Payment urging method and device, electronic equipment and storage medium
By screening and clustering the payer's data, determining the difficulty of urging and selecting the target urging objects, the problem of inefficiency in the existing payment urging methods is solved, and efficient urging for different payers is achieved.
Patent Information
- Application Number
- CN202411074930.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-22
AI Technical Summary
The existing payment urge method is inefficient, and the urge to handle cases is not distributed in a timely manner or the business personnel are unable to urge to handle the payer, resulting in inefficiency.
By obtaining payer data and urged data, the target data is screened using a pre-trained classification model, and clustering is performed to determine the difficulty of urged, and the target urged object is selected from the candidate urged object for urged object.
It effectively avoids the situation of untimely urging and business personnel being unable to urging, improves the efficiency of payment and urging, and realizes matching the corresponding urging parties for urging and urging.
Smart Images

Figure CN119004175A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and in particular to a payment urging method and device, an electronic device and a storage medium. Background Art
[0002] At present, the main method for urging payment is to have business managers assign urging cases to business personnel, and then have business personnel urge the payers of the assigned urging cases. However, this method is based on manual allocation and urging of urging cases, which results in untimely allocation of urging cases or inability of business personnel to urge the payers during the urging process, resulting in low efficiency of payment urging. How to improve the efficiency of payment urging has become an urgent problem to be solved. Summary of the invention
[0003] The main purpose of the embodiments of the present application is to propose a payment reminder method and device, an electronic device and a storage medium, aiming to improve the efficiency of payment reminders.
[0004] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a payment urging method, the method comprising:
[0005] Obtaining payer data and expediter data; wherein the expediter data includes candidate expediting objects;
[0006] Filtering target data from the payer data according to a pre-trained classification model; wherein the target data includes a target object;
[0007] Performing clustering processing on the target objects to obtain the difficulty of urging the target objects;
[0008] Selecting a target urging object from the candidate urging objects according to the urging difficulty;
[0009] The target object is urged to be handled through the target urging object.
[0010] In some embodiments, filtering out target data from the payer data according to a pre-trained classification model includes:
[0011] Obtaining an initial classification model before training, and obtaining a true classification label corresponding to the payer data;
[0012] Extracting features from the payer data using the initial classification model to obtain payer features;
[0013] Predicting the payer category to which the payer data belongs according to the payer characteristics to obtain a predicted payer category;
[0014] Calculating the residual between the predicted payer category and the true classification label;
[0015] Calculating a gradient estimate of the payer characteristic based on the residual;
[0016] Training the initial classification model according to the gradient estimation value to obtain the pre-trained classification model;
[0017] Performing a final prediction on the payer category to which the payer data belongs according to the pre-trained classification model, and outputting a payer classification result; wherein the payer classification result is used to indicate the payment expediting risk level of the payer;
[0018] The target data is screened out from the payer data according to the payer classification result.
[0019] In some embodiments, the payer features include payer category features and combined category features; the extracting features from the payer data using the initial classification model to obtain the payer features includes:
[0020] Performing initial feature extraction on the payer data using the initial classification model to obtain payer category features; the payer category features are used to represent discrete payer category features;
[0021] Converting the payer category feature into a numerical category feature; the numerical category feature is used to represent a numerical payer category feature;
[0022] Identifying associations between the payer category characteristics;
[0023] The payer category features are combined according to the association relationship to obtain a combined category feature.
[0024] In some embodiments, the finally predicting the payer category to which the payer data belongs according to the pre-trained classification model and outputting the payer classification result includes:
[0025] Traversing each leaf node of the classification tree in the classification model to obtain a category score of the payer category to which the payer data output by each leaf node belongs; the leaf node stores the payer data and the category score of the payer category to which the payer data belongs;
[0026] A final prediction is made on the payer category to which the payer data belongs based on the category score, and the payer classification result is output.
[0027] In some embodiments, clustering the target objects to obtain the difficulty of urging the target objects includes:
[0028] Obtaining the number of clusters of the target object and obtaining cluster distribution parameters of the target object;
[0029] Calculating the cluster distribution probability of the target object according to the cluster quantity and the cluster distribution parameter;
[0030] The difficulty of urging the target object is determined according to the cluster distribution probability.
[0031] In some embodiments, calculating the cluster distribution probability of the target object according to the cluster quantity and the cluster distribution parameter includes:
[0032] Calculating the initial distribution probability of the target object according to the number of clusters and the cluster distribution parameters;
[0033] Optimizing the cluster distribution parameters of the target object according to the initial distribution probability to obtain optimized cluster distribution parameters;
[0034] The cluster distribution of the target object is calculated according to the cluster quantity and the optimized cluster distribution parameters; the cluster distribution probability is used to represent the maximum probability of the cluster category to which the target object belongs.
[0035] In some embodiments, selecting a target urging object from the candidate urging objects according to the urging difficulty includes:
[0036] Scoring the candidate expediting objects for expediting ability according to the preset expediting ability index to obtain an initial ability score;
[0037] Obtaining the indicator weight of the expediting capability indicator;
[0038] Performing a weighted operation on the initial capability score according to the indicator weight to obtain a target capability score;
[0039] The target urging object is selected from the candidate urging objects according to the urging difficulty and the target ability score.
[0040] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a payment urging device, the device comprising:
[0041] A payment urging data acquisition module is used to acquire payer data and urging party data; wherein the urging party data includes candidate urging objects;
[0042] A payer screening module, used to screen target data from the payer data according to a pre-trained classification model; wherein the target data includes a target object;
[0043] A payer clustering module, used for clustering the target objects to obtain the difficulty of urging the target objects;
[0044] An expedited object selection module is used to select a target expedited object from the candidate expedited objects according to the expedited difficulty;
[0045] The payment urging module is used to urge the target object through the target urging object.
[0046] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.
[0047] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.
[0048] The payment collection method and device, electronic device and storage medium proposed in the present application obtain the payer data and the collection party data; wherein the collection party data includes candidate collection objects; the target data is screened out from the payer data according to the pre-trained classification model; wherein the target data includes the target object; the target object is clustered to obtain the collection difficulty of the target object; the target collection object is selected from the candidate collection objects according to the collection difficulty; and the target object is collected through the target collection object. The present application can classify the payers according to the collection difficulty by screening and clustering the payers, so as to facilitate the screening of the corresponding collection parties for different payers, and can match the corresponding collection parties for collection for payers with different collection difficulties, effectively avoiding the situation where the collection is not timely or the business personnel are unable to collect the payers in the collection process, thereby improving the efficiency of payment collection. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 is a flow chart of a payment urging method provided in an embodiment of the present application;
[0050] Figure 2 yes Figure 1 Flow chart of step S102 in FIG.
[0051] Figure 3 yes Figure 2 Flow chart of step S202 in FIG.
[0052] Figure 4 yes Figure 2 Flow chart of step S207 in FIG.
[0053] Figure 5 yes Figure 1 Flow chart of step S103 in FIG.
[0054] Figure 6 yes Figure 5 Flow chart of step S502 in FIG.
[0055] Figure 7 yes Figure 1 Flow chart of step S104 in FIG.
[0056] Figure 8 It is a structural diagram of a payment urging device provided in an embodiment of the present application;
[0057] Fig. 9 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION
[0058] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0059] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.
[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.
[0061] First, the nouns involved in this application are analyzed:
[0062] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0063] Based on this, the embodiments of the present application provide a payment reminder method and device, an electronic device and a storage medium, aiming to improve the efficiency of payment reminder.
[0064] The payment urging method and device, electronic device and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the payment urging method in the embodiments of the present application is described.
[0065] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0066] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.
[0067] The payment urging method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The payment urging method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the payment urging method, etc., but is not limited to the above forms.
[0068] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.
[0069] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user payment data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.
[0070] Figure 1 is an optional flow chart of the payment urging method provided in the embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S105.
[0071] Step S101, obtaining payer data and expediter data; wherein the expediter data includes candidate expediting objects.
[0072] Step S102, filtering out target data from the payer data according to the pre-trained classification model; wherein the target data includes a target object.
[0073] Step S103, clustering the target objects to obtain the difficulty of urging the target objects.
[0074] Step S104, selecting a target urging object from the candidate urging objects according to the urging difficulty.
[0075] Step S105, urging the target object through the target urging object.
[0076] Steps S101 to S105 shown in the embodiment of the present application are performed by obtaining payer data and urging party data; wherein the urging party data includes candidate urging objects; filtering out target data from the payer data according to a pre-trained classification model; wherein the target data includes a target object; clustering the target object to obtain the difficulty of urging the target object; selecting the target urging object from the candidate urging objects according to the difficulty of urging; and urging the target object through the target urging object. The present application can classify payers according to the difficulty of urging by screening and clustering payers, so as to facilitate screening out corresponding urging parties for different payers, and can match corresponding urging parties for urging payers with different urging difficulties, effectively avoiding the situation where manual urging is not timely or the business personnel are unable to urge the payer during the urging process, thereby improving the efficiency of payment urging.
[0077] In step S101 of some embodiments, the payer data refers to the user data to be urged during the payment reminder process, and the payer data includes but is not limited to user asset data (such as loan data, other bank liability data, guarantee type, etc.), overdue information, overdue days, number of user reminders, historical payment records (such as historical transaction transfer-in behavior, historical transaction transfer-out behavior, etc.) and credit reports.
[0078] Specifically, the payer data can be obtained from the bank's internal systems (such as loan information systems, bank transaction systems, bad debt disposal systems, etc.).
[0079] Specifically, the collection party data refers to the collection party that handles the collection business during the payment collection process. The collection party data includes but is not limited to the collection data that the collection party can undertake, the collection cycle, the collection complaint rate, the payer satisfaction level, etc. The collection data also includes multiple candidate collection objects, and each candidate collection corresponds to a collection company.
[0080] Specifically, the data of the urging party can be obtained from the internal institutional database of the bank.
[0081] See also Figure 2 In some embodiments, step S102 may include but is not limited to steps S201 to S208:
[0082] Step S201, obtaining an initial classification model before training, and obtaining a real classification label corresponding to the payer data.
[0083] Step S202: extract features from payer data using the initial classification model to obtain payer features.
[0084] Step S203: predicting the payer category to which the payer data belongs according to the payer characteristics to obtain a predicted payer category.
[0085] Step S204, calculating the residual between the predicted payer category and the true classification label.
[0086] Step S205, calculating the gradient estimate of the payer characteristic based on the residual.
[0087] Step S206, training the initial classification model according to the gradient estimation value to obtain a pre-trained classification model.
[0088] Step S207, making a final prediction on the payer category to which the payer data belongs according to the pre-trained classification model, and outputting the payer classification result; wherein the payer classification result is used to indicate the payment expediting risk level of the payer.
[0089] Step S208: Filter out target data from the payer data according to the payer classification result.
[0090] In step S201 of some embodiments, the initial classification model is a symmetric classification tree model constructed based on the CatBoost algorithm.
[0091] Specifically, the real classification label refers to the real payer category to which the payer data belongs. The real classification label may include the first payer category, the second payer category and the third payer category; the first payer category is used to indicate the user category with a low risk level for expediting; the second payer category is used to indicate the user category with a medium risk level for expediting; and the third payer category is used to indicate the user category with a high risk level for expediting.
[0092] Specifically, the first payer category represents payers who respond to reminders in a timely manner and make payments; the second payer category represents payers who do not respond to reminders in a timely manner or who require multiple reminders before making payments; the third payer category represents payers who have a low probability of asset recovery, have multiple reminders at the same time, or have difficulty in making reminder payments.
[0093] In step S202 of some embodiments, payer characteristics refer to characteristics of payer data represented by numerical values; payer characteristics include payer category characteristics and combined payer characteristics.
[0094] For example, the payer characteristics are discrete characteristics such as overdue information, overdue days, number of user reminders, and historical payment records. The discrete characteristics can be mapped to numerical values, and the payer characteristics can be represented by numerical values of 0, 1, 2, and 3.
[0095] See also Figure 3 Step S202 may include but is not limited to steps S301 to S304:
[0096] Step S301, using the initial classification model to perform initial feature extraction on the payer data to obtain payer category features; the payer category features are used to represent discrete payer category features.
[0097] Step S302, converting the payer category feature into a numerical category feature; the numerical category feature is used to represent a numerical payer category feature.
[0098] Step S303, identifying the association relationship between the payer category features.
[0099] Step S304: Combining the payer category features according to the association relationship to obtain a combined category feature.
[0100] In step S301 of some embodiments, the payer category feature refers to a discrete payer category feature, and the type of each feature (such as numerical, discrete, etc.) is identified through a CatBoost model, and non-numerical payer features are used as payer category features.
[0101] For example, overdue information, overdue days, number of times the user has reminded the user, and historical payment records are discrete payer category features.
[0102] In step S302 of some embodiments, the numerical category feature refers to a payer category feature expressed in a numerical type.
[0103] Specifically, the payer category feature can be converted into a numerical category feature by the following formula:
[0104]
[0105] Among them, x i,k represents the numerical category feature corresponding to the i-th payer category feature among the k numerical category features; k represents the number of numerical category features; n represents the number of payer category features; x j,k Y represents the numerical category feature corresponding to the j-th payer category feature among the k numerical category features; j represents the true numerical category feature corresponding to the j-th payer category feature; a represents the weight coefficient; p represents the prior probability of converting the payer category feature into the corresponding numerical category feature.
[0106] In step S303 of some embodiments, the association relationship may be determined based on actual business.
[0107] For example, there is an association relationship with user A between the numerical category features of 0 (indicating user A), 1 (indicating overdue information of user A), and 2 (indicating historical payment records of user A).
[0108] In step S304 of some embodiments, the combined category feature refers to a feature obtained by combining numerical category features that have an associated relationship.
[0109] For example, there is an association relationship between the numerical category features of 0 (representing user A), 1 (representing overdue information of user A) and 2 (representing historical payment records of user A), and the combined category feature after combining the numerical category features is 012.
[0110] In this embodiment, by combining the payer category features according to the association relationship, the interaction between different category features can be captured, the feature dimension of the model is enriched, and the accuracy of the subsequent model in predicting the payer category probability is improved.
[0111] In step S203 of some embodiments, the predicted payer category refers to the prediction result of the initial classification model on the payer category to which the payer data belongs. The predicted payer category still includes the first payer category, the second payer category and the third payer category; please refer to the description of the real classification label of step S201 for details, and the number of times will not be repeated.
[0112] In step S204 of some embodiments, the residual refers to the difference between the predicted payer category and the true classification label.
[0113] Specifically, for each payer feature m, the corresponding true classification label is ym, and the predicted payer category is The residual is the true classification label ym minus the predicted payer category The difference obtained.
[0114] In step S205 of some embodiments, the gradient estimate is used to characterize the loss function for predicting the payer category to which the payer data belongs.
[0115] Specifically, the gradient estimate includes the gradient value and the Hessian matrix value; the gradient value represents the first-order derivative of the loss function. Since the derivative vector of the gradient points to the direction where the loss function grows fastest, it is used to represent the sensitivity of the loss function to the predicted output; the Hessian matrix value represents the second-order derivative of the loss function and is used to calculate the curvature of the loss function, which helps to adjust the step size of the model to further balance the speed and stability of model classification during the model optimization process.
[0116] Specifically, the gradient estimates of the first m-1 payer characteristics are obtained, and a model is constructed based on the gradient estimates of the first m-1 payer characteristics and the first m-1 payer characteristics, so that for each payer characteristic, there is a separate model G m Correspondingly, and through G m Obtain an estimate of the gradient with respect to each payer characteristic.
[0117] In this embodiment, since the model does not include m payer features for training, the gradient estimate value output by the model for each payer feature is an unbiased gradient estimate value, which overcomes the subsequent prediction deviation of the model and ensures the accuracy of the model prediction.
[0118] In step S206 of some embodiments, the pre-trained classification model refers to a trained Catboost model.
[0119] Specifically, the initial classification model can be trained using the following formula:
[0120]
[0121] in, represents the predicted payer category of the mth payer feature at the tth iteration; represents the predicted payer category of the mth payer feature at the t-1th iteration; v represents the learning rate of the initial classification model; g m represents the gradient value of the mth payer feature; h m The Hessian matrix value representing the characteristics of the mth payer; represents the predicted payer category probability of the initial classification model for the mth payer feature at the tth iteration.
[0122] Furthermore, in this embodiment, when the predicted payer category of the mth payer feature at the tth iteration is obtained, the actual classification label ym is subtracted from the predicted payer category Get a new residual; optimize the loss function of the initial classification model according to the new residual, and continuously adjust the initial classification model parameters until the new residual is greater than or equal to the residual threshold, stop optimizing the initial classification model, and get a pre-trained classification model.
[0123] In step S207 of some embodiments, it should be noted that the payer classification result refers to the payer category to which the payer data belongs. Since the payer result is used to indicate the payment expediting risk level of the payer, the payer classification result still includes the first payer category, the second payer category and the third payer category.
[0124] See also Figure 4 In some embodiments, step S207 may include but is not limited to steps S401 to S402:
[0125] Step S401, traverse each leaf node of the classification tree in the classification model to obtain the category score of the payer category to which the payer data output by each leaf node belongs; the leaf node stores the payer data and the category score of the payer category to which the payer data belongs.
[0126] Step S402, making a final prediction on the payer category to which the payer data belongs according to the category score, and outputting the payer classification result.
[0127] In step S401 of some embodiments, specifically, by obtaining the gradient estimation value corresponding to each leaf node, the gradient estimation value can be used as the category score of the payer category to which the payer data belongs.
[0128] For example, the payee stored in leaf node A is user C. User C has no overdue information and three historical payment records, indicating that user C has paid three installments of reminder payments. The category score of leaf node A belonging to the first payee category is 96.6 points; the category score of leaf node A belonging to the second payee category is 0.4 points; and the category score of leaf node A belonging to the second payee category is 3 points.
[0129] In step S402 of some embodiments, a final prediction is made on the payer category to which the payer data belongs based on the category score, and a payer classification result is output.
[0130] For example, the classification model outputs that the category score of the first payer category to which the payer data belongs is 96.6 points; the category score of the second payer category to which the payer data belongs is 0.4 points; and the category score of the third payer category of the payer data is 3 points. Based on the above category scores, it is determined that the payer data belongs to the first payer category.
[0131] In this embodiment, by dividing the payer data into different categories, it is possible to configure different payment reminder strategies for payers of different categories in the subsequent payment reminder process, thereby effectively improving the efficiency of payment reminder.
[0132] In step S208 of some embodiments, the target data refers to the payee data that can automatically perform payment reminders, and the target data includes a target object; wherein the target object refers to a user who can automatically pay the reminder amount.
[0133] Specifically, the payer classification results have been determined as the first payer category, the second payer category and the third payer category.
[0134] Among them, the first payer category has a better effect on user payment urging, and the first payer category accounts for about 10% of the total payers. The customer manager can urge users to make payments by themselves, thereby saving payment urging costs; the third payer category has a low probability of user asset recovery, and there are multiple urging businesses or payers that are difficult to urge payment at the same time, and the third payer category accounts for about 10% of the total payers. The customer manager can also urge users to make payments by themselves, so as to carry out special urging for complex payment urging businesses to ensure the recovery rate of payment urging; the second payer category is for payers who do not respond to urging in time or need multiple urgings before paying, and the second payer category accounts for about 80% of the total payers. Automatic payment urging can be achieved through subsequent models, so the payer data corresponding to the second payer category is used as the target data.
[0135] In this embodiment, different reminder methods are adopted for different payer categories. While ensuring the overall payment reminder effect, the reminder cost is effectively reduced and the benefits of payment reminder are maximized. Since the second payer category accounts for a large proportion and the recovery rate of payment reminders is high, in order to ensure the efficiency of payment reminders, automatic payment reminders are performed for the second payer category. This can avoid the situation where reminder cases are not assigned in time or business personnel are unable to urge the payer, effectively improving the efficiency of payment reminders.
[0136] See also Figure 5 In some embodiments, step S103 may include but is not limited to steps S501 to S503:
[0137] Step S501, obtaining the number of clusters of the target object and obtaining the cluster distribution parameters of the target object.
[0138] Step S502: Calculate the cluster distribution probability of the target object according to the cluster quantity and cluster distribution parameters.
[0139] Step S503, determining the difficulty of urging the target object according to the cluster distribution probability.
[0140] In step S501 of some embodiments, the number of clusters refers to the number of types of data that the target data belongs to the same type.
[0141] Specifically, the number of clusters can be determined based on the target data of the actual business, where the target data includes the basic data of the target object and the asset data of the target object. The basic data of the target object includes user loss of contact information, user default information, etc.; the asset data of the target object includes overdue information within the bank (such as in-bank liabilities, in-bank registered asset information, etc.), collection and disposal information (such as user entrusted disposal), collection result information (user entrusted litigation), overdue information outside the bank (such as non-bank liabilities, user registered asset scale, etc.), and property clue information (such as mortgages, car loans, property auctions, etc.).
[0142] Specifically, the number of clusters can be determined by the following formula:
[0143]
[0144] Among them, CH represents the Karanska-Harabas index, which is used to evaluate the clustering effect of the target data. The higher the Karanska-Harabas index, the better the clustering effect. BSS represents the between-class sum of squares, that is, the sum of the squares of the distances between the center points of different categories and the center point of the target data. WSS represents the within-class sum of squares, that is, the sum of the squares of the distances between the data points within each category and the center point of the category. q represents the number of clusters. n represents the number of target objects.
[0145] Furthermore, in this embodiment, the cluster distribution parameters include a mean vector, a covariance matrix, and a mixing coefficient.
[0146] In this embodiment, by selecting the q value that maximizes the Kalanska-Harabas index as the optimal number of clusters, the optimal clustering parameters can be determined, which is convenient for subsequently improving the accuracy of cluster distribution probability calculation.
[0147] In step S502 of some embodiments, the cluster distribution probability refers to the probability that each target object belongs to each target cluster, and the cluster distribution probability can be obtained by clustering based on a Gaussian mixture clustering model. Since the Gaussian mixture clustering model is implemented based on the Gaussian mixture clustering algorithm, and the algorithm is used to describe that the same type of data belongs to the same probability distribution, different target objects can be represented by several Gaussian distributions, that is, each target object is not absolutely divided into a certain category, but the probability of each target object belonging to each target cluster is calculated.
[0148] See also Figure 6In some embodiments, step S502 includes but is not limited to steps S601 to S603:
[0149] Step S601, calculating the initial distribution probability of the target object according to the number of clusters and cluster distribution parameters.
[0150] Step S602: optimizing the cluster distribution parameters of the target object according to the initial distribution probability to obtain optimized cluster distribution parameters.
[0151] Step S603, calculating the cluster distribution probability of the target object according to the number of clusters and the optimized cluster distribution parameters; the cluster distribution probability is used to represent the maximum probability of the cluster category to which the target object belongs.
[0152] In step S601 of some embodiments, the initial distribution probability refers to the initial probability that each target object belongs to each cluster.
[0153] Specifically, the initial distribution probability can be calculated by the following formula:
[0154]
[0155] Among them, p(x) represents the initial distribution probability of the x-th target object; Q represents the number of clusters; α q Indicates the mixing coefficient corresponding to the cluster category of the qth cluster number; μ q represents the mean vector corresponding to the cluster category of the qth number of clusters; ∑q represents the covariance matrix of the cluster category of the qth number of clusters.
[0156] In this embodiment, by calculating the initial distribution probability corresponding to each target object, the initial distribution probability of each target object belonging to each Gaussian model can be determined, that is, the closer the target object is to the center of the Gaussian distribution, the greater the probability, and the target object is more likely to belong to the cluster cluster, which effectively improves the accuracy of target object clustering.
[0157] In step S602 of some embodiments, the cluster distribution parameters are optimized to represent the maximum probability of the optimal cluster category to which the target object belongs.
[0158] Specifically, the cluster distribution parameters of the target object can be optimized by using the expectation maximization algorithm.
[0159] Specifically, the optimized mean vector can be obtained by the following formula:
[0160]
[0161] in, represents the optimized mean vector; N represents the number of target objects; p(x) represents the initial distribution probability; x nRepresents the nth target object.
[0162] Specifically, the optimized covariance matrix can be obtained by the following formula:
[0163]
[0164] in, represents the optimized covariance matrix; N represents the number of target objects; p(x) represents the initial distribution probability; x n represents the nth target object; μ q Represents the mean vector corresponding to the cluster category of the qth cluster number; (x n -μ q ) T It means (x n -μ q ) is the transposed matrix of .
[0165] Specifically, the optimized mixing coefficient can be obtained by the following formula:
[0166]
[0167] in, represents the optimized mixing coefficient; N represents the number of target objects; p(x) represents the initial distribution probability; x n represents the nth target object; μ q Represents the mean vector corresponding to the cluster category of the qth cluster number.
[0168] Furthermore, the optimized cluster distribution parameters include an optimized mean vector, an optimized covariance matrix and an optimized mixing coefficient.
[0169] In this embodiment, the cluster distribution parameters of the target object are optimized by the expectation maximization algorithm, and the optimized cluster distribution parameters can be determined to maximize the cluster probability corresponding to the target object, thereby further ensuring the accuracy of the target object clustering.
[0170] In step S603 of some embodiments, the cluster distribution probability is used to represent the maximum probability of the cluster category to which the target object belongs.
[0171] Specifically, the cluster distribution probability needs to be adjusted in combination with the weight value corresponding to the initial distribution probability of each target object;
[0172] Specifically, the weight value can be determined by the following formula:
[0173]
[0174] Among them, w nrepresents the weight value corresponding to the initial distribution probability of the nth target object; N represents the number of target objects; p(x) represents the initial distribution probability.
[0175] Specifically, the cluster distribution process of the target object is optimized according to the number of clusters, the optimized cluster distribution and the weight value corresponding to the initial distribution probability of each target object. The final cluster distribution probability corresponding to each target object can be determined according to the optimized parameters, further improving the accuracy of target object clustering.
[0176] In step S503 of some embodiments, the collection difficulty may be used to describe the collection collection period and collection collection rate of the target object's payment collection.
[0177] Specifically, according to the cluster distribution probability, it can be determined that the cluster corresponding to each target object represents the difficulty of group collection for the object. For example, cluster D1 can represent the difficult collection group with medium risk and high overdue amount; cluster D2 can represent the relatively difficult collection group with medium risk and low overdue amount.
[0178] Specifically, the difficulty of urging matters may be evaluated by analyzing the average or median of the cluster.
[0179] For example, the average overdue days and average overdue amount of target object h1 are obtained as 60 and 5000 yuan, and the average overdue days and average overdue amount of target object h1 are compared with the preset urging conditions, where the urging conditions include the preset overdue threshold of 30 days and the preset overdue amount greater than 2000 yuan. Since the average overdue days and average overdue amount of target object h1 are greater than the preset overdue threshold and the preset overdue amount, it means that the urging difficulty is difficult.
[0180] See for example Figure 7 In some embodiments, step S104 may include but is not limited to steps S701 to S704:
[0181] Step S701, scoring the expediency capabilities of candidate expediency objects according to preset expediency capability indicators to obtain initial capability scores.
[0182] Step S702, obtaining the indicator weight of the expediting capability indicator.
[0183] Step S703, performing a weighted operation on the initial capability score according to the indicator weight to obtain a target capability score.
[0184] Step S704: Select a target urging object from the candidate urging objects according to the urging difficulty and the target ability score.
[0185] In step S701 of some embodiments, the expediting capability indicator is used to represent the expediting capability of the candidate expediting object, including but not limited to expediting cycle, expediting success rate, payer contact rate, expediting cost, expediting risk and payer satisfaction.
[0186] Specifically, the initial capability score is used to characterize the expediting capability of the candidate expediting object.
[0187] Specifically, the initial capability score can be used to comprehensively evaluate the candidate urging objects from five dimensions, including historical recovery performance, discovery of property clues, repair of lost contact, post-score of the urging manager / team, and urging timeliness, based on capability indicators such as the candidate urging objects' urging cycle, urging success rate, payer contact rate, urging cost, urging risk, and payer satisfaction, and output the candidate urging objects' capability scores.
[0188] For example, the historical recovery performance of candidate expediting object E is 85 points, the discovery of property clues is 80 points, and the repair of lost contact is 90 points. The post-expediting manager / team score is 90 points, and the expediting time limit is 85 points. The average score of the five dimensions is 86 points. The corresponding initial ability score of candidate expediting object E is 86 points.
[0189] In step S702 of some embodiments, the indicator weight refers to the importance of each expediency indicator to the overall expediency capability score when evaluating the expediency capability of the candidate expediency object.
[0190] Specifically, the indicator weights can be adjusted based on actual business or combined with expert advice to determine the weights of each indicator from five dimensions: historical recovery performance indicators of candidate follow-up objects, discovery of property clues, repair of lost contact, post-scoring indicators of follow-up managers / teams, and follow-up timeliness.
[0191] Among them, historical recovery performance indicators (including historical collection costs, historical collection success rates, etc.), discovery of property clues (including overdue amounts of payers, payer assets, etc.), repair of lost contact (including payer contact rate), collection manager / team post-scoring indicators (including collection risks, payer satisfaction, etc.) and collection timeliness (including historical collection cycles, etc.).
[0192] For example, the weight of the historical recovery performance indicator is 25%; the weight of the property clue discovery indicator is 15%; the weight of the lost contact repair indicator is 20%; the weight of the follow-up manager / team post-scoring indicator is 20%; and the weight of the follow-up timeliness indicator is 20%.
[0193] In step S703 of some embodiments, the target ability score is the target ability score obtained by comprehensive weighted accumulation of the initial ability scores.
[0194] For example, the historical recovery performance of candidate expediting object E is 85 points, the discovery of property clues is 80 points, and the repair of lost contact is 90 points. The post-expediting manager / team score is 90 points, and the expediting time limit is 85 points. Then the target ability score of candidate expediting object E is (85x25%)+(80x15%)+(90x20%)+(90x20%)+(85x20%)=86.25 points.
[0195] In this embodiment, the indicator weights are determined in combination with expert knowledge to determine the target capability score, so as to improve the efficiency of finding the target urging object in the future.
[0196] In step S704 of some embodiments, by dividing the target objects into different clusters, and each type of cluster corresponds to a different candidate urging object, the initial candidate urging objects with high similarity corresponding to each type of target object can be preliminarily screened out, and then the target urging object can be screened out from the initial candidate urging objects.
[0197] Specifically, by obtaining the urging difficulty score of the target object, the difference between the target ability score and the urging difficulty score is calculated, and the target urging object is detected based on the difference and a preset score threshold.
[0198] For example, the difficulty score of target object H is 78 points; the target ability score of the corresponding initial candidate object is 86.25 points; then the difference between the target ability score and the difficulty score is 8.25 points, and the score threshold is 5 points, indicating that the candidate object can be used as the target object.
[0199] Furthermore, in this embodiment, if there are multiple initial candidate urging objects whose target capability scores and urging difficulty scores have differences greater than the score threshold, the initial candidate urging object with the largest difference is selected as the target urging object.
[0200] In this embodiment, the target urging object is selected from the candidate urging objects according to the urging difficulty and the target ability score. The urging difficulty and the integration of expert knowledge can realize the precise sorting of the candidate urging objects, which not only improves the accuracy of the target urging object selection, but also realizes the automatic matching of the corresponding urging objects for the payer to urge the payment on the premise of meeting the precise urging requirements, thereby improving the efficiency of payment urging.
[0201] In step S105 of some embodiments, for example, for target objects with a short overdue period, stable income and good historical repayment record, reminders can be made through the target collection object by phone or text message; for target objects with a long overdue period, unstable income and poor historical repayment record, collection can be carried out through a combination of telephone collection and door-to-door visits by the target collection object. If the collection is fruitless, further legal means will be considered; for target customers with a large overdue amount but good willingness and ability to repay, the target collection object can negotiate a repayment plan with the target object, which may include payment methods such as installments or deferred repayment.
[0202] In this embodiment, different target reminder objects are matched with different target objects, so that payment reminders can be implemented in a targeted manner, which not only ensures the payment reminder effect, but also saves reminder resources, thereby improving the efficiency of payment reminders.
[0203] The embodiment of the present application obtains payer data and urging party data; wherein the urging party data includes candidate urging objects; screens out target data from the payer data according to a pre-trained classification model; wherein the target data includes target objects; clusters the target objects to obtain the difficulty of urging the target objects; selects the target urging objects from the candidate urging objects according to the difficulty of urging; and urges the target objects through the target urging objects. The present application can classify payers according to the difficulty of urging by screening and clustering payers, so as to facilitate screening out corresponding urging parties for different payers, and can match corresponding urging parties for urging payers with different urging difficulties, effectively avoiding the situation where manual urging is not timely or the business personnel are unable to urge the payers during the urging process, thereby improving the efficiency of payment urging.
[0204] See also Figure 8 The embodiment of the present application also provides a payment urging device, which can implement the above-mentioned payment urging method, and the device includes:
[0205] A payment urging data acquisition module is used to acquire payer data and urging party data; wherein the urging party data includes candidate urging objects;
[0206] A payer screening module, used to screen target data from the payer data according to a pre-trained classification model; wherein the target data includes a target object;
[0207] A payer clustering module, used for clustering the target objects to obtain the difficulty of urging the target objects;
[0208] An expedited object selection module is used to select a target expedited object from the candidate expedited objects according to the expedited difficulty;
[0209] The payment urging module is used to urge the target object through the target urging object.
[0210] The specific implementation of the payment urging device is substantially the same as the specific implementation of the payment urging method described above, and will not be described in detail herein.
[0211] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the payment urging method when executing the computer program. The electronic device can be any smart terminal including a tablet computer, a car computer, etc.
[0212] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:
[0213] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;
[0214] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the payment urging method of the embodiment of this application;
[0215] Input / output interface 903, used to implement information input and output;
[0216] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);
[0217] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);
[0218] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .
[0219] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned payment urging method is implemented.
[0220] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed 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.
[0221] The payment urging method, payment urging device, electronic device and storage medium provided in the embodiments of the present application obtain payer data and urging party data; wherein the urging party data includes candidate urging objects; target data is screened out from the payer data according to a pre-trained classification model; wherein the target data includes a target object; the target object is clustered to obtain the difficulty of urging the target object; the target urging object is selected from the candidate urging objects according to the difficulty of urging; and the target object is urged through the target urging object. The present application can classify payers according to the difficulty of urging by screening and clustering payers, so that corresponding urging parties can be screened out for different payers, and corresponding urging parties can be matched for payers with different urging difficulties for urging, which effectively avoids the situation where manual urging is not timely or business personnel are unable to urge the payer during the urging process, thereby improving the efficiency of payment urging.
[0222] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.
[0223] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.
[0224] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0225] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.
[0226] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device 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 devices.
[0227] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.
[0228] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0229] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0230] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0231] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.
[0232] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.
Claims
1. A payment urging method, characterized in that: The method comprises: Obtaining payer data and expediter data; wherein the expediter data includes candidate expediting objects; Filtering target data from the payer data according to a pre-trained classification model; wherein the target data includes a target object; Performing clustering processing on the target objects to obtain the difficulty of urging the target objects; Selecting a target urging object from the candidate urging objects according to the urging difficulty; The target object is urged to be handled through the target urging object.
2. The method according to claim 1, characterized in that The step of filtering out target data from the payer data according to the pre-trained classification model includes: Obtaining an initial classification model before training, and obtaining a true classification label corresponding to the payer data; Extracting features from the payer data using the initial classification model to obtain payer features; Predicting the payer category to which the payer data belongs according to the payer characteristics to obtain a predicted payer category; Calculating the residual between the predicted payer category and the true classification label; Calculating a gradient estimate of the payer characteristic based on the residual; Training the initial classification model according to the gradient estimation value to obtain the pre-trained classification model; Performing a final prediction on the payer category to which the payer data belongs according to the pre-trained classification model, and outputting a payer classification result; wherein the payer classification result is used to indicate the payment expediting risk level of the payer; The target data is screened out from the payer data according to the payer classification result.
3. The method according to claim 2, characterized in that The payer characteristics include payer category characteristics and combination category characteristics; The extracting features of the payer data using the initial classification model to obtain payer features includes: Performing initial feature extraction on the payer data using the initial classification model to obtain payer category features; the payer category features are used to represent discrete payer category features; Converting the payer category feature into a numerical category feature; the numerical category feature is used to represent a numerical payer category feature; Identifying associations between the payer category characteristics; The payer category features are combined according to the association relationship to obtain a combined category feature.
4. The method according to claim 2, characterized in that: The finally predicting the payer category to which the payer data belongs according to the pre-trained classification model and outputting the payer classification result includes: Traversing each leaf node of the classification tree in the classification model to obtain a category score of the payer category to which the payer data output by each leaf node belongs; the leaf node stores the payer data and the category score of the payer category to which the payer data belongs; A final prediction is made on the payer category to which the payer data belongs based on the category score, and the payer classification result is output.
5. The method according to claim 1, characterized in that The clustering process of the target objects to obtain the difficulty of urging the target objects includes: Obtaining the number of clusters of the target object and obtaining cluster distribution parameters of the target object; Calculating the cluster distribution probability of the target object according to the cluster quantity and the cluster distribution parameter; The difficulty of urging the target object is determined according to the cluster distribution probability.
6. The method according to claim 5, characterized in that The calculating the cluster distribution probability of the target object according to the cluster quantity and the cluster distribution parameter includes: Calculating the initial distribution probability of the target object according to the number of clusters and the cluster distribution parameters; Optimizing the cluster distribution parameters of the target object according to the initial distribution probability to obtain optimized cluster distribution parameters; The cluster distribution of the target object is calculated according to the cluster quantity and the optimized cluster distribution parameters; the cluster distribution probability is used to represent the maximum probability of the cluster category to which the target object belongs.
7. The method according to any one of claims 1 to 6, characterized in that: The step of selecting a target urging object from the candidate urging objects according to the urging difficulty includes: Scoring the candidate expediting objects for expediting ability according to the preset expediting ability index to obtain an initial ability score; Obtaining the indicator weight of the expediting capability indicator; Performing a weighted operation on the initial capability score according to the indicator weight to obtain a target capability score; The target urging object is selected from the candidate urging objects according to the urging difficulty and the target ability score.
8. A payment urging device, characterized in that: The device comprises: A payment urging data acquisition module is used to acquire payer data and urging party data; wherein the urging party data includes candidate urging objects; A payer screening module, used to screen target data from the payer data according to a pre-trained classification model; wherein the target data includes a target object; A payer clustering module, used for clustering the target objects to obtain the difficulty of urging the target objects; An expedited object selection module is used to select a target expedited object from the candidate expedited objects according to the expedited difficulty; The payment urging module is used to urge the target object through the target urging object.
9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the payment urging method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the payment urging method according to any one of claims 1 to 7 is implemented.
Citation Information
Patent Citations
Bill overdue intelligent collection method and system
CN111383093A
Loan collection scheme pushing method and device
CN112967131A
Multi-dimensional case allocation method based on credit card post-loan collection service
CN113642908A
Classification method and device for electricity users and computer equipment
CN115712689A
Power grid company supervision method and system for improving full-service process efficiency
CN117408431A
