Potential lost customer identification method and device

Abstract

The embodiment of the invention discloses a potential lost customer identification method and device, relates to the field of artificial intelligence, and can be used in the financial science and technology field, and the method comprises the steps: obtaining the historical payment record of a target customer at a bank end, and determining the historical payment feature data of the target customer according to the historical payment record; the historical payment feature data is input into a preset potential lost customer identification model, a first identification result corresponding to the target customer output by the potential lost customer identification model is obtained, the potential lost customer identification model is obtained by federated learning of the bank terminal and the charging enterprise terminal as participants, and the first identification result is obtained by the bank terminal and the charging enterprise terminal in the federated learning process. The bank end and the charging enterprise end generate first training samples according to local historical payment records of clients, and then train a local first classification model according to the first training samples. According to the invention, the accuracy and efficiency of mining potential lost customers of bank life payment services can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology and can be used in the field of financial technology. Specifically, it relates to a method and device for identifying potential lost customers. Background Art

[0002] At present, banks can handle agency payment services. Customers can pay for their daily life in the daily life payment section of the bank client, such as water bills, electricity bills, gas bills, Internet bills, and telephone bills. When customers pay for their daily life through banks, banks will collect the money on their behalf and transfer the money to the corresponding charging companies (for example, water companies, power companies, gas companies, and telecommunications companies).

[0003] At present, when bank business personnel are trying to win back customers who pay for their living expenses, they need to first identify potential lost customers of the living expense service, and then win back the customers by issuing coupons and other means. At present, when bank business personnel are trying to find potential lost customers of the living expense service, they often manually identify and judge which customers are potential lost customers based on the customers' living expense payment history data. The mining efficiency of potential lost customers is low and the accuracy cannot be guaranteed, which affects the subsequent recovery of potential lost customers. It can be seen that the prior art urgently needs a more accurate and efficient solution for mining potential lost customers of the bank's living expense payment service. Summary of the invention

[0004] In order to solve at least one technical problem in the above-mentioned background technology, the present invention proposes a method and device for identifying potential lost customers.

[0005] In order to achieve the above object, according to one aspect of the present invention, a method for identifying potential lost customers is provided, which is applied to a bank, and comprises:

[0006] Acquire the historical payment records of the target customer at the bank end, and determine the historical payment characteristic data of the target customer according to the historical payment records;

[0007] The historical payment feature data is input into a preset potential churn customer identification model to obtain a first identification result corresponding to the target customer output by the potential churn customer identification model, wherein the potential churn customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a first training sample based on the local customer historical payment records, and then each trains the local first classification model based on the first training sample.

[0008] Optionally, the method for identifying potential lost customers further includes:

[0009] The potential lost customer identification model is trained through multiple rounds of communication iterations of federated learning. In each round of communication iteration, the local first classification model is trained according to the local first training sample. After the training, the model parameters of the local first classification model are transmitted to the federated learning system, so that the federated learning system calculates the global model parameters of this round of communication iteration according to the model parameters transmitted by the bank end in this round of communication iteration and the model parameters transmitted by the charging enterprise end in this round of communication iteration. In each round of communication iteration of federated learning, the charging enterprise end trains the local first classification model according to the local first training sample. After the training, the model parameters of the local first classification model are transmitted to the federated learning system;

[0010] The parameters of the local first classification model are updated according to the global model parameters of this round of communication iteration.

[0011] Optionally, the method for identifying potential lost customers further includes:

[0012] Extract historical payment feature data for model training from local customer historical payment records used for model training;

[0013] The historical payment feature data used for model training is labeled with identification results to generate the first training sample.

[0014] Optionally, the historical payment characteristic data includes: the average number of payments in each time period, the average amount of each payment, the number of payments in each time period, and payment channel information.

[0015] Optionally, the first classification model is specifically a decision tree model.

[0016] Optionally, the first identification result includes: a potential lost customer and not a potential lost customer;

[0017] The method further comprises:

[0018] If the first identification result corresponding to the target customer is not a potential lost customer and the target customer has not activated the payment entrusted withholding at the bank end, determining the entrusted withholding characteristic data of the target customer according to the historical payment record and the entrusted withholding information of the target customer;

[0019] The entrusted withholding feature data is input into a preset entrusted withholding potential customer identification model to obtain a second identification result corresponding to the target customer output by the entrusted withholding potential customer identification model, wherein the entrusted withholding potential customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a second training sample based on the local customer's historical payment records and entrusted withholding information, and then each trains the local second classification model based on the second training sample.

[0020] Optionally, the method for identifying potential lost customers further includes:

[0021] The entrusted withholding potential customer identification model is trained through multiple rounds of communication iterations of federated learning. In each round of communication iteration, the local second classification model is trained according to the local second training sample. After the training, the model parameters of the local second classification model are transmitted to the federated learning system, so that the federated learning system calculates the global model parameters of this round of communication iteration according to the model parameters transmitted by the bank end in this round of communication iteration and the model parameters transmitted by the charging enterprise end in this round of communication iteration. In each round of communication iteration of federated learning, the charging enterprise end trains the local second classification model according to the local second training sample. After the training, the model parameters of the local second classification model are transmitted to the federated learning system;

[0022] The parameters of the local second classification model are updated according to the global model parameters of this round of communication iteration.

[0023] Optionally, the entrusted withholding characteristic data includes: whether the payment method of the payment item supports entrusted withholding information, the number of entrusted withholding agreements signed by customers, the average number of payments in each time period, the average amount of each payment, the number of payments in each time period, and payment channel information.

[0024] In order to achieve the above object, according to another aspect of the present invention, a device for identifying potential lost customers is provided, which is applied to a bank, and comprises:

[0025] A historical payment characteristic data generating unit, used for obtaining the historical payment records of the target customer at the bank end, and determining the historical payment characteristic data of the target customer according to the historical payment records;

[0026] A potential churn customer identification unit is used to input the historical payment feature data into a preset potential churn customer identification model to obtain a first identification result corresponding to the target customer output by the potential churn customer identification model, wherein the potential churn customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a first training sample based on the local customer historical payment records, and then each trains a local first classification model based on the first training sample.

[0027] In order to achieve the above-mentioned purpose, according to another aspect of the present invention, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for identifying potential lost customers when executing the computer program.

[0028] In order to achieve the above object, according to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program / instruction is stored, and when the computer program / instruction is executed by a processor, the steps of the above-mentioned potential lost customer identification method are implemented.

[0029] The beneficial effects of the present invention are:

[0030] The embodiment of the present invention obtains the historical payment records of the target customer at the bank end, determines the historical payment feature data of the target customer based on the historical payment records, and then inputs the historical payment feature data into a preset potential lost customer identification model to obtain the first identification result corresponding to the target customer output by the potential lost customer identification model, thereby realizing automatic identification of whether the target customer is a potential lost customer and improving the efficiency of potential lost customer mining. In addition, the potential lost customer identification model of the present invention is obtained by the bank end and the charging enterprise end as participants in federated learning, and the data of the charging enterprise end is also used in the training of the model. Compared with the model training using only the data of the bank end, the recognition accuracy of the model is higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0032] Figure 1 is a flow chart of a method for identifying potential lost customers according to an embodiment of the present invention;

[0033] Figure 2 This is a flow chart of the potential lost customer identification model training according to an embodiment of the present invention;

[0034] Figure 3 is a flow chart of generating a first training sample according to an embodiment of the present invention;

[0035] Figure 4 This is a flow chart of identifying a potential customer who is entrusted with withholding according to an embodiment of the present invention;

[0036] Figure 5 This is a flow chart of the training of a potential customer identification model for entrusted withholding according to an embodiment of the present invention;

[0037] Figure 6 is a schematic diagram of the architecture between various systems of the present invention;

[0038] Figure 7 is a structural block diagram of a device for identifying potential lost customers according to an embodiment of the present invention;

[0039] Figure 8 It is a schematic diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0041] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0042] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or apparatus.

[0043] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0044] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations.

[0045] It should be noted that the potential lost customer identification method and device of the present invention can be used in the field of financial technology, and can also be used in any field other than the field of financial technology. The application field of the potential lost customer identification method and device of the present invention is not limited.

[0046] Figure 6 It is a schematic diagram of the architecture between various systems of the present invention, such as Figure 6 As shown, the present invention relates to a bank side, a charging enterprise side and a federated learning system. The potential lost customer identification model and the entrusted withholding potential customer identification model of the present invention are obtained by the bank side and the charging enterprise side as participants in the federated learning.

[0047] The federated learning system is responsible for coordinating and managing the calculation process between the two parties, the bank side and the charging enterprise side. The bank side and the charging enterprise side can use the federated averaging algorithm to jointly build a model. Each participant trains its own model locally, and then transmits and aggregates the model parameters through the federated averaging algorithm to obtain a shared global model. The federated learning system includes a client and a server. The client is set up on the charging enterprise side and the bank side respectively, and the server side is responsible for coordinating and managing the entire federated learning process.

[0048] Figure 1 Flow chart of the method for identifying potential lost customers according to an embodiment of the present invention. Figure 1 As shown, in one embodiment of the present invention, the potential lost customer identification method of the present invention includes step S101 and step S102.

[0049] Step S101, obtaining the historical payment records of the target customer at the bank, and determining the historical payment characteristic data of the target customer according to the historical payment records.

[0050] In the present invention, this step obtains the historical data of the customer's living payment in the living payment section of the bank client, that is, the historical payment record, and then extracts the historical payment feature data of the target customer from the historical payment record.

[0051] In one embodiment of the present invention, the historical payment characteristic data includes: the average number of payments in each time period, the average amount of each payment, the number of payments in each time period, and payment channel information.

[0052] Step S102, input the historical payment feature data into a preset potential churn customer identification model, and obtain a first identification result corresponding to the target customer output by the potential churn customer identification model, wherein the potential churn customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a first training sample based on the local customer historical payment records, and then each trains the local first classification model based on the first training sample.

[0053] In one embodiment of the present invention, the first classification model is specifically a decision tree model.

[0054] The potential lost customer identification model of the present invention is obtained by the bank side and the charging enterprise side as participants in the federated learning. The training of the model also uses the data of the charging enterprise side. Compared with the model training using only the data of the bank side, the recognition accuracy of the model is higher. When performing federated learning, the charging enterprise side generates training samples based on the historical payment records of local customers, and then trains the local model. Since the charging enterprise side can correspond to multiple payment channels at the same time, such as third-party payment platforms such as Alipay and WeChat, and bank payment channels, the historical payment records in the charging enterprise side contain the historical data of customers paying through their respective payment channels. The data is relatively comprehensive and the data volume is large. The historical payment records of the charging enterprise side are used to generate training samples for model training, and the training effect is better. Compared with the model training using only the data of the bank side, the recognition accuracy of the model is higher.

[0055] In one embodiment of the present invention, if the first identification result corresponding to the target customer is a potential lost customer, the potential lost customer recovery process is entered. Specifically, in the potential lost customer recovery process, the bank can issue a payment coupon of a preset amount to the target customer to attract the customer to continue to use the bank's life payment service to pay.

[0056] In one embodiment of the present invention, the historical payment characteristic data specifically includes:

[0057] Payment frequency: the average number of payments made by a customer in each time period.

[0058] Payment amount: the average amount of each payment made by the customer.

[0059] Number of payments: The number of payments made by the customer in each time period.

[0060] Payment history: the customer’s payment records over the past period of time, such as the payment status in the last three months.

[0061] Payment channels: channels for customers to pay, such as WeChat, Alipay, bank mobile banking, apps operated by charging companies, etc.

[0062] Figure 2 : is a flow chart of the potential lost customer identification model training according to an embodiment of the present invention. Figure 2 As shown, in one embodiment of the present invention, the training process of the potential lost customer identification model in the above step S102 includes step S201 and step S202.

[0063] Step S201, train the potential churn customer identification model through multiple rounds of communication iterations of federated learning, and in each round of communication iteration, train the local first classification model according to the local first training sample, and transmit the model parameters of the local first classification model to the federated learning system after the training, so that the federated learning system calculates the global model parameters of this round of communication iteration according to the model parameters transmitted by the bank end in this round of communication iteration and the model parameters transmitted by the charging enterprise end in this round of communication iteration, wherein in each round of communication iteration of federated learning, the charging enterprise end trains the local first classification model according to the local first training sample, and transmits the model parameters of the local first classification model to the federated learning system after the training.

[0064] Step S202: update the parameters of the local first classification model according to the global model parameters of this round of communication iteration.

[0065] In one embodiment of the present invention, the specific process of federated learning of the present invention is as follows:

[0066] Initialization: Both parties (bank and charging company) initialize local model parameters w i .

[0067] Communication rounds: Federated learning consists of multiple rounds of communication iterations, each of which includes the following steps:

[0068] a. Local model training: Each participant uses local data and current local model parameters w i Perform local model training to obtain a local model

[0069] b. Model parameter transmission: The participants transmit the encrypted local model parameters Transmitted to the federated computing unit through a secure channel.

[0070] c. Model parameter aggregation: The federated computing unit receives and decrypts the parameters transmitted by all participants and calculates the average value to obtain the global model parameter w global :

[0071]

[0072] d. Model parameters are transmitted back to the participants: The federated computing unit transmits the encrypted global model parameters w global Transmitted back to each party.

[0073] Convergence and update: Repeat multiple rounds of communication iterations until the model converges or reaches a predetermined number of rounds. After each round of iteration, each participant uses the global model parameter w global to update the local model.

[0074] Figure 3 is a flow chart of generating the first training sample according to an embodiment of the present invention. Figure 3 As shown, in one embodiment of the present invention, the process of generating the first training sample includes step S301 and step S302.

[0075] Step S301, extracting historical payment feature data for model training from local customer historical payment records for model training;

[0076] Step S302: Add recognition result labels to the historical payment feature data used for model training to generate a first training sample.

[0077] In one embodiment of the present invention, the first recognition result includes: a potential lost customer and a non-potential lost customer. When generating training samples, the present invention adds a first recognition result label to the historical payment feature data used for model training to obtain the first training sample.

[0078] Similarly, when the charging enterprise trains the local model, it also needs to generate training samples. The process of generating training samples on the charging enterprise side includes: extracting historical payment feature data used for model training from the local customer historical payment records used for model training; marking the recognition result label on the historical payment feature data used for model training to generate the first training sample.

[0079] In some embodiments of the present invention, the present invention adopts a decision tree model for model training. The present invention constructs a CART algorithm model to complete parameter fitting and data training to predict the loss of paying customers. The specific process may include:

[0080] a. Feature extraction

[0081] Select a feature and the corresponding split point so that the data can obtain the best classification or regression results in the split subset.

[0082] b. Data segmentation

[0083] The data set is divided into subsets according to the selected features and split points.

[0084] c. Constructing CART decision tree

[0085] Recursively apply steps a and b to the subset until a termination condition is reached, such as the number of samples in the subset is less than a preset threshold or the depth of the tree reaches a preset maximum value.

[0086] d. Pruning optimization

[0087] By comparing the model performance before and after pruning, we decide whether to keep the branch to avoid overfitting.

[0088] Assuming that the present invention is to predict whether a customer is a potential churn customer (1 indicates yes, 0 indicates no) based on the customer's payment frequency (average number of payments in each time period) and payment amount (average amount of each payment), the training sample can be in the form of the following Table 1:

[0089] Customer ID Payment frequency Payment amount Is it a potential lost customer? 1 10 200 0 2 5 100 1 3 8 150 1 4 3 50 0

[0090] Table 1

[0091] Feature selection and data segmentation: For the payment frequency feature, assuming that the present invention selects a segmentation point of 7, the data is divided into two subsets, one group of users with a payment frequency less than or equal to 7, and the other group with a payment frequency greater than 7.

[0092] Building a tree: Continue to apply the CART algorithm on the subsets, repeating the process of feature selection and splitting the data until the termination condition is met.

[0093] Pruning: After the tree is built, pruning is performed to compare the model performance before and after pruning to avoid overfitting.

[0094] In the present invention, if the target customer is not a potential lost customer and the target customer has not opened payment entrusted withholding at the bank end, the present invention further determines whether the target customer is a potential customer for entrusted withholding, and then sends the determination result to the business personnel, so that the business personnel can accurately know which customers are potential customers for entrusted withholding, which is convenient for the business personnel to market the entrusted withholding business for living payment to these customers. Compared with the prior art solution of business personnel manually mining potential customers for entrusted withholding, the accuracy and efficiency are significantly improved.

[0095] Figure 4 is a flow chart of identifying a potential customer who entrusts withholding according to an embodiment of the present invention, such as Figure 4 As shown, in one embodiment of the present invention, the method of the present invention further includes step S401 and step S402.

[0096] Step S401, if the first identification result corresponding to the target customer is not a potential lost customer and the target customer has not activated the payment entrusted withholding at the bank end, determine the entrusted withholding characteristic data of the target customer according to the historical payment record and the entrusted withholding information of the target customer.

[0097] Step S402, input the entrusted withholding feature data into a preset entrusted withholding potential customer identification model, and obtain a second identification result corresponding to the target customer output by the entrusted withholding potential customer identification model, wherein the entrusted withholding potential customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a second training sample based on the local customer's historical payment records and entrusted withholding information, and then each trains the local second classification model based on the second training sample.

[0098] In one embodiment of the present invention, the entrusted withholding characteristic data includes: whether the payment method of the payment item supports entrusted withholding information, the number of entrusted withholding agreements signed by customers, the average number of payments in each time period, the average amount of each payment, the number of payments in each time period, and payment channel information.

[0099] In one embodiment of the present invention, the second identification result includes: a potential customer for entrusted withholding and a potential customer for entrusted withholding.

[0100] In one embodiment of the present invention, the entrusted withholding characteristic data specifically includes:

[0101] Status of signing of payment entrustment and withholding agreements: the number of entrusted withholding agreements signed by customers.

[0102] Payment method for payment items: whether entrusted deduction is supported.

[0103] Payment frequency: the average number of payments made by a customer in each time period.

[0104] Payment amount: the average amount of each payment made by the customer.

[0105] Number of payments: The number of payments made by the customer in each time period.

[0106] Payment history: the customer’s payment records over the past period of time, such as the payment status in the last three months.

[0107] Payment channels: channels for customers to pay, such as WeChat, Alipay, bank mobile banking, apps operated by charging companies, etc.

[0108] In one embodiment of the present invention, if the second identification result corresponding to the target customer is a potential customer for entrusted withholding, the marketing process for potential customers for entrusted withholding will be entered. Specifically, in the marketing process for potential customers for entrusted withholding, the bank can issue a coupon for entrusted withholding payment of a preset amount to the target customer to attract the customer to open an entrusted withholding for living payment in the bank.

[0109] Figure 5 : is a flow chart of the training of the potential customer identification model for entrusted withholding according to an embodiment of the present invention, such as Figure 5 As shown, in one embodiment of the present invention, the training process of the entrusted withholding potential customer identification model in the above step S402 includes step S501 and step S502.

[0110] Step S501, train the entrusted withholding potential customer identification model through multiple rounds of communication iterations of federated learning, and in each round of communication iteration, train the local second classification model according to the local second training samples, and transmit the model parameters of the local second classification model to the federated learning system after the training, so that the federated learning system calculates the global model parameters of this round of communication iteration according to the model parameters transmitted by the bank end in this round of communication iteration and the model parameters transmitted by the charging enterprise end in this round of communication iteration, wherein in each round of communication iteration of federated learning, the charging enterprise end trains the local second classification model according to the local second training samples, and transmits the model parameters of the local second classification model to the federated learning system after the training.

[0111] Step S502: update the parameters of the local second classification model according to the global model parameters of this round of communication iteration.

[0112] In one embodiment of the present invention, the second classification model is specifically a decision tree model.

[0113] In the present invention, when the bank and the charging enterprise are training the local model, they need to generate a second training sample. The process of generating the second training sample by the bank and the charging enterprise includes: extracting the entrusted withholding feature data used for model training from the local customer historical payment records used for model training and the entrusted withholding information used for model training; marking the second recognition result label on the entrusted withholding feature data used for model training to generate a second training sample.

[0114] In the present invention, the training processes of the potential customer identification model for entrusted with deduction and the potential churned customer identification model are the same except for the different training samples. The implementation example of the training process of the potential customer identification model for entrusted with deduction can refer to the implementation example of the training process of the potential churned customer identification model, and the details will not be repeated here.

[0115] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0116] Based on the same inventive concept, an embodiment of the present invention also provides a potential lost customer identification device, which can be used to implement the potential lost customer identification method described in the above embodiment, as described in the following embodiment. Since the principle of solving the problem by the potential lost customer identification device is similar to that of the potential lost customer identification method, the embodiment of the potential lost customer identification device can refer to the embodiment of the potential lost customer identification method, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0117] Figure 7 is a structural block diagram of a device for identifying potential lost customers according to an embodiment of the present invention. Figure 7 As shown, in one embodiment of the present invention, the potential lost customer identification device of the present invention comprises:

[0118] A historical payment characteristic data generating unit 1 is used to obtain the historical payment records of the target customer at the bank end, and determine the historical payment characteristic data of the target customer according to the historical payment records;

[0119] The potential churn customer identification unit 2 is used to input the historical payment feature data into a preset potential churn customer identification model to obtain a first identification result corresponding to the target customer output by the potential churn customer identification model, wherein the potential churn customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a first training sample based on the local customer historical payment records, and then each trains the local first classification model based on the first training sample.

[0120] In one embodiment of the present invention, the potential lost customer identification device of the present invention further includes:

[0121] The first training unit is used to train the potential churn customer identification model through multiple rounds of federated learning communication iterations. In each round of communication iteration, the local first classification model is trained according to the local first training sample. After the training, the model parameters of the local first classification model are transmitted to the federated learning system, so that the federated learning system calculates the global model parameters of this round of communication iteration according to the model parameters transmitted by the bank end in this round of communication iteration and the model parameters transmitted by the charging enterprise end in this round of communication iteration. In each round of communication iteration of federated learning, the charging enterprise end trains the local first classification model according to the local first training sample. After the training, the model parameters of the local first classification model are transmitted to the federated learning system; the parameters of the local first classification model are updated according to the global model parameters of this round of communication iteration.

[0122] In one embodiment of the present invention, the potential lost customer identification device of the present invention further includes:

[0123] A feature extraction unit, used to extract historical payment feature data for model training from local customer historical payment records used for model training;

[0124] The training sample generating unit is used to label the historical payment feature data used for model training with identification result labels to generate the first training sample.

[0125] In one embodiment of the present invention, the first identification result includes: a potential lost customer and not a potential lost customer;

[0126] In one embodiment of the present invention, the potential lost customer identification device of the present invention further includes:

[0127] a proxy withholding characteristic data generating unit, configured to determine the proxy withholding characteristic data of the target customer according to the historical payment record and the proxy withholding information of the target customer, if the first identification result corresponding to the target customer is not a potential lost customer and the target customer has not opened the proxy withholding for payment at the bank end;

[0128] The entrusted withholding potential customer identification unit is used to input the entrusted withholding feature data into a preset entrusted withholding potential customer identification model to obtain a second identification result corresponding to the target customer output by the entrusted withholding potential customer identification model, wherein the entrusted withholding potential customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a second training sample based on the local customer's historical payment records and entrusted withholding information, and then each trains the local second classification model based on the second training sample.

[0129] In one embodiment of the present invention, the potential lost customer identification device of the present invention further includes:

[0130] The second training unit trains the entrusted deduction potential customer identification model through multiple rounds of communication iterations of federated learning. In each round of communication iteration, the local second classification model is trained according to the local second training samples. After the training, the model parameters of the local second classification model are transmitted to the federated learning system, so that the federated learning system calculates the global model parameters of this round of communication iteration according to the model parameters transmitted by the bank end in this round of communication iteration and the model parameters transmitted by the charging enterprise end in this round of communication iteration. In each round of communication iteration of federated learning, the charging enterprise end trains the local second classification model according to the local second training samples. After the training, the model parameters of the local second classification model are transmitted to the federated learning system; the parameters of the local second classification model are updated according to the global model parameters of this round of communication iteration.

[0131] In order to achieve the above object, according to another aspect of the present application, a computer device is also provided. Figure 8 As shown, the computer device includes a memory, a processor, a communication interface and a communication bus. A computer program that can be executed on the processor is stored in the memory. When the processor executes the computer program, the steps in the above-mentioned embodiment method are implemented.

[0132] The processor may be a central processing unit (CPU). The processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0133] The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs and units, such as the corresponding program units in the above method embodiments of the present invention. The processor executes various functional applications of the processor and works data processing by running the non-transitory software programs, instructions and modules stored in the memory, that is, implementing the method in the above method embodiments.

[0134] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0135] The one or more units are stored in the memory, and when executed by the processor, the method in the above embodiment is performed.

[0136] The specific details of the above-mentioned computer device can be understood by referring to the corresponding related descriptions and effects in the above-mentioned embodiments, and will not be repeated here.

[0137] In order to achieve the above purpose, according to another aspect of the present application, a computer-readable storage medium is also provided, wherein the computer-readable storage medium stores a computer program, and the computer program implements the steps in the above-mentioned potential lost customer identification method when executed in a computer processor. Those skilled in the art can understand that the implementation of all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and the program can include the processes of the embodiments of the above-mentioned methods when executed. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviated: HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0138] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a computer program product is further provided, comprising a computer program / instruction, which implements the steps of the above-mentioned method for identifying potential lost customers when executed by a processor.

[0139] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0140] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for identifying potential lost customers, characterized in that: Applied to the bank side, the method includes: Acquire the historical payment records of the target customer at the bank end, and determine the historical payment characteristic data of the target customer according to the historical payment records; The historical payment feature data is input into a preset potential churn customer identification model to obtain a first identification result corresponding to the target customer output by the potential churn customer identification model, wherein the potential churn customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a first training sample based on the local customer historical payment records, and then each trains the local first classification model based on the first training sample.

2. The method for identifying potential lost customers according to claim 1, characterized in that: Also includes: The potential lost customer identification model is trained through multiple rounds of communication iterations of federated learning. In each round of communication iteration, the local first classification model is trained according to the local first training sample. After the training, the model parameters of the local first classification model are transmitted to the federated learning system, so that the federated learning system calculates the global model parameters of this round of communication iteration according to the model parameters transmitted by the bank end in this round of communication iteration and the model parameters transmitted by the charging enterprise end in this round of communication iteration. In each round of communication iteration of federated learning, the charging enterprise end trains the local first classification model according to the local first training sample. After the training, the model parameters of the local first classification model are transmitted to the federated learning system; The parameters of the local first classification model are updated according to the global model parameters of this round of communication iteration.

3. The method for identifying potential lost customers according to claim 2, characterized in that: Also includes: Extract historical payment feature data for model training from local customer historical payment records used for model training; The historical payment feature data used for model training is labeled with identification results to generate the first training sample.

4. The method for identifying potential lost customers according to claim 1, characterized in that: The historical payment characteristic data includes: the average number of payments in each time period, the average amount of each payment, the number of payments in each time period, and payment channel information.

5. The method for identifying potential lost customers according to claim 1, characterized in that: The first classification model is specifically a decision tree model.

6. The method for identifying potential lost customers according to claim 1, characterized in that: The first identification result includes: a potential lost customer and not a potential lost customer; The method further comprises: If the first identification result corresponding to the target customer is not a potential lost customer and the target customer has not activated the payment entrusted withholding at the bank end, determining the entrusted withholding characteristic data of the target customer according to the historical payment record and the entrusted withholding information of the target customer; The entrusted withholding feature data is input into a preset entrusted withholding potential customer identification model to obtain a second identification result corresponding to the target customer output by the entrusted withholding potential customer identification model, wherein the entrusted withholding potential customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a second training sample based on the local customer's historical payment records and entrusted withholding information, and then each trains the local second classification model based on the second training sample.

7. The method for identifying potential lost customers according to claim 6, characterized in that: Also includes: The entrusted withholding potential customer identification model is trained through multiple rounds of communication iterations of federated learning. In each round of communication iteration, the local second classification model is trained according to the local second training sample. After the training, the model parameters of the local second classification model are transmitted to the federated learning system, so that the federated learning system calculates the global model parameters of this round of communication iteration according to the model parameters transmitted by the bank end in this round of communication iteration and the model parameters transmitted by the charging enterprise end in this round of communication iteration. In each round of communication iteration of federated learning, the charging enterprise end trains the local second classification model according to the local second training sample. After the training, the model parameters of the local second classification model are transmitted to the federated learning system; The parameters of the local second classification model are updated according to the global model parameters of this round of communication iteration.

8. The method for identifying potential lost customers according to claim 6, characterized in that: The entrusted withholding characteristic data includes: whether the payment method of the payment item supports entrusted withholding information, the number of entrusted withholding agreements signed by customers, the average number of payments in each time period, the average amount of each payment, the number of payments in each time period, and payment channel information.

9. A device for identifying potential lost customers, characterized in that: Applied to a bank, the device comprises: A historical payment characteristic data generating unit, used for obtaining the historical payment records of the target customer at the bank end, and determining the historical payment characteristic data of the target customer according to the historical payment records; A potential churn customer identification unit is used to input the historical payment feature data into a preset potential churn customer identification model to obtain a first identification result corresponding to the target customer output by the potential churn customer identification model, wherein the potential churn customer identification model is obtained by the bank and the charging enterprise as participants in federated learning. During the federated learning process, the bank and the charging enterprise each generate a first training sample based on the local customer historical payment records, and then each trains a local first classification model based on the first training sample.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images