Customer information approval method and device, electronic equipment and medium

Abstract

The invention provides a customer information approval method and device, electronic equipment and a medium, and can be applied to the technical field of big data and the technical field of artificial intelligence. The method comprises the steps of obtaining information of a to-be-approved customer, and calculating a target single customer value of the to-be-approved customer based on the information of the to-be-approved customer; inputting the target single customer value into a customer value model for customer value classification, and outputting a customer value grade; in response to the fact that the customer value level is high, inputting the information of the to-be-approved customer and the target single customer value into a pre-trained passing rate prediction model to perform passing rate prediction of customer information approval, and outputting a passing rate prediction result; and examining and approving the customer to be examined and approved based on the passing rate prediction result, and outputting an examination and approval result.

Description

Technical Field

[0001] The present invention relates to the fields of big data technology and artificial intelligence technology, and more specifically to a customer information approval method, device, electronic device and medium. Background Art

[0002] In the banking industry, high-value users are a type of customer group that banks need to maintain specially. Such customers usually have higher credit ratings, more stable repayment capabilities, and more potential financial needs. Therefore, banks need to pay special attention to and provide services for the business applications of high-value users. At present, the user value model used to evaluate customer value tends to use single-dimensional indicators, such as the customer's consumption amount or transaction frequency, and omits the customer's life cycle value; at the same time, the existing bank approval probability model roughly calculates the probability of repaying a credit card through a set threshold, and further obtains the approval result based on this probability. However, due to the low quality of customer behavior and transaction data, the large time span, and the deviation introduced by model selection, the measurement of user value is prone to errors; at the same time, since customer data is mostly high-dimensional data and the data volume is large, the calculation amount is large in the credit approval process, and mutual exclusion is prone to occur.

[0003] Therefore, in the current banking system, during the approval process, it is easy for the user value of new customers to be unclearly defined, and for customers who should have been approved to be mistakenly rejected by the system, thus affecting the customer experience. Summary of the invention

[0004] In view of the above problems, according to the first aspect of the present invention, a customer information approval method is provided, the method comprising: obtaining information of customers to be approved, and calculating a target single customer value of the customers to be approved based on the information of the customers to be approved; inputting the target single customer value into a customer value model to perform customer value classification, and outputting a customer value grade; in response to the customer value grade being a high value, inputting the information of the customers to be approved and the target single customer value into a pre-trained pass rate prediction model to perform pass rate prediction of customer information approval, and outputting a pass rate prediction result; and approving the customers to be approved based on the pass rate prediction result, and outputting an approval result.

[0005] According to some exemplary embodiments, the pass rate prediction model is trained by the LightGBM model; the information of the customer to be approved and the target single customer value are input into the pre-trained pass rate prediction model to predict the pass rate of customer information approval, and the pass rate prediction result is output, which specifically includes: preprocessing and feature engineering the information of the customer to be approved and the target single customer value to obtain a discretized customer information feature histogram; inputting the customer information feature histogram into N decision trees, each of the N decision trees contains at least one leaf node, and obtaining the prediction results of the N decision trees based on the leaf nodes, wherein N is a positive integer; and based on the prediction results of the N decision trees, using voting or majority voting, selecting the prediction result with the largest number of occurrences as the pass rate prediction result.

[0006] According to some exemplary embodiments, the information of the customer to be approved and the target single customer value are preprocessed and feature engineered, specifically including: data cleaning of the information of the customer to be approved and the target single customer value, feature selection based on expert experience, and obtaining preliminary customer features; using a mutually exclusive bundling algorithm to bundle high-dimensional sparse features in the preliminary customer features to obtain customer bundled features; and using a histogram algorithm to construct a discretized feature histogram of the customer bundled features to obtain a customer information feature histogram.

[0007] According to some exemplary embodiments, the training of the pass rate prediction model by the LightGBM model specifically includes: generating a decision tree by using a Leaf-wise algorithm, wherein the Leaf-wise algorithm is used to recursively process the splitting of the leaf nodes and the depth of the nodes in the N decision trees until a stopping condition is met.

[0008] According to some exemplary embodiments, the training of the pass rate prediction model by the LightGBM model specifically includes: using the new customer personal credit card approval record form and the customer credit card issuance record form to screen new customer applications; based on the customer number in the new customer application, fully associating the information with the customer portrait asset; and based on the content of the full information association, establishing a training set for model training.

[0009] According to some exemplary embodiments, the use of a new customer personal credit card approval record table and a customer credit card issuance record table to screen new customer applications specifically includes: judging whether a same customer association can be established based on the customer number in the new customer personal credit card approval record table corresponding to the customer credit card issuance record table; if the same customer association establishment fails, using the approval record in the new customer personal credit card approval record table corresponding to the customer number as the new customer application; and if the same customer association is successfully established, obtaining the first credit card issuance date corresponding to the customer number from the customer credit card issuance record table, and using the approval record corresponding to the customer number that is no later than the first credit card issuance date as the new customer application.

[0010] According to some exemplary embodiments, calculating the target single customer value of the customer to be approved based on the information of the customer to be approved specifically includes: calculating the single customer full life cycle revenue and the single customer acquisition cost based on the information of the customer to be approved; calculating the difference between the single customer full life cycle revenue and the single customer acquisition cost to obtain a first difference; and using the first difference as the target single customer value.

[0011] According to some exemplary embodiments, the calculation of the single-customer life cycle revenue specifically includes: obtaining the single-customer daily revenue and the single-customer daily maintenance cost, calculating the difference between the single-customer daily revenue and the single-customer daily maintenance cost, and obtaining a second difference; obtaining a customer retention curve, and calculating the single-customer retention time using the area between the customer retention curve and the coordinate axis; and calculating the product of the second difference and the single-customer retention time to obtain the single-customer life cycle revenue.

[0012] According to some exemplary embodiments, calculating the single customer acquisition cost specifically includes: obtaining the total customer acquisition cost and the number of new customers in a target time period; and calculating the quotient of the total customer acquisition cost and the number of new customers to obtain the single customer acquisition cost.

[0013] According to some exemplary embodiments, the customer value model is an RFM model; inputting the target single customer value into the customer value model for customer value classification and outputting a customer value grade specifically includes: calculating the RFM score of the customer to be approved based on the target single customer value; and based on a preset threshold, classifying the RFM score of the customer to be approved and outputting a customer value grade.

[0014] According to a second aspect of the present invention, a customer information approval device is proposed, the device comprising: a target single customer value calculation module, used to: obtain the information of the customer to be approved, and calculate the target single customer value of the customer to be approved based on the information of the customer to be approved; a customer value level acquisition module, used to: input the target single customer value into a customer value model to classify the customer value, and output the customer value level; an inference result generation module, used to: take the speech semantic information as input, use a pre-built upper and lower reasoning model to perform relational reasoning in the speech semantic information, and generate an inference result, the inference result is used to represent the content understanding of the speech semantic information; a pass rate prediction result output module, used to: in response to the customer value level being high value, input the information of the customer to be approved and the target single customer value into a pre-trained pass rate prediction model to predict the pass rate of customer information approval, and output the pass rate prediction result; and an approval result output module, used to: approve the customer to be approved based on the pass rate prediction result, and output the approval result.

[0015] According to some exemplary embodiments, the target single customer value calculation module may include a calculation module, a first difference acquisition unit and a target single customer value acquisition unit.

[0016] According to some exemplary embodiments, the calculation module may be used to calculate the life cycle revenue of a single customer and the cost of acquiring a single customer based on the information of the customer to be approved.

[0017] According to some exemplary embodiments, the first difference acquisition unit may be used to calculate the difference between the single-customer full-lifecycle revenue and the single-customer acquisition cost to obtain a first difference.

[0018] According to some exemplary embodiments, the target single-customer value acquisition unit may be configured to use the first difference as the target single-customer value.

[0019] According to some exemplary embodiments, the calculation module may include a single-customer full-lifecycle revenue calculation module and a single-customer acquisition cost calculation module.

[0020] According to some exemplary embodiments, the single-customer full-lifecycle revenue calculation module may include a second difference calculation unit, a single-customer retention time calculation unit, and a single-customer full-lifecycle revenue calculation unit.

[0021] According to some exemplary embodiments, the second difference calculation unit may be used to obtain a single-customer daily revenue and a single-customer daily maintenance cost, calculate a difference between the single-customer daily revenue and the single-customer daily maintenance cost, and obtain a second difference.

[0022] According to some exemplary embodiments, the single customer retention duration calculation unit may be used to obtain a customer retention curve, and calculate the single customer retention duration using the area between the customer retention curve and a coordinate axis.

[0023] According to some exemplary embodiments, the single-customer full-lifecycle revenue calculation unit may be used to calculate the product of the second difference and the single-customer retention time to obtain the single-customer full-lifecycle revenue.

[0024] According to some exemplary embodiments, the single customer acquisition cost calculation module may include a total customer acquisition cost and new customer quantity acquisition unit and a single customer acquisition cost acquisition unit.

[0025] According to some exemplary embodiments, the total customer acquisition cost and new customer number acquisition unit may be used to acquire the total customer acquisition cost and the number of new customers within a target time period.

[0026] According to some exemplary embodiments, the single customer acquisition cost acquisition unit may be used to calculate a quotient of the total customer acquisition cost and the number of new customers to obtain the single customer acquisition cost.

[0027] According to some exemplary embodiments, the customer value level acquisition module may include an RFM score calculation unit and a customer value level acquisition unit.

[0028] According to some exemplary embodiments, the RFM score calculation unit may be configured to calculate the RFM score of the customer to be approved based on the target single customer value.

[0029] According to some exemplary embodiments, the customer value level acquisition unit may be used to perform customer value classification on the RFM scores of the customers to be approved based on a preset threshold value, and output a customer value level.

[0030] According to some exemplary embodiments, the pass rate prediction result output module may include a preprocessing and feature engineering module, a prediction unit, and a prediction result output unit.

[0031] According to some exemplary embodiments, the preprocessing and feature engineering module may be used to preprocess and feature engineer the information of the pending customers and the target single customer value to obtain a discretized customer information feature histogram.

[0032] According to some exemplary embodiments, the prediction unit can be used to input the customer information feature histogram into N decision trees, each of the N decision trees includes at least one leaf node, and obtain prediction results of the N decision trees based on the leaf nodes, where N is a positive integer.

[0033] According to some exemplary embodiments, the prediction result output unit may be used to select the prediction result with the largest number of occurrences as the pass rate prediction result based on the prediction results of the N decision trees by voting or majority vote.

[0034] According to some exemplary embodiments, the preprocessing and feature engineering module may further include a feature selection unit, a feature bundling unit, and a feature histogram construction unit.

[0035] According to some exemplary embodiments, the feature selection unit may be used to perform data cleaning on the information of the customer to be approved and the target single customer value, perform feature selection based on expert experience, and obtain preliminary customer features.

[0036] According to some exemplary embodiments, the feature bundling unit may be used to utilize a mutually exclusive bundling algorithm to perform bundling processing on high-dimensional sparse features in the customer preliminary features to obtain customer bundled features.

[0037] According to some exemplary embodiments, the feature histogram construction unit may be used to construct a discretized feature histogram from the customer bundling features using a histogram algorithm to obtain a customer information feature histogram.

[0038] According to some exemplary embodiments, the client information approval device may further include a training set building module.

[0039] According to some exemplary embodiments, the training set building module may include a new customer application screening module, an association unit and a training set building unit.

[0040] According to some exemplary embodiments, the new customer application screening module may be used to screen new customer applications using a new customer personal credit card approval record table and a customer credit card issuance record table.

[0041] According to some exemplary embodiments, the association unit may be used to perform full information association with a customer profile asset based on a customer number in the new customer application.

[0042] According to some exemplary embodiments, the training set establishing unit may be used to establish a training set for model training based on the content of the full information association.

[0043] According to some exemplary embodiments, the new customer application screening module may include a judgment unit, a first processing unit, and a second processing unit.

[0044] According to some exemplary embodiments, the judgment unit may be used to judge whether the same customer association can be established based on the customer number in the new customer personal credit card approval record table corresponding to the customer credit card issuance record table.

[0045] According to some exemplary embodiments, the first processing unit may be configured to use the approval record in the new customer personal credit card approval record table corresponding to the customer number as the new customer application if the same customer association establishment fails.

[0046] According to some exemplary embodiments, the second processing unit can be used to obtain the first credit card issuance date corresponding to the customer number from the customer credit card issuance record table if the same customer association is successfully established, and use the approval record corresponding to the customer number that is no later than the first credit card issuance date as the new customer application.

[0047] According to a third aspect of the present invention, there is provided an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the method as described above.

[0048] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the method as described above.

[0049] According to a fifth aspect of the present invention, there is provided a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method as described above is implemented.

[0050] One or more of the above embodiments have the following advantages or beneficial effects: according to the customer information approval method provided by the present invention, by calculating the target single customer value of each customer to be approved, the calculation result of the customer value model can be optimized, thereby obtaining a more accurate customer value classification result, which is helpful for the accurate screening of high-value users, and enables the computer to allocate resources more specifically, thereby improving the calculation speed, thereby improving the prediction efficiency of the pass rate prediction model, and reducing the calculation time and resource consumption; at the same time, by screening high-value users and predicting the pass rate, it is ensured that high-value customers obtain approval results faster, thereby improving customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0052] Figure 1 The application scenario diagram of the customer information approval method, device, equipment, and medium according to an embodiment of the present invention is schematically shown.

[0053] Figure 2 The flowchart of the customer information approval method according to an embodiment of the present invention is schematically shown.

[0054] Figure 3 The flowchart of the method for calculating the target single customer value of customers to be approved according to an embodiment of the present invention is schematically shown.

[0055] Figure 4 The flowchart of the method for calculating the full life cycle revenue of a single customer according to an embodiment of the present invention is schematically shown.

[0056] Figure 5 The flowchart of the method for calculating the single customer acquisition cost according to an embodiment of the present invention is schematically shown.

[0057] Figure 6 The flowchart of the method for classifying customer value according to an embodiment of the present invention is schematically shown.

[0058] Figure 7 A flowchart of a method for outputting a pass rate prediction result through a LightGBM model according to an embodiment of the present invention is schematically shown.

[0059] Figure 8 The flowchart of the method of preprocessing and feature engineering according to an embodiment of the present invention is schematically shown.

[0060] Fig. 9 The flowchart of the method for establishing a training set for model training according to an embodiment of the present invention is schematically shown.

[0061] Fig.10 The flowchart of the method for screening new customer applications according to an embodiment of the present invention is schematically shown.

[0062] Fig.11 The structure block diagram of the client information approval device according to an embodiment of the present invention is schematically shown.

[0063] Fig.12 A block diagram of an electronic device suitable for a customer information approval method according to an embodiment of the present invention is schematically shown. DETAILED DESCRIPTION

[0064] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present invention.

[0065] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0066] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0067] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0068] In the technical solution of the present invention, the acquisition, storage and application of user personal information involved all comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

[0069] First, the technical terms described in this article are explained and described as follows.

[0070] LightGBM (Light Gradient Boosting Machine) is a machine learning framework based on gradient boosting trees. It is a powerful ensemble learning algorithm for processing structured data, such as tabular data. LightGBM builds multiple decision trees in an efficient and accurate way and integrates them into a more powerful prediction model. It has efficient performance and the ability to process large-scale data.

[0071] Decision tree: A machine learning model based on a tree structure, used for classification and regression tasks. In classification tasks, decision trees gradually split data into different categories based on input features. In regression tasks, decision trees predict continuous values. Each node of the decision tree represents a feature, each branch represents the value of a feature, and the leaf node represents the final prediction result.

[0072] Mutually exclusive bundling algorithm: High-dimensional data is often sparse. In the sparse feature space, many features are mutually exclusive (i.e., they will not take non-zero values ​​at the same time), so mutually exclusive features can be safely bundled. In addition, a bias can be added to ensure the feasibility of feature bundling: for example, to bind two features A and B, the original value of feature A is in the interval [0, 10), and the original value of feature B is in the interval [0, 20). A bias constant 10 can be added to the value of feature B to change its value range to [10, 30). The feature value range after A and B features are bound is [0, 30), thereby realizing the fusion of features A and B.

[0073] Histogram algorithm: Divide the continuous values ​​of the feature into discrete bins, and use bins to construct the feature histogram during the training process. The histogram of a leaf can be obtained by subtracting the histogram of its parent node from the histogram of its sibling node, thereby traversing the k buckets of the histogram and approximately finding the optimal branching point, thereby greatly reducing the amount of calculation, which is particularly effective when processing high-dimensional sparse data.

[0074] Leaf-wise algorithm: A splitting strategy for building decision trees, usually used in the gradient boosting tree algorithm. With the same number of splits, it can reduce more errors and obtain better accuracy.

[0075] Gradient-based One-Side Sampling (GOSS): In the gradient boosting decision tree, the data samples do not have initial weights at the beginning. When calculating information gain, the importance of samples with different gradients is different. According to the definition of information gain, the larger the gradient, the greater the corresponding contribution when calculating information gain. Therefore, when calculating information gain during training, the lightGBM model will retain samples with large gradients and randomly collect some samples with small gradients (to ensure the original data distribution). This is more accurate than random sampling with a uniform sampling ratio.

[0076] As part of the financial industry, the banking industry has a large customer base, including high-value users. These customers contribute greatly to the bank's business and profits, and usually have higher credit and greater financial needs. Therefore, banks need to provide special maintenance and services for these customers. However, the information data of high-value users often has the following characteristics: high latitude, large scale and complexity, which means that banks need to process a large amount of complex data, resulting in a significant increase in computing costs; at the same time, due to the high dimensionality of the data, mutually exclusive features are prone to occur, affecting modeling and decision-making.

[0077] Currently, the models used by banks in the customer approval process may lead to false rejections in some cases, which has a negative impact on customer experience.

[0078] In addition, the current definition of user value tends to use single-dimensional indicators, such as customer transaction amount or transaction frequency, which ignores other factors affecting customers, such as lifetime income and acquisition cost, leading to deviations in the evaluation of user value.

[0079] Based on this, an embodiment of the present invention provides a customer information approval method, characterized in that the method includes: obtaining information of customers to be approved, and calculating the target single customer value of the customers to be approved based on the information of the customers to be approved; inputting the target single customer value into the customer value model to classify the customer value, and outputting the customer value level; in response to the customer value level being high value, inputting the information of the customers to be approved and the target single customer value into a pre-trained pass rate prediction model to predict the pass rate of customer information approval, and outputting the pass rate prediction result; and approving the customers to be approved based on the pass rate prediction result, and outputting the approval result. According to the customer information approval method provided by the present invention, through customer value classification, the computer can more accurately distinguish users of different value levels, which means that the computer can allocate resources more specifically, thereby improving the calculation speed, thereby improving the prediction efficiency of the pass rate prediction model, and reducing the calculation time and resource consumption; at the same time, through the screening of high-value users and the pass rate prediction, it is ensured that high-value customers obtain approval results faster, thereby improving customer satisfaction.

[0080] It should be noted that the customer information approval method, device, equipment and medium determined by the present invention can be used in the fields of big data technology and artificial intelligence technology, can also be used in the financial field, and can also be used in various fields other than the fields of big data technology and artificial intelligence technology and the financial field. The application fields of the customer information approval method, device, equipment and medium provided by the embodiments of the present invention are not limited.

[0081] In the technical solution of the present invention, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0082] Figure 1 The application scenario diagram of the customer information approval method, device, equipment, and medium according to an embodiment of the present invention is schematically shown.

[0083] like Figure 1 As shown, the application scenario 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0084] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0085] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0086] The server 105 may be a server that provides various services, such as a background management server (only an example) that provides support for websites browsed by users using the terminal devices 101, 102, and 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0087] It should be noted that the client information approval method provided in the embodiment of the present invention can generally be executed by the server 105. Accordingly, the client information approval device provided in the embodiment of the present invention can generally be set in the server 105. The client information approval method provided in the embodiment of the present invention can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the client information approval device provided in the embodiment of the present invention can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105.

[0088] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0089] Figure 2 The flowchart of the customer information approval method according to an embodiment of the present invention is schematically shown.

[0090] like Figure 2 As shown, the customer information approval method 200 of this embodiment may include operations S210 to S240.

[0091] In operation S210, information of customers to be approved is obtained, and based on the information of the customers to be approved, a target single customer value of the customers to be approved is calculated.

[0092] In the embodiment of the present invention, it is necessary to collect relevant information of the customer to be approved, which may include the customer's personal information (such as name, age, gender), financial status (such as income, expenditure, debt), credit history, transaction history, expenditure, income, expenses, etc. This information is usually stored in the database or system of the bank.

[0093] In the embodiment of the present invention, in order to further help the bank understand the long-term value of each customer and thus optimize the evaluation result of customer value, an operation of calculating the target single customer value of the customer to be approved is also provided.

[0094] Figure 3 The flowchart of the method for calculating the target single customer value of customers to be approved according to an embodiment of the present invention is schematically shown.

[0095] like Figure 3 As shown, the method for calculating the target single customer value of customers to be approved in this embodiment may include operations S310 to S330, and operations S310 to S330 may at least partially execute operation S210.

[0096] In operation S310, based on the information of the customer to be approved, the life cycle revenue of a single customer and the cost of acquiring a single customer are calculated.

[0097] Figure 4 The flowchart of the method for calculating the full life cycle revenue of a single customer according to an embodiment of the present invention is schematically shown.

[0098] like Figure 4 As shown, the method for calculating the lifetime revenue of a single customer in this embodiment may include operations S410 to S430, and operations S410 to S430 may at least partially execute operation S310.

[0099] In operation S410, a single-customer daily revenue and a single-customer daily maintenance cost are obtained, and a difference between the single-customer daily revenue and the single-customer daily maintenance cost is calculated to obtain a second difference.

[0100] In the embodiment of the present invention, the daily income of a single customer and the daily maintenance cost of a single customer refer to the daily income obtained by the bank from each customer and the cost incurred for maintaining the customer. The records of the daily income of a single customer and the daily maintenance cost of a single customer can be obtained from the data source of the bank system. Among them, calculating the difference between the daily income and the daily maintenance cost, that is, the daily net income, can help the bank understand the profitability of each customer.

[0101] In operation S420, a customer retention curve is obtained, and the single customer retention time is calculated using the area between the customer retention curve and the coordinate axis.

[0102] In an embodiment of the present invention, the customer retention curve shows the customer retention status over a period of time. For example, the customer retention curve can be a curve about time, showing the customer retention status from the time of acquisition.

[0103] In an embodiment of the present invention, the retention status of customers over a period of time can be obtained, which can be at daily, weekly or monthly time intervals. The retention data can record the customer's activity at different time points, the situation of continuing to use products or services, etc.; using the obtained retention data, a customer retention curve can be drawn, the horizontal axis represents time (usually in days or months), the vertical axis represents the retention rate (or the number of retained people), and the shape of the curve reflects the customer's retention status at different time points; calculating the retention time of a single customer involves measuring the area between the customer retention curve and the coordinate axis, which represents the sum of the customer's retention time since acquisition, and then, by dividing by the total retention rate (or the total number of customers), the average retention time can be obtained.

[0104] In an embodiment of the present invention, the single customer retention time can help banks understand customer retention status and retention time, thereby helping banks determine which customers are more likely to maintain a long-term cooperative relationship with the enterprise, so as to formulate better customer relationship management strategies, including customer maintenance, customized services and promotional activities.

[0105] In operation S430, the product of the second difference and the single-customer retention time is calculated to obtain the single-customer full-life cycle revenue.

[0106] In an embodiment of the present invention, the single customer life cycle revenue helps to evaluate the value contribution of customer retention to the bank. If the second difference is positive, it means that the customer's daily net income increases and the retention time is longer, which has a positive impact on the bank's long-term profit potential.

[0107] Figure 5 The flowchart of the method for calculating the single customer acquisition cost according to an embodiment of the present invention is schematically shown.

[0108] like Figure 5As shown, the method for calculating the single customer acquisition cost of this embodiment may include operations S510 to S520, and operations S510 to S520 may at least partially execute operation S310.

[0109] In operation S510, the total customer acquisition cost and the number of new customers in the target time period are obtained.

[0110] In an embodiment of the present invention, customer acquisition cost refers to the cost invested by a bank in acquiring new customers within a certain period of time, including advertising expenses, marketing expenses, sales expenses, etc.; the number of new customers can be obtained from channels such as new user registrations and the number of new customers purchasing products or services.

[0111] In operation S520, the quotient of the total customer acquisition cost and the number of new customers is calculated to obtain the single customer acquisition cost.

[0112] In an embodiment of the present invention, the cost per customer acquisition helps banks make decisions, such as determining customer acquisition budgets, selecting customer acquisition channels, and improving customer acquisition strategies. By monitoring changes in the cost per customer acquisition, banks can evaluate the effectiveness of customer acquisition activities and optimize them based on cost-effectiveness.

[0113] Return to reference Figure 3 In operation S320, the difference between the single-customer lifetime revenue and the single-customer acquisition cost is calculated to obtain a first difference.

[0114] In operation S330, the first difference is used as the target single customer value.

[0115] In the embodiment of the present invention, customer value classification based on target single customer value allows the bank to classify and analyze customers to be approved more finely.

[0116] Return to reference Figure 2 In operation S220, the target single customer value is input into the customer value model to classify the customer value and output the customer value level.

[0117] In an embodiment of the present invention, the customer value model is an RFM model.

[0118] Figure 6 The flowchart of the method for classifying customer value according to an embodiment of the present invention is schematically shown.

[0119] like Figure 6 As shown, the method for classifying customer value in this embodiment may include operations S610 to S620, and operations S610 to S620 may at least partially execute operation S220.

[0120] In operation S610, the RFM score of the customer to be approved is calculated based on the target single customer value.

[0121] In an embodiment of the present invention, the RFM score generally includes the following aspects: a Recency score, which is measured based on the time of the most recent interaction or purchase behavior of the customer to be approved. If the customer's most recent interaction with the bank is more recent, the Recency score is higher, indicating that the customer is more active; a Frequency score, which indicates the number of interactions or purchases made by the customer within a certain period of time. Customers who interact frequently receive a higher Frequency score, indicating that they are highly engaged; a Monetary score, which represents the total transaction amount or income of the customer within a certain period of time. Customers with high transaction amounts receive a higher Monetary score, indicating that they contribute more to the bank.

[0122] In an embodiment of the present invention, the target single customer value can be used to adjust the weight of the RFM score. For example, if the single customer value is positive, the weights of Recency, Frequency, and Monetary can be increased to emphasize high-value customers. If the single customer value is negative, the weights can be adjusted accordingly to focus on reducing customer acquisition costs and increasing customer value. Furthermore, based on the specific value of the single customer value, weight and score rules can also be defined to ensure that different single customer value levels have different effects on the RFM score.

[0123] In operation S620, based on a preset threshold, the RFM scores of the customers to be approved are classified into customer value categories, and a customer value grade is output.

[0124] In an embodiment of the present invention, the customers to be approved may be classified into high-value, medium-value, and low-value customers according to a preset threshold and a calculated RFM score.

[0125] Return to reference Figure 2 In operation S230, in response to the customer value level being high, the information of the customer to be approved and the target single customer value are input into a pre-trained pass rate prediction model to predict the pass rate of customer information approval, and the pass rate prediction result is output.

[0126] In an embodiment of the present invention, the pass rate prediction model is trained by the LightGBM model.

[0127] In an embodiment of the present invention, a Leaf-wise algorithm is used to generate a decision tree, wherein the Leaf-wise algorithm is used to recursively process the splitting of the leaf nodes and the depth of the nodes in the decision tree until a stopping condition is met.

[0128] Specifically, the Leaf-wise algorithm starts at the root node of the tree. First, the root node is split into two child nodes to minimize impurity. Next, the Leaf-wise algorithm does not generate the entire decision tree hierarchy at once like other algorithms. Instead, it selects a leaf node and tries to split it, where the selection of the leaf node is based on a pre-established strategy, usually determined by the depth of the current tree, the impurity before the split, etc. The selected leaf node is split to obtain two new leaf nodes. The splitting process is similar to the root node splitting, selecting the best features and splitting points to minimize impurity, and this process continues until the stopping condition is met.

[0129] In an embodiment of the present invention, the stopping condition may be the maximum depth of the tree, the number of samples in a leaf node is less than a preset threshold, the splitting no longer significantly reduces the impurity, etc.

[0130] In an embodiment of the present invention, gradient-based single-side sampling can also be used for sampling and weighted training to improve the effect of model training.

[0131] In an embodiment of the present invention, since the number of samples of different categories in the information of customers to be approved varies greatly, the model tends to predict categories with larger numbers during training. Therefore, gradient-based unilateral sampling and weighted training can be used to solve this problem so that model training can better identify these categories.

[0132] In an embodiment of the present invention, the number of samples of the minority class can be increased by unilateral sampling. Specifically, it can be achieved by undersampling (reducing the majority class samples) or oversampling (increasing the minority class samples), which helps to balance the weights of different classes and reduce the overfitting of the model to the majority class.

[0133] In the embodiments of the present invention, in addition to sampling, different weights can be given to samples of different categories during the training process. For the majority category (usually the negative category), a lower weight is assigned; for the minority category (usually the positive category), a higher weight is assigned. These weights will play a role in the loss function of the model, affect the update of the model parameters, and make the model pay more attention to important or rare categories, which helps to balance the importance between categories and improve the generalization ability of the model.

[0134] In the embodiments of the present invention, sampling and weighted training can reduce the risk of overfitting. By introducing more minority category samples or increasing their weights, the model can more easily capture the characteristics of these categories, thereby reducing the risk of overfitting.

[0135] Figure 7A flowchart of a method for outputting a pass rate prediction result through a LightGBM model according to an embodiment of the present invention is schematically shown.

[0136] like Figure 7 As shown, the method of outputting the pass rate prediction result through the LightGBM model in this embodiment may include operations S710 to S730, and operations S710 to S730 may at least partially execute operation S230.

[0137] In operation S710, preprocessing and feature engineering are performed on the information of the customer to be approved and the target single customer value to obtain a discretized customer information feature histogram.

[0138] Figure 8 The flowchart of the method of preprocessing and feature engineering according to an embodiment of the present invention is schematically shown.

[0139] like Figure 8 As shown, the preprocessing and feature engineering method of this embodiment may include operations S810 to S830, and operations S810 to S830 may at least partially execute operation S710.

[0140] In operation S810, data cleaning is performed on the information of the customer to be approved and the target single customer value, and feature selection is performed based on expert experience to obtain preliminary customer features.

[0141] In an embodiment of the present invention, data cleaning may include removing duplicate values, processing missing values, processing outliers, etc. to ensure the quality and accuracy of the data.

[0142] In an embodiment of the present invention, based on expert experience, the most relevant and valuable features can be selected from the original data for subsequent modeling and analysis. Expert experience can help determine which features are important for evaluating the target single customer value of a customer.

[0143] In operation S820, a mutually exclusive bundling algorithm is used to bundle the high-dimensional sparse features in the preliminary customer features to obtain customer bundled features.

[0144] In an embodiment of the present invention, in order to solve the problem that the sample quantity of the customer information to be approved is huge, redundant, and prone to mutual exclusion, a mutually exclusive bundling algorithm can be further used to reduce the dimension and improve the interpretability of the data. Among them, the mutually exclusive feature bundling algorithm is used to process high-dimensional sparse features in the customer's preliminary features, that is, features in the customer's preliminary features with a large number of values, but a limited number of samples for each value. Such features may cause dimensionality disasters during the modeling process, and the number of samples with different values ​​is unbalanced.

[0145] In operation S830, a histogram algorithm is used to construct a discretized feature histogram of the customer bundling features to obtain a customer information feature histogram.

[0146] In an embodiment of the present invention, the bundled features can also be constructed into a discretized feature histogram, that is, the continuous numerical features are converted into a set of discrete intervals or boxes for modeling and analysis. The distribution of customer information can be obtained through the customer information feature histogram, and each box represents the number of samples or other statistical information within a specific range.

[0147] Return to reference Figure 7 In operation S720, the customer information feature histogram is input into N decision trees, each of the N decision trees includes at least one leaf node, and prediction results of the N decision trees are obtained based on the leaf nodes, where N is a positive integer.

[0148] In operation S730, based on the prediction results of the N decision trees, a prediction result with the largest number of occurrences is selected as a pass rate prediction result by voting or majority vote.

[0149] In addition, in order to avoid noise caused by irrelevant information, sample screening of the training set can be performed during the training process and the sample screening strategy can be refined.

[0150] Fig. 9 The flowchart of the method for establishing a training set for model training according to an embodiment of the present invention is schematically shown.

[0151] like Fig. 9 As shown, the method for establishing a training set for model training in this embodiment may include operations S910 to S930.

[0152] In operation S910, the new customer's application is screened using the new customer's personal credit card approval record table and the customer credit card issuance record table.

[0153] In an embodiment of the present invention, personal credit card approval records and customer credit card issuance records of new customers who have submitted applications and completed approval within a certain period of time and do not contain the aforementioned warning signs can be selected as sample base tables.

[0154] Fig.10 The flowchart of the method for screening new customer applications according to an embodiment of the present invention is schematically shown.

[0155] like Fig.10 As shown, the method for screening new customer applications in this embodiment may include operations S1010 to S1030.

[0156] In operation S1010, based on the customer number in the new customer's personal credit card approval record table, corresponding to the customer credit card issuance record table, it is determined whether the same customer association can be established.

[0157] In the embodiment of the present invention, the credit card approval record and the card issuance record of the same customer can be identified by the customer number.

[0158] In operation S1020, if the same customer association establishment fails, the approval record in the new customer personal credit card approval record table corresponding to the customer number is used as the new customer application.

[0159] In an embodiment of the present invention, if it is determined in operation S1010 that the same customer association cannot be established, it means that the customer has not obtained the first credit card, so all applications of the customer can be determined as new customer applications.

[0160] In operation S1030, if the same customer association is successfully established, the first credit card issuance date corresponding to the customer number is obtained from the customer credit card issuance record table, and the approval record corresponding to the customer number that is no later than the first credit card issuance date is used as the new customer application.

[0161] In an embodiment of the present invention, if it is determined in operation S1010 that the same customer association can be established, it means that the customer has obtained a credit card, so all applications before the customer obtains the first credit card can be determined as new customer applications.

[0162] Return to reference Fig. 9 In operation S920, based on the customer number in the new customer application, full information association is performed with the customer portrait asset.

[0163] In the embodiment of the present invention, the customer portrait generally includes the customer's basic information, historical transaction records, financial status, etc. Through full information association, the complete information of the new customer can be obtained for subsequent model training and analysis.

[0164] In an embodiment of the present invention, the customer number applied for by a new customer can be used to perform full information association with the customer portrait asset, and the data of the corresponding customer portrait asset within the selected time period can be obtained, which together with the approval record constitutes a complete sample set.

[0165] In operation S930, a training set for model training is established based on the content of the full information association.

[0166] Return to reference Figure 2 In operation S240, the customer to be approved is approved based on the pass rate prediction result, and the approval result is output.

[0167] In an embodiment of the present invention, before the approval is conducted, an approval threshold can be selected to determine which customers' applications will be approved. The selection of the threshold is usually based on business needs and risk management strategies. If the predicted pass rate result of a customer to be approved is greater than or equal to the set threshold, the customer's credit card application will be approved, otherwise it will be rejected.

[0168] In the embodiment of the present invention, the approval result can be "approved" or "rejected". If the customer's application is approved, the bank can continue to process the application and issue a credit card; if the customer's application is rejected, the bank needs to send a rejection letter to the customer and explain the reason for the rejection.

[0169] According to the customer information approval method provided by the present invention, through customer value classification, the computer can more accurately distinguish users of different value levels and allocate resources more targetedly, thereby improving the calculation speed, thereby improving the prediction efficiency of the pass rate prediction model, and reducing the calculation time and resource consumption; at the same time, through the screening of high-value users and the prediction of the pass rate, it is ensured that high-value customers obtain the approval results faster, thereby improving customer satisfaction. Specifically, it brings the following beneficial effects:

[0170] 1. By calculating the target single customer value of each customer to be approved, the calculation results of the RFM model can be optimized, thereby obtaining more accurate customer value classification results, which is helpful for accurate screening of high-value users and improving user satisfaction and loyalty;

[0171] 2. Using the RFM model to classify customers can better understand their consumption behavior and value, which helps banks to specify personalized strategies for different customer groups, thereby improving the effectiveness of marketing and product sales;

[0172] 3. The Leaf-wise algorithm can generate decision trees more efficiently, thereby reducing the number of node splits during the training process and effectively improving the training speed on large-scale data sets. Therefore, banks can update models more quickly and respond to customer information approval needs, improving efficiency;

[0173] 4. Leaf-wise algorithms also help reduce the memory footprint of the model, thereby reducing memory pressure and enabling banks to run the model on smaller hardware configurations;

[0174] 5. In the credit card approval scenario of a bank, customer information may contain a large number of sparse features. The mutually exclusive bundling algorithm helps to process high-dimensional and sparse features. It can bundle related features together, thereby reducing the complexity of the model and improving the training and prediction efficiency.

[0175] 6. The histogram algorithm is used to construct a discretized feature histogram after the exclusive bundling algorithm. This method allows parallel computing and effectively utilizes the advantages of multi-core processors, thereby further improving the training speed of the model;

[0176] 7. Gradient-based one-sided sampling and weighted training are conducive to processing unbalanced data sets such as customer information. They can ensure that the model can better learn information from minority categories, improve the training effect and generalization performance of the model, and in the bank's credit card approval scenario, they can help accurately identify high-risk scenarios and reduce credit risks.

[0177] It should be noted that the customer information approval method provided by the present invention is applied to the credit approval of bank customers, which is only a preferred embodiment and is not intended to limit the specific application of the present invention. That is, the customer information approval method provided in the embodiment of the present invention can also be applied to scenarios such as approval of other physical enterprises.

[0178] Based on the above customer information approval method, the present invention also provides a customer information approval device. Fig.11 The device is described in detail.

[0179] Fig.11 The structure block diagram of the client information approval device according to an embodiment of the present invention is schematically shown.

[0180] like Fig.11 As shown, the customer information approval device 1100 according to this embodiment includes a target single customer value calculation module 1110 , a customer value level acquisition module 1120 , a pass rate prediction result output module 1130 and an approval result output module 1140 .

[0181] The target single customer value calculation module 1110 can be used to obtain information of pending customers, and calculate the target single customer value of the pending customers based on the pending customers' information. In one embodiment, the target single customer value calculation module 1110 can be used to perform the operation S210 described above, which will not be described in detail here.

[0182] The customer value level acquisition module 1120 may be used to input the target single customer value into the customer value model to perform customer value classification and output the customer value level. In one embodiment, the customer value level acquisition module 1120 may be used to perform the operation S220 described above, which will not be described in detail here.

[0183] The pass rate prediction result output module 1130 can be used to respond to the customer value level being high value, input the information of the customer to be approved and the target single customer value into a pre-trained pass rate prediction model to predict the pass rate of customer information approval, and output a pass rate prediction result. In one embodiment, the pass rate prediction result output module 1130 can be used to perform the operation S230 described above, which will not be repeated here.

[0184] The approval result output module 1140 can be used to approve the client to be approved based on the pass rate prediction result and output the approval result. In one embodiment, the approval result output module 1140 can be used to perform the operation S240 described above, which will not be described in detail here.

[0185] According to an embodiment of the present invention, the target single customer value calculation module 1110 may include a calculation module, a first difference acquisition unit and a target single customer value acquisition unit.

[0186] The calculation module may be used to calculate the life cycle revenue of a single customer and the cost of acquiring a single customer based on the information of the customer to be approved. In one embodiment, the calculation module may be used to perform the operation S310 described above, which will not be described in detail here.

[0187] The first difference acquisition unit may be used to calculate the difference between the single customer lifecycle revenue and the single customer acquisition cost to obtain a first difference. In one embodiment, the first difference acquisition unit may be used to perform the operation S320 described above, which will not be described in detail herein.

[0188] The target single customer value acquisition unit may be used to use the first difference as the target single customer value. In one embodiment, the target single customer value acquisition unit may be used to perform the operation S330 described above, which will not be described in detail here.

[0189] According to an embodiment of the present invention, the calculation module may include a single-customer full-lifecycle revenue calculation module and a single-customer acquisition cost calculation module.

[0190] According to an embodiment of the present invention, the single-customer full-lifecycle revenue calculation module may include a second difference calculation unit, a single-customer retention time calculation unit and a single-customer full-lifecycle revenue calculation unit.

[0191] The second difference calculation unit can be used to obtain the single-customer daily revenue and the single-customer daily maintenance cost, calculate the difference between the single-customer daily revenue and the single-customer daily maintenance cost, and obtain a second difference. In one embodiment, the second difference calculation unit can be used to perform the operation S410 described above, which will not be repeated here.

[0192] The single customer retention time calculation unit can be used to obtain a customer retention curve, and calculate the single customer retention time using the area between the customer retention curve and the coordinate axis. In one embodiment, the single customer retention time calculation unit can be used to perform the operation S420 described above, which will not be repeated here.

[0193] The single customer life cycle revenue calculation unit can be used to calculate the product of the second difference and the single customer retention time to obtain the single customer life cycle revenue. In one embodiment, the single customer life cycle revenue calculation unit can be used to perform the operation S430 described above, which is not repeated here.

[0194] According to an embodiment of the present invention, the single customer acquisition cost calculation module may include a total customer acquisition cost and new customer quantity acquisition unit and a single customer acquisition cost acquisition unit.

[0195] The total customer acquisition cost and new customer number acquisition unit may be used to acquire the total customer acquisition cost and the number of new customers in the target time period. In one embodiment, the total customer acquisition cost and new customer number acquisition unit may be used to perform the operation S510 described above, which will not be described in detail here.

[0196] The single customer acquisition cost acquisition unit may be used to calculate the quotient of the total customer acquisition cost and the number of new customers to obtain the single customer acquisition cost. In one embodiment, the single customer acquisition cost acquisition unit may be used to perform the operation S520 described above, which will not be described in detail here.

[0197] According to an embodiment of the present invention, the customer value level acquisition module 1120 may include an RFM score calculation unit and a customer value level acquisition unit.

[0198] The RFM score calculation unit may be used to calculate the RFM score of the customer to be approved based on the target single customer value. In one embodiment, the RFM score calculation unit may be used to perform the operation S610 described above, which will not be described in detail here.

[0199] The customer value level acquisition unit may be used to classify the RFM scores of the pending customers based on a preset threshold and output a customer value level. In one embodiment, the customer value level acquisition unit may be used to perform the operation S620 described above, which will not be described in detail here.

[0200] According to an embodiment of the present invention, the pass rate prediction result output module 1130 may include a preprocessing and feature engineering module, a prediction unit and a prediction result output unit.

[0201] The preprocessing and feature engineering module can be used to preprocess and feature engineer the information of the pending customer and the target single customer value to obtain a discretized customer information feature histogram. In one embodiment, the preprocessing and feature engineering module can be used to perform the operation S710 described above, which will not be repeated here.

[0202] The prediction unit may be used to input the customer information feature histogram into N decision trees, each of the N decision trees including at least one leaf node, and obtain prediction results of the N decision trees based on the leaf nodes, where N is a positive integer. In one embodiment, the prediction unit may be used to perform the operation S720 described above, which will not be described in detail herein.

[0203] The prediction result output unit can be used to select the prediction result with the most occurrences as the pass rate prediction result based on the prediction results of the N decision trees by voting or majority vote. In one embodiment, the prediction result output unit can be used to perform the operation S730 described above, which will not be repeated here.

[0204] According to an embodiment of the present invention, the preprocessing and feature engineering module may further include a feature selection unit, a feature bundling unit and a feature histogram construction unit.

[0205] The feature selection unit can be used to clean the information of the pending customer and the target single customer value, perform feature selection based on expert experience, and obtain preliminary customer features. In one embodiment, the feature selection unit can be used to perform the operation S810 described above, which will not be repeated here.

[0206] The feature bundling unit may be used to use a mutually exclusive bundling algorithm to perform bundling processing on the high-dimensional sparse features in the customer preliminary features to obtain customer bundled features. In one embodiment, the feature bundling unit may be used to perform the operation S820 described above, which will not be described in detail herein.

[0207] The feature histogram construction unit may be used to construct a discretized feature histogram from the customer bundling feature using a histogram algorithm to obtain a customer information feature histogram. In one embodiment, the feature histogram construction unit may be used to perform the operation S830 described above, which will not be described in detail herein.

[0208] According to an embodiment of the present invention, the client information approval device 1100 may further include a training set building module.

[0209] According to an embodiment of the present invention, the training set establishment module may include a new customer application screening module, an association unit and a training set establishment unit.

[0210] The new customer application screening module can be used to screen new customer applications using the new customer personal credit card approval record table and the customer credit card issuance record table. In one embodiment, the new customer application screening module can be used to perform the operation S910 described above, which will not be repeated here.

[0211] The association unit may be used to associate the customer profile asset with the customer number in the new customer application. In one embodiment, the association unit may be used to perform the operation S920 described above, which will not be described in detail here.

[0212] The training set establishment unit may be used to establish a training set for model training based on the content of the full information association. In one embodiment, the training set establishment unit may be used to perform the operation S930 described above, which will not be described in detail here.

[0213] According to an embodiment of the present invention, the new customer application screening module may include a judgment unit, a first processing unit and a second processing unit.

[0214] The judgment unit can be used to judge whether the same customer association can be established based on the customer number in the new customer personal credit card approval record table corresponding to the customer credit card issuance record table. In one embodiment, the judgment unit can be used to perform the operation S1010 described above, which will not be repeated here.

[0215] The first processing unit may be used to, if the same customer association establishment fails, use the approval record in the new customer personal credit card approval record table corresponding to the customer number as the new customer application. In one embodiment, the first processing unit may be used to perform the operation S1020 described above, which will not be described in detail here.

[0216] The second processing unit can be used to obtain the first credit card issuance date corresponding to the customer number from the customer credit card issuance record table if the same customer association is successfully established, and use the approval record corresponding to the customer number that is no later than the first credit card issuance date as the new customer application. In one embodiment, the second processing unit can be used to perform the operation S1030 described above, which will not be repeated here.

[0217] According to an embodiment of the present invention, any multiple modules among the target single customer value calculation module 1110, the customer value level acquisition module 1120, the pass rate prediction result output module 1130 and the approval result output module 1140 can be combined in one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the target single customer value calculation module 1110, the customer value level acquisition module 1120, the pass rate prediction result output module 1130 and the approval result output module 1140 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation modes of software, hardware and firmware or in a proper combination of any of them. Alternatively, at least one of the target single customer value calculation module 1110, the customer value level acquisition module 1120, the pass rate prediction result output module 1130 and the approval result output module 1140 can be at least partially implemented as a computer program module, and when the computer program module is executed, the corresponding function can be executed.

[0218] Fig.12 A block diagram of an electronic device suitable for a customer information approval method according to an embodiment of the present invention is schematically shown.

[0219] like Fig.12 As shown, the electronic device 1200 according to an embodiment of the present invention includes a processor 1201, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1202 or a program loaded from a storage part 1208 into a random access memory (RAM) 1203. The processor 1201 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1201 may also include an onboard memory for caching purposes. The processor 1201 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0220] In RAM 1203, various programs and data required for the operation of electronic device 1200 are stored. Processor 1201, ROM 1202 and RAM 1203 are connected to each other via bus 1204. Processor 1201 performs various operations of the method flow according to the embodiment of the present invention by executing the program in ROM 1202 and / or RAM 1203. It should be noted that the program can also be stored in one or more memories other than ROM 1202 and RAM 1203. Processor 1201 can also perform various operations of the method flow according to the embodiment of the present invention by executing the program stored in the one or more memories.

[0221] According to an embodiment of the present invention, the electronic device 1200 may further include an input / output (I / O) interface 1205, which is also connected to the bus 1204. The electronic device 1200 may further include one or more of the following components connected to the I / O interface 1205: an input portion 1206 including a keyboard, a mouse, etc.; an output portion 1207 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 1208 including a hard disk, etc.; and a communication portion 1209 including a network interface card such as a LAN card, a modem, etc. The communication portion 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to the I / O interface 1205 as needed. A removable medium 1211, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1210 as needed, so that a computer program read therefrom is installed into the storage portion 1208 as needed.

[0222] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiment; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0223] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present invention, the computer-readable storage medium may include the ROM 1202 and / or RAM 1203 described above and / or one or more memories other than ROM 1202 and RAM 1203.

[0224] The embodiment of the present invention also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the method provided by the embodiment of the present invention.

[0225] The computer program executes the above functions defined in the system / device of the embodiment of the present invention when it is executed by the processor 1201. According to the embodiment of the present invention, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0226] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 1209, and / or installed from the removable medium 1211. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0227] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1209, and / or installed from the removable medium 1211. When the computer program is executed by the processor 1201, the above functions defined in the system of the embodiment of the present invention are performed. According to the embodiment of the present invention, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0228] According to an embodiment of the present invention, the program code for executing the computer program provided by the embodiment of the present invention can be written in any combination of one or more programming languages, specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, such as Java, C++, python, "C" language or similar programming languages. The program code can be executed completely on the user computing device, partially on the user device, partially on the remote computing device, or completely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0229] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0230] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination advantageously. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A customer information approval method, characterized in that: The method comprises: Acquire information of customers to be approved, and calculate the target single customer value of the customers to be approved based on the information of the customers to be approved; Inputting the target single customer value into the customer value model to classify the customer value and output the customer value level; In response to the customer value level being high, inputting the information of the customer to be approved and the target single customer value into a pre-trained pass rate prediction model to predict the pass rate of customer information approval, and outputting a pass rate prediction result; and The customers to be approved are approved based on the pass rate prediction result, and the approval result is output.

2. The method according to claim 1, characterized in that The pass rate prediction model is trained by the LightGBM model; The step of inputting the information of the customer to be approved and the target single customer value into a pre-trained pass rate prediction model to predict the pass rate of customer information approval, and outputting the pass rate prediction result specifically includes: Preprocessing and feature engineering the information of the customer to be approved and the target single customer value to obtain a discretized customer information feature histogram; Inputting the customer information feature histogram into N decision trees, each of the N decision trees comprising at least one leaf node, and obtaining prediction results of the N decision trees based on the leaf nodes, wherein N is a positive integer; and Based on the prediction results of the N decision trees, the prediction result with the largest number of occurrences is selected as the pass rate prediction result by voting or majority vote.

3. The method according to claim 2, characterized in that The preprocessing and feature engineering of the information of the customer to be approved and the target single customer value to obtain a discretized customer information feature histogram specifically includes: Perform data cleaning on the information of the pending customers and the target single customer value, perform feature selection based on expert experience, and obtain preliminary customer features; Using a mutually exclusive bundling algorithm, bundling high-dimensional sparse features in the customer preliminary features to obtain customer bundled features; and The customer bundling features are constructed into a discretized feature histogram using a histogram algorithm to obtain a customer information feature histogram.

4. The method according to claim 2 or 3, characterized in that: The method of training the pass rate prediction model by the LightGBM model specifically includes: A decision tree is generated using a Leaf-wise algorithm, wherein the Leaf-wise algorithm is used to recursively process the splitting of the leaf nodes and the depth of the nodes in the N decision trees until a stopping condition is met.

5. The method according to claim 4, characterized in that The method of training the pass rate prediction model by the LightGBM model specifically includes: Use the new customer personal credit card approval record form and customer credit card issuance record form to screen new customer applications; Based on the customer number in the new customer application, fully associate the customer profile asset; and Based on the content of the full information association, a training set for model training is established.

6. The method according to claim 5, characterized in that The use of the new customer personal credit card approval record form and the customer credit card issuance record form to screen new customer applications specifically includes: Based on the customer number in the new customer's personal credit card approval record table, corresponding to the customer credit card issuance record table, determining whether the same customer association can be established; If the same customer association establishment fails, the approval record in the new customer personal credit card approval record table corresponding to the customer number is used as the new customer application; and If the same customer association is successfully established, the first credit card issuance date corresponding to the customer number is obtained from the customer credit card issuance record table, and the approval record corresponding to the customer number that is no later than the first credit card issuance date is used as the new customer application.

7. The method according to any one of claims 1 to 3, 5 to 6, characterized in that: The calculating the target single customer value of the customer to be approved based on the information of the customer to be approved specifically includes: Based on the information of the pending customers, calculate the lifetime revenue of a single customer and the cost of acquiring a single customer; Calculating the difference between the single-customer lifetime revenue and the single-customer acquisition cost to obtain a first difference; and The first difference is used as the target single customer value.

8. The method according to claim 7, characterized in that The calculation of the lifetime revenue of a single customer specifically includes: Obtain a single-customer daily revenue and a single-customer daily maintenance cost, calculate the difference between the single-customer daily revenue and the single-customer daily maintenance cost, and obtain a second difference: Obtaining a customer retention curve, and calculating a single customer retention duration using the area between the customer retention curve and the coordinate axis; as well as The product of the second difference and the single customer retention time is calculated to obtain the single customer lifetime revenue.

9. The method according to claim 7, characterized in that: Calculate the cost of acquiring a single customer, including: Obtain the total customer acquisition cost and number of new customers during the target time period; and The quotient of the total customer acquisition cost and the number of new customers is calculated to obtain the single customer acquisition cost.

10. The method according to claim 1, characterized in that The customer value model is an RFM model; The step of inputting the target single customer value into a customer value model to classify the customer value and output the customer value level specifically includes: Calculating the RFM score of the customer to be approved based on the target single customer value; and Based on a preset threshold, the RFM scores of the customers to be approved are classified into customer value categories, and a customer value grade is output.

11. A customer information approval device, characterized in that: The device comprises: A target single customer value calculation module is used to: obtain information of customers to be approved, and calculate the target single customer value of the customers to be approved based on the information of the customers to be approved; A customer value level acquisition module is used to: input the target single customer value into a customer value model to classify the customer value and output the customer value level; A pass rate prediction result output module is used to: in response to the customer value level being high value, input the information of the customer to be approved and the target single customer value into a pre-trained pass rate prediction model to predict the pass rate of customer information approval, and output a pass rate prediction result; and The approval result output module is used to: approve the customers to be approved based on the pass rate prediction result and output the approval result.

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

13. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the method according to any one of claims 1 to 10.

Images