Credit card collection method and device
The credit card collection prediction model is used to identify the customer's collection success rate and screen the collection strategy based on this rate, which solves the problem of inefficient collection of existing credit card collection and realizes a more efficient collection process.
Patent Information
- Application Number
- CN202411848682.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-16
- Publication Date
- 2025-03-18
AI Technical Summary
The existing credit card collection methods are inefficient, and the collection personnel need a lot of manual operations and data integration, resulting in low collection efficiency.
By obtaining the customer information of the customers to be collected and entering them into the trained credit card collection prediction model, the collection success rate is obtained, and the collection strategy corresponding to the collection success rate is screened from the collection strategy database, and collecting according to the strategy is carried out.
By identifying customers with high success rates and low success rates, optimize collection strategies, reduce invalid collection, and improve overall collection efficiency.
Smart Images
Figure CN119648391A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and in particular to a credit card collection method and device. Background Art
[0002] In recent years, with the popularization of credit card business, the number of overdue customers and the overdue amount have continued to increase, and credit card collection work has faced new challenges.
[0003] At present, the credit card collection process generally relies on manual execution. When a customer's credit card is not repaid on time, the system will automatically send reminders by text message, phone call, email, etc. For some customers who do not respond to automatic reminders, the bank will arrange collection personnel to manually collect by phone or text message. However, the arrears of credit card customers vary. Collection personnel need to understand each customer's repayment ability, historical records, communication progress and other information one by one, which requires a lot of manual operations and data integration, resulting in low efficiency of credit card collection.
[0004] Therefore, how to improve the efficiency of credit card collection has become an urgent problem to be solved in this field. Summary of the invention
[0005] The present application provides a credit card collection method and device, the purpose of which is to improve the efficiency of credit card collection.
[0006] In order to achieve the above objectives, this application provides the following technical solutions:
[0007] A credit card collection method, comprising:
[0008] Obtain customer information of customers to be collected;
[0009] Inputting the customer information of the customer to be collected into a trained credit card collection prediction model to obtain a collection success rate; the trained credit card collection prediction model is obtained by training the credit card collection prediction model based on sample customer information;
[0010] Filtering out a collection strategy corresponding to the collection success rate from the collection strategy library and marking it as a target collection strategy;
[0011] Credit card collection is performed on the customer to be collected according to the target collection strategy.
[0012] Optionally, the step of inputting the customer information of the customer to be collected into a trained credit card collection prediction model to obtain a collection success rate includes:
[0013] Inputting the customer information of the customer to be collected into the encoding module of the trained credit card collection prediction model to obtain an encoded hidden layer vector;
[0014] Inputting the encoded hidden layer vector into a decoding module of a trained credit card collection prediction model to obtain a hidden layer vector carrying label data;
[0015] The hidden layer vector carrying the label data is input into the prediction module of the trained credit card collection prediction model to obtain the collection success rate.
[0016] Optionally, the process of training a credit card collection prediction model based on sample customer information to obtain the trained credit card collection prediction model includes:
[0017] Obtain sample customer information;
[0018] Inputting the sample customer information into a credit card collection prediction model to obtain a sample collection success rate;
[0019] Calculate the error between the sample collection success rate and the actual collection success rate corresponding to the sample customer information to obtain a loss function;
[0020] Determining whether the loss function converges;
[0021] If the loss function does not converge, the model parameters of the credit card collection prediction model are adjusted, and the step of inputting the sample customer information into the credit card collection prediction model to obtain the sample collection success rate is returned;
[0022] If the loss function converges, the credit card collection prediction model is determined as the trained credit card collection prediction model.
[0023] Optionally, obtaining sample customer information includes:
[0024] Screen target customers from the credit card database;
[0025] Query the target customer's personal attribute information and credit card overdue information from the database;
[0026] Sending the target customer's personal attribute information and credit card overdue information to the server, so that the server generates customer information according to the target customer's personal attribute information and credit card overdue information;
[0027] When receiving the customer information sent by the server, performing data preprocessing on the customer information to obtain preprocessed customer information;
[0028] The preprocessed customer information is subjected to feature enhancement using an autoencoder to obtain sample customer information.
[0029] Optionally, the step of selecting a collection strategy corresponding to the collection success rate from the collection strategy library and marking it as a target collection strategy includes:
[0030] Filter out the collection level corresponding to the collection success rate from all collection levels, and mark it as the target collection level;
[0031] A collection strategy corresponding to the target collection level is selected from the collection strategy library and marked as the target collection strategy.
[0032] A credit card collection device, comprising:
[0033] An acquisition unit, used to acquire customer information of customers to be collected;
[0034] A prediction unit, used for inputting the customer information of the customer to be collected into a trained credit card collection prediction model to obtain a collection success rate; the trained credit card collection prediction model is obtained by training the credit card collection prediction model based on sample customer information;
[0035] A screening unit, used to screen out a collection strategy corresponding to the collection success rate from the collection strategy library, and mark it as a target collection strategy;
[0036] The collection unit is used to collect credit card debts from the customer to be collected according to the target collection strategy.
[0037] Optionally, the prediction unit is specifically used for:
[0038] Inputting the customer information of the customer to be collected into the encoding module of the trained credit card collection prediction model to obtain an encoded hidden layer vector;
[0039] Inputting the encoded hidden layer vector into a decoding module of a trained credit card collection prediction model to obtain a hidden layer vector carrying label data;
[0040] The hidden layer vector carrying the label data is input into the prediction module of the trained credit card collection prediction model to obtain the collection success rate.
[0041] Optionally, the prediction unit is specifically used for:
[0042] Obtain sample customer information;
[0043] Inputting the sample customer information into a credit card collection prediction model to obtain a sample collection success rate;
[0044] Calculate the error between the sample collection success rate and the actual collection success rate corresponding to the sample customer information to obtain a loss function;
[0045] Determining whether the loss function converges;
[0046] If the loss function does not converge, the model parameters of the credit card collection prediction model are adjusted, and the step of inputting the sample customer information into the credit card collection prediction model to obtain the sample collection success rate is returned;
[0047] If the loss function converges, the credit card collection prediction model is determined as the trained credit card collection prediction model.
[0048] Optionally, the prediction unit is specifically used for:
[0049] Screen target customers from the credit card database;
[0050] Query the target customer's personal attribute information and credit card overdue information from the database;
[0051] Sending the target customer's personal attribute information and credit card overdue information to the server, so that the server generates customer information according to the target customer's personal attribute information and credit card overdue information;
[0052] When receiving the customer information sent by the server, performing data preprocessing on the customer information to obtain preprocessed customer information;
[0053] The preprocessed customer information is subjected to feature enhancement using an autoencoder to obtain sample customer information.
[0054] Optionally, the collection unit is specifically used for:
[0055] Filter out the collection level corresponding to the collection success rate from all collection levels, and mark it as the target collection level;
[0056] A collection strategy corresponding to the target collection level is selected from the collection strategy library and marked as the target collection strategy.
[0057] The technical solution provided by this application obtains the customer information of the customer to be collected; inputs the customer information of the customer to be collected into a trained credit card collection prediction model to obtain the collection success rate; selects the collection strategy corresponding to the collection success rate from the collection strategy library and marks it as the target collection strategy; and performs credit card collection on the customer to be collected according to the target collection strategy. In this application, customers with high success rates and low success rates are identified through the credit card collection prediction model, the collection strategy is optimized, and corresponding collection measures are taken according to the collection success rates of different customers, thereby reducing ineffective collection and improving overall collection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0059] Figure 1 A flowchart of a credit card collection method provided in an embodiment of the present application;
[0060] Figure 2 A flowchart of a method for obtaining a collection success rate provided in an embodiment of the present application;
[0061] Figure 3 A schematic diagram of a model structure provided in an embodiment of the present application;
[0062] Figure 4 A flowchart of a method for training a credit card collection prediction model provided in an embodiment of the present application;
[0063] Figure 5 A flow chart of obtaining sample customer information provided in an embodiment of the present application;
[0064] Figure 6 A schematic diagram of a credit card collection process provided in an embodiment of the present application;
[0065] Figure 7 A schematic diagram of the architecture of a credit card collection device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0066] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0067] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.
[0068] like Figure 1FIG. 1 is a flowchart of a credit card collection method provided in an embodiment of the present application, comprising the following steps:
[0069] S101: Acquire customer information of customers to be collected.
[0070] Among them, customer information includes customer's personal attribute information (such as name, age, etc.) and credit card expected information (such as overdue amount, overdue days).
[0071] S102: Input the customer information of the customer to be collected into the trained credit card collection prediction model to obtain the collection success rate.
[0072] Among them, the trained credit card collection prediction model is obtained by training the credit card collection prediction model based on sample customer information.
[0073] Optionally, the credit card collection prediction model includes but is not limited to a Transformer model.
[0074] Optionally, in another embodiment of the present application, the specific implementation of step S102 is as follows: Figure 2 As shown, the following steps are included:
[0075] S201: Input the customer information of the customer to be collected into the encoding module of the trained credit card collection prediction model to obtain the encoded hidden layer vector.
[0076] It can be understood that the customer information of the customer to be collected is input into the encoding module of the trained credit card collection prediction model, and the encoding module uses the Transformer encoder to encode the customer information of the customer to be collected to obtain the encoded hidden layer vector.
[0077] S202: Input the encoded hidden vector into the decoding module of the trained credit card collection prediction model to obtain a hidden vector carrying label data.
[0078] The decoding module includes but is not limited to: a fully connected layer.
[0079] It can be understood that the encoded hidden vector is input into the decoding module of the trained credit card collection prediction model. The decoding module uses a decoder to decode the encoded hidden vector to obtain the decoded hidden vector, and then through error back propagation, the parameters are continuously updated to make and become closer and closer, and finally a hidden vector that can well represent the original data information is obtained, that is, a hidden vector carrying label data.
[0080] S203: Input the hidden layer vector carrying the label data into the prediction module of the trained credit card collection prediction model to obtain the collection success rate.
[0081] It can be understood that the latent vector carrying the label data is input into the prediction module of the trained credit card collection prediction model, and the prediction module uses a classification algorithm to calculate the latent vector carrying the label data to obtain the collection success rate.
[0082] It should be noted that the use of layer normalization and residual connections in each sublayer of the Transformer model helps to speed up the convergence of the model during training, and can more effectively propagate signals during the training process to avoid the problem of gradient disappearance or explosion. The multi-head attention mechanism can capture patterns in the data from different angles, calculate the attention scores of each data feature in parallel, and allocate computing power according to the importance of the feature. For features that are more important to the current prediction task, the model assigns a larger weight value to it, and assigns a lower weight value to unimportant features. Specifically, the model structure is as follows: Figure 3 shown.
[0083] Optionally, in another embodiment of the present application, a specific implementation method of training a credit card collection prediction model based on sample customer information to obtain a trained credit card collection prediction model is as follows: Figure 4 As shown, the following steps are included:
[0084] S401: Obtain sample customer information.
[0085] Among them, the sample customer information includes the customer's personal attribute information and credit card overdue information.
[0086] Optionally, in another embodiment of the present application, the specific implementation of step S301 is as follows: Figure 5 As shown, the following steps are included:
[0087] S501: Filter out target customers from the credit card database.
[0088] Optionally, target customers may be screened out from user loan information in a credit card database.
[0089] It should be noted that all customers can be screened out from the credit card database as target customers, or specific customers can be screened out as target customers based on business needs.
[0090] S502: Query the target customer's personal attribute information and credit card overdue information from the database.
[0091] S503: Send the target customer's personal attribute information and credit card overdue information to the server, so that the server generates customer information according to the target customer's personal attribute information and credit card overdue information.
[0092] It can be understood that the personal attribute information and credit card overdue information of the target customer are sent to the server so that the server can integrate the personal attribute information and credit card overdue information of the target customer, thereby extracting derivative information such as risk level labels, consumer behavior labels, credit scores, etc., and generating user portraits (i.e., customer information) based on the derivative information.
[0093] S504: When the customer information sent by the server is received, data preprocessing is performed on the customer information to obtain preprocessed customer information.
[0094] It is understandable that since the data types in customer information are different, including both discrete data and continuous data, directly feeding it into the model will affect the learning effect of the model. Therefore, before model training, customer information needs to be preprocessed, including data cleaning, data deduplication, missing value filling and data normalization.
[0095] Specifically, a small number of duplicate values and outliers are deleted, missing values are filled using K nearest neighbors for discrete variables, and one-hot encoding is used for them, and specific values (such as 0) are used to fill missing values for continuous variables, and normalization is performed using a normalization formula. The specific expression of the normalization formula is shown in formula (1).
[0096] (1)
[0097] In formula (1), x is the data to be normalized, x min is the minimum value of preprocessed data, x max is the maximum value of preprocessed data, x scale This is the normalized data.
[0098] It should be noted that after data preprocessing of customer information, the preprocessed customer information is divided into training set (70%) and validation set (30%). The credit card collection prediction model is trained with the training set. After the credit card collection prediction model is trained, the validation set is used to verify the credit card collection prediction model. The verification process is consistent with the training process.
[0099] S505: Use an autoencoder to perform feature enhancement on the preprocessed customer information to obtain sample customer information.
[0100] Among them, the autoencoder includes but is not limited to: Autoencoder neural network model.
[0101] It should be noted that in order to solve the high-dimensional and dense characteristics of the pre-processed customer information, the Autoencoder neural network model is used to combine the attributes of different features to enhance the features of the pre-processed customer information, thereby increasing the dimension of the feature vector in the sample customer information.
[0102] S402: Input the sample customer information into the credit card collection prediction model to obtain the sample collection success rate.
[0103] Among them, the sample collection success rate indicates the collection success rate corresponding to the sample customer information.
[0104] S403: Calculate the error between the sample collection success rate and the actual collection success rate corresponding to the sample customer information to obtain a loss function.
[0105] It should be noted that for continuous variables, square error loss is used, and for discrete variables, cross loss is used. Specifically, the specific form of the loss function is shown in formula (2).
[0106] (2)
[0107] In formula (2), loss is the loss function, is a set of continuous variables, is a set of discrete variables, x i is the input vector of the ith variable, is the vector reconstructed by autoencoding of the jth variable, is the prediction result of the j-th variable obtained by softmax (i.e., the sample collection success rate), x j is the input vector of the j-th variable.
[0108] S404: Determine whether the loss function converges.
[0109] If the loss function has not converged, step S405 is executed; if the loss function has converged, step S406 is executed.
[0110] It can be understood that whether the loss function converges is judged. If the loss function converges, that is, the loss function gradually decreases with the increase of training steps, and finally tends to be stable or no longer changes significantly, it means that the credit card collection prediction model training is completed at this time.
[0111] S405: Adjust the model parameters of the credit card collection prediction model and return to step S402.
[0112] It should be noted that if the loss function does not converge, it means that the loss function does not decrease with the increase of training steps. At this time, the model parameters in the credit card collection prediction model are adjusted, and the process returns to step S402 until the loss function converges.
[0113] S406: Determine the credit card collection prediction model as the trained credit card collection prediction model.
[0114] S103: Filter out a collection strategy corresponding to the collection success rate from the collection strategy library, and mark it as a target collection strategy.
[0115] The collection strategies include but are not limited to: SMS reminder strategy, telephone collection strategy, and letter generation strategy.
[0116] SMS reminder strategy: When receiving the SMS collection queue, the intelligent SMS reminder sends repayment reminders, overdue penalty interest and collection notices to the target customers via SMS. At the same time, based on the customer's repayment situation, it automatically selects the appropriate SMS template and sends it in batches during a specific time period. Intelligent SMS reminders not only improve collection efficiency, but also reduce manual mis-sending and missed sending.
[0117] Telephone collection strategy: When receiving a telephone collection queue, the intelligent telephone collection will proactively call the customer according to the established strategy. Through voice recognition and natural language processing technology, it can judge the customer's repayment intention in real time based on the nodes where the customer interacts with the virtual customer service, and generate a corresponding customer portrait. The interaction nodes include whether the call is answered normally, the number of conversations, the length of the call, the customer's repayment intention level, etc. During operation, it will track and analyze the collection status and customer feedback, use machine learning functions to optimize the collection voice content, and continuously improve and optimize the collection strategy.
[0118] For example, if the customer answers the call himself, the customer's repayment intention is intelligently judged based on the conversation content. When the customer's current repayment intention is not high, the outbound call specialist is reminded to intervene in time, giving full play to the advantages of manual seats. If the customer does not answer the call himself, the intelligent telephone collection will label it, stop repeated outbound calls to such accounts, prevent customer complaints, and further improve the effectiveness of collection work. After completing the outbound call operation, the case is divided into labels such as promised repayment, refusal to repay, inability to repay, anti-collection, and high complaint risk according to the collection results, and feedback is sent to the server to complete the process loop and provide data support for the next round of predictions.
[0119] The letter generation strategy, when receiving the collection queue, fills the customer overdue data returned by the database into the letter template and generates Word files in batches. At the same time, it can automatically adapt to the changes in the letter template format based on actual needs.
[0120] Optionally, in another embodiment of the present application, a specific implementation of step S103 includes:
[0121] The collection level corresponding to the collection success rate is screened out from all collection levels and marked as the target collection level.
[0122] Optionally, the collection level can be divided into the first level, the second level, the third level, the fourth level and the fifth level, the first level> the second level> the third level> the fourth level> the fifth level. The collection level can also be divided into high level, medium level and low level.
[0123] It should be noted that the various intervals of collection success rate are divided in advance, and each interval corresponds to a collection level.
[0124] For example, the success rate of 0.8-1.0 is the first level, the success rate of 0.6-0.8 is the second level, the success rate of 0.4-0.6 is the third level, the success rate of 0.2-0.4 is the fourth level, and the others are the fifth level.
[0125] A collection strategy corresponding to the target collection level is selected from the collection strategy library and marked as the target collection strategy.
[0126] For example, when the level is 1 or 2, the SMS reminder strategy is matched first, and when the level is 3, 4, or 5, the system prioritizes matching the telephone collection strategy or the letter generation strategy.
[0127] In addition, the business can modify the strategy priority based on the aforementioned customer information to achieve differentiated customization of collection strategies.
[0128] Specifically, differentiated collection strategies include: adjusting the content of outbound calls based on the debt situation and repayment ability of the collection target; providing appropriate interest reduction and repayment plans and installment repayment plans to increase customers' willingness to repay; real-time tracking of changes in the payment period of overdue accounts, and timely freezing or stopping payment of credit card accounts with malicious overdrafts; and intervening in door-to-door collection or legal letter collection when necessary. The above-mentioned credit card collection prediction module can realize pre-overdue reminders and overdue contact under the premise of effective resource allocation, further improving the collection effect.
[0129] S104: Collect credit card debts from customers to be collected according to the target collection strategy.
[0130] It is understandable that credit card collection is carried out on the collection customers according to the target collection strategy, the collection results are obtained, the collection results are monitored and counted in real time, and the collection results are fed back to the collection prediction model for optimization. For cases where collection is unsuccessful, the collection strategy is automatically adjusted, and a new collection strategy is predicted and executed again. If the collection is successful, wait for the subsequent evaluation of the repayment situation, and enter the collection date and collection method. At the same time, based on the analysis results, the loan collection information is generated and pushed to the automatic collection platform and the custody manager through multiple channels such as the data application platform, and the configuration management of closing or opening each push channel is carried out.
[0131] In order to better illustrate the above content, the following is a detailed description combined with specific application scenarios. Figure 6 , identify sample target customers, and then generate customer portraits (i.e., sample customer information), including basic information and derivative information; preprocess the sample customer information to obtain preprocessed sample customer information, and use the preprocessed sample customer information to train the credit card prediction model to obtain a trained credit card prediction model; input the customer information of the customer to be collected into the trained credit card collection prediction model to obtain the collection success rate; match the collection strategy according to the collection success rate and execute it to obtain the collection result; push the collection result to the message management center, and adjust the credit card collection prediction model according to the collection result.
[0132] In summary, the credit card collection prediction model is used to identify customers with high and low success rates, optimize the collection strategy, and take corresponding collection measures based on the collection success rates of different customers, thereby reducing ineffective collection and improving overall collection efficiency.
[0133] like Figure 7 , which is a schematic diagram of the architecture of a credit card collection device provided in an embodiment of the present application, the credit card collection device includes: an acquisition unit 100, a prediction unit 200, a screening unit 300 and a collection unit 400.
[0134] The acquisition unit 100 is used to acquire the customer information of the customer to be collected.
[0135] The prediction unit 200 is used to input the customer information of the customer to be collected into the trained credit card collection prediction model to obtain the collection success rate; the trained credit card collection prediction model is trained based on the sample customer information.
[0136] The prediction unit 200 is specifically used for: inputting the customer information of the customer to be collected into the encoding module of the trained credit card collection prediction model to obtain the encoded hidden vector; inputting the encoded hidden vector into the decoding module of the trained credit card collection prediction model to obtain the hidden vector carrying the label data; inputting the hidden vector carrying the label data into the prediction module of the trained credit card collection prediction model to obtain the collection success rate.
[0137] The prediction unit 200 is specifically used to: obtain sample customer information; input the sample customer information into the credit card collection prediction model to obtain a sample collection success rate; calculate the error between the sample collection success rate and the actual collection success rate corresponding to the sample customer information to obtain a loss function; determine whether the loss function converges; if the loss function does not converge, adjust the model parameters of the credit card collection prediction model, and return to execute the step of inputting the sample customer information into the credit card collection prediction model to obtain the sample collection success rate; if the loss function converges, determine the credit card collection prediction model as a trained credit card collection prediction model.
[0138] The prediction unit 200 is specifically used for: querying the personal attribute information and credit card overdue information of the target customer from the database; sending the personal attribute information and credit card overdue information of the target customer to the server, so that the server generates customer information according to the personal attribute information and credit card overdue information of the target customer; when receiving the customer information sent by the server, performing data preprocessing on the customer information to obtain preprocessed customer information; and using the autoencoder to perform feature enhancement on the preprocessed customer information to obtain sample customer information.
[0139] The screening unit 300 is used to screen out a collection strategy corresponding to the collection success rate from the collection strategy library and mark it as a target collection strategy.
[0140] The collection unit 400 is used to collect credit card debts from customers to be collected according to a target collection strategy.
[0141] The collection unit 400 is specifically used to: select a collection level corresponding to the collection success rate from all collection levels, and mark it as the target collection level; select a collection strategy corresponding to the target collection level from the collection strategy library, and mark it as the target collection strategy.
[0142] In summary, the credit card collection prediction model is used to identify customers with high and low success rates, optimize the collection strategy, and take corresponding collection measures based on the collection success rates of different customers, thereby reducing ineffective collection and improving overall collection efficiency.
[0143] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.
[0144] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0145] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A credit card collection method, characterized in that: include: Obtain customer information of customers to be collected; Inputting the customer information of the customer to be collected into a trained credit card collection prediction model to obtain a collection success rate; the trained credit card collection prediction model is obtained by training the credit card collection prediction model based on sample customer information; Filtering out a collection strategy corresponding to the collection success rate from the collection strategy library and marking it as a target collection strategy; Credit card collection is performed on the customer to be collected according to the target collection strategy.
2. The method according to claim 1, characterized in that The step of inputting the customer information of the customer to be collected into the trained credit card collection prediction model to obtain the collection success rate includes: Inputting the customer information of the customer to be collected into the encoding module of the trained credit card collection prediction model to obtain an encoded hidden layer vector; Inputting the encoded hidden layer vector into a decoding module of a trained credit card collection prediction model to obtain a hidden layer vector carrying label data; The hidden layer vector carrying the label data is input into the prediction module of the trained credit card collection prediction model to obtain the collection success rate.
3. The method according to claim 1, characterized in that The process of training a credit card collection prediction model based on sample customer information to obtain the trained credit card collection prediction model includes: Obtain sample customer information; Inputting the sample customer information into a credit card collection prediction model to obtain a sample collection success rate; Calculate the error between the sample collection success rate and the actual collection success rate corresponding to the sample customer information to obtain a loss function; Determining whether the loss function converges; If the loss function does not converge, the model parameters of the credit card collection prediction model are adjusted, and the step of inputting the sample customer information into the credit card collection prediction model to obtain the sample collection success rate is returned; If the loss function converges, the credit card collection prediction model is determined as the trained credit card collection prediction model.
4. The method according to claim 1, characterized in that The obtaining of sample customer information includes: Screen target customers from the credit card database; Query the target customer's personal attribute information and credit card overdue information from the database; Sending the target customer's personal attribute information and credit card overdue information to the server, so that the server generates customer information according to the target customer's personal attribute information and credit card overdue information; When receiving the customer information sent by the server, performing data preprocessing on the customer information to obtain preprocessed customer information; The preprocessed customer information is subjected to feature enhancement using an autoencoder to obtain sample customer information.
5. The method according to claim 1, characterized in that The step of selecting a collection strategy corresponding to the collection success rate from the collection strategy library and marking it as a target collection strategy includes: Filter out the collection level corresponding to the collection success rate from all collection levels, and mark it as the target collection level; A collection strategy corresponding to the target collection level is selected from the collection strategy library and marked as the target collection strategy.
6. A credit card collection device, characterized in that: include: An acquisition unit, used to acquire customer information of customers to be collected; A prediction unit, used for inputting the customer information of the customer to be collected into a trained credit card collection prediction model to obtain a collection success rate; the trained credit card collection prediction model is obtained by training the credit card collection prediction model based on sample customer information; A screening unit, used to screen out a collection strategy corresponding to the collection success rate from the collection strategy library, and mark it as a target collection strategy; The collection unit is used to collect credit card debts from the customer to be collected according to the target collection strategy.
7. The device according to claim 6, characterized in that The prediction unit is specifically used for: Inputting the customer information of the customer to be collected into the encoding module of the trained credit card collection prediction model to obtain an encoded hidden layer vector; Inputting the encoded hidden layer vector into a decoding module of a trained credit card collection prediction model to obtain a hidden layer vector carrying label data; The hidden layer vector carrying the label data is input into the prediction module of the trained credit card collection prediction model to obtain the collection success rate.
8. The device according to claim 6, characterized in that The prediction unit is specifically used for: Obtain sample customer information; Inputting the sample customer information into a credit card collection prediction model to obtain a sample collection success rate; Calculate the error between the sample collection success rate and the actual collection success rate corresponding to the sample customer information to obtain a loss function; Determining whether the loss function converges; If the loss function does not converge, the model parameters of the credit card collection prediction model are adjusted, and the step of inputting the sample customer information into the credit card collection prediction model to obtain the sample collection success rate is returned; If the loss function converges, the credit card collection prediction model is determined as the trained credit card collection prediction model.
9. The device according to claim 6, characterized in that The prediction unit is specifically used for: Screen target customers from the credit card database; Query the target customer's personal attribute information and credit card overdue information from the database; Sending the target customer's personal attribute information and credit card overdue information to the server, so that the server generates customer information according to the target customer's personal attribute information and credit card overdue information; When receiving the customer information sent by the server, performing data preprocessing on the customer information to obtain preprocessed customer information; The preprocessed customer information is subjected to feature enhancement using an autoencoder to obtain sample customer information.
10. The device according to claim 6, characterized in that The collection unit is specifically used for: Filter out the collection level corresponding to the collection success rate from all collection levels, and mark it as the target collection level; A collection strategy corresponding to the target collection level is selected from the collection strategy library and marked as the target collection strategy.