here.

[0222] This application also provides an electronic device, such as... Figure 10 The diagram shown is a schematic representation of an electronic device structure provided in an embodiment of this application, including: a processor 1001, a memory 1002, and optionally, a bus. The memory 1002 stores machine-readable instructions executable by the processor 1001 (e.g., ...). Figure 9 The device includes the execution instructions corresponding to the generation module 901, the first training module 902, the second training module 903, the third training model 904, and the prediction module 905. When the electronic device is running, the processor 1001 communicates with the memory 1002 via a bus. When the machine-readable instructions are executed by the processor 1001, the above-mentioned potential customer mining method is performed.

[0223] This application also provides a computer-readable storage medium storing a computer program, which is executed by the processor 1001 to perform the steps of the above-described potential customer mining method.

[0224] Those skilled in the art can clearly understand the specific working process of the system and the device described above for the convenience and brevity of description, which can refer to the corresponding process in the method embodiment, and will not be repeated in the present application. In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented by other ways. The device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and the actual implementation can have another division, for example, a plurality of modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some communication interface, device or module, which can be electrical, mechanical or other forms.

[0225] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. When the functions are realized in the form of software functional units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or say the part of the prior art or the part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0226] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. A method of excavating a cryptid, characterized by, The method comprises the following steps: generating a first training sample set according to historical product purchase records and a customer information set, the first training sample set comprising a first positive sample set and a first negative sample set, the first positive sample set comprising customer information of customers who have purchased any product, and the first negative sample set comprising customer information of customers who have not purchased any product; training an initial mining model based on the first training sample set to obtain a first intermediate model, inputting each customer information in the first positive sample set into the first intermediate model to obtain a prediction result of the first intermediate model, clustering products in the product set to obtain a plurality of product categories, and obtaining a second training sample set based on the prediction result of the first intermediate model and product clustering information; the second training sample set is obtained based on the prediction result of the first intermediate model and the product clustering information, which comprises inputting each customer information in the first positive sample set into the first intermediate model to obtain the prediction result of the first intermediate model, the prediction result of the first intermediate model being used to indicate the probability of each customer purchasing any product, selecting a first preset number of customers with high probability of purchasing any product as target customers according to the prediction result of the first intermediate model, and obtaining the second training sample set according to the target customer information and the plurality of product categories; the second training sample set is obtained according to the target customer information and the plurality of product categories, which comprises determining the product category to which at least one product purchased by each customer belongs according to the historical product transaction record of the customer in the information of each target customer, generating a second positive sample set according to the target customer information and the product category corresponding to each customer, the second positive sample set comprising customer information of customers who have purchased each product category, and generating a second negative sample set from negative customer information that does not belong to the target customer information according to the negative customer information; training the first intermediate model based on the second training sample set to obtain a second intermediate model, inputting each customer information in the second positive sample set into the second intermediate model to obtain a prediction result of the second intermediate model, and obtaining a third training sample set based on the prediction result of the second intermediate model, the third training sample set comprising a third positive sample set and a third negative sample set; the third training sample set is obtained based on the prediction result of the second intermediate model, which comprises inputting each customer information in the second positive sample set into the second intermediate model to obtain the prediction result of the second intermediate model, the prediction result of the second intermediate model indicating the probability of each customer purchasing each product category, sorting the probabilities indicated by the prediction result of the second intermediate model according to the prediction result of the second intermediate model, selecting a second preset number of customers with high probability as potential customers in each product category, and obtaining a third training sample set according to each candidate customer information; The third training sample set is obtained according to each candidate customer information, and includes the following steps: determining the purchased products corresponding to each candidate customer information according to each candidate customer information and historical product purchase records; forming a customer-product pair by combining the candidate customer information in each product category and the products in the product category purchased by the candidate customer information, and taking the customer-product pair as a third positive sample set; selecting a product that has not been purchased by each customer who does not belong to the candidate customer information from the customer information set to form a customer-product pair, taking the customer-product pair as a negative sample, thereby generating a third negative sample set; The second intermediate model is trained based on the third training sample set to obtain a target potential customer mining model; The third positive sample set is input into the target potential customer mining model for prediction to obtain potential customer information; the target potential customer mining model sorts the probability of each customer purchasing each product from high to low, and determines the prediction result of the target potential customer mining model according to a preset number of customers in the sequence, and takes the customer information in the prediction result of the target potential customer mining model as the potential customer information.

2. The method of claim 1, wherein, The first training sample set is generated according to the historical product purchase records and the customer information set, including: Determine the purchased customers according to the historical product purchase records; Read the customer information of the purchased customers from the customer information set according to the identification of the purchased customers, and add the customer information of the purchased customers to the first positive sample set; Read the customer information of customers who do not belong to the purchased customers from the customer information set to obtain candidate non-purchased customer information; Perform undersampling processing on the candidate non-purchased customer information to obtain the first negative sample set.

3. The method of claim 1, wherein, Determine the target customer information, including: Sort each customer information in the customer information set from high to low according to the probability indicated by the prediction result of the first intermediate model to obtain a customer information sequence; According to the order of the customer information in the customer information sequence, filter the first preset number of customer information in the customer information sequence as the target customer information.

4. A subterranean excavation apparatus characterized by, It includes: A generation module is configured to generate a first training sample set according to historical product purchase records and a customer information set, the first training sample set including a first positive sample set and a first negative sample set, the first positive sample set including customer information that has purchased any product, and the first negative sample set including customer information that has not purchased any product; The first training module is configured to train an initial mining model based on the first training sample set to obtain a first intermediate model, input each customer information in the first positive sample set into the first intermediate model to obtain a prediction result of the first intermediate model, cluster products in the product set to obtain a plurality of product categories, and obtain a second training sample set based on the prediction result of the first intermediate model and product clustering information, wherein the second training sample set includes a second positive sample set and a second negative sample set; the obtaining of the second training sample set based on the prediction result of the first intermediate model and the product clustering information includes: inputting each customer information in the first positive sample set into the first intermediate model to obtain the prediction result of the first intermediate model, wherein the prediction result of the first intermediate model is used to indicate a probability of each customer to purchase any product; according to the prediction result of the first intermediate model, a first preset number of customers with a high probability to purchase any product are selected as target customers; and according to the target customer information and the plurality of product categories, the second training sample set is obtained; the obtaining of the second training sample set according to the target customer information and the plurality of product categories includes: determining a product category to which at least one product that each customer has historically purchased belongs according to a transaction record of a historical product of the customer in the information of each target customer; generating the second positive sample set according to the target customer information and the product category corresponding to each customer, wherein the second positive sample set includes customer information that has purchased each product category; and reading negative customer information that does not belong to the target customer information from the customer information set, and generating the second negative sample set according to the negative customer information. The second training module is configured to train the first intermediate model based on the second training sample set to obtain a second intermediate model, input each customer information in the second positive sample set into the second intermediate model to obtain a prediction result of the second intermediate model, and obtain a third training sample set based on the prediction result of the second intermediate model, wherein the third training sample set includes a third positive sample set and a third negative sample set; the obtaining of the third training sample set based on the prediction result of the second intermediate model includes inputting each customer information in the second positive sample set into the second intermediate model to obtain the prediction result of the second intermediate model, wherein the prediction result of the second intermediate model indicates a probability of each customer purchasing each product category; sorting the probabilities indicated by the prediction result of the second intermediate model, and regarding a second preset number of customers with high probabilities in each product category as potential customers, and regarding the customer information of the potential customers as candidate customer information; and obtaining the third training sample set based on each candidate customer information; the obtaining of the third training sample set based on each candidate customer information includes determining the purchased products corresponding to each candidate customer information based on each candidate customer information and historical product purchase records, and regarding the candidate customer information and the products in the product category purchased by the candidate customer information as a customer-product pair to form the third positive sample set; and regarding a customer-product pair formed by randomly selecting a product not purchased by each customer not belonging to the candidate customer information from the customer information set as a negative sample to form the third negative sample set. The third training module is configured to train the second intermediate model based on the third training sample set to obtain a target potential customer mining model. The prediction module is configured to input the third positive sample set into the target potential customer mining model to obtain potential customer information; the target potential customer mining model obtains the probability of each customer purchasing each product, sorts the probabilities of each customer purchasing each product from high to low, determines the prediction result of the target potential customer mining model based on a preset number of customers before sorting, and regards the customer information in the prediction result of the target potential customer mining model as the potential customer information.

5. An electronic device, comprising: The method comprises the following steps: The processor and the memory, wherein the memory stores machine readable instructions executable by the processor, and when the electronic device is running, the processor executes the machine readable instructions to perform the steps of the potential customer mining method according to any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to perform the steps of the potential customer mining method according to any one of claims 1 to 3.

Images