Business processing method, apparatus, program product, computer device, and medium

Abstract

The application discloses a business processing method, device, program product, computer equipment and medium, and relates to the technical field of business processing. The method comprises the following steps: acquiring an object set containing N business objects; the object set corresponds to a processing deadline; when the processing deadline is reached, if there is an unprocessed business object in the object set, the unprocessed business object is determined as overdue; acquiring M predicted business objects from the object set; the M predicted business objects are business objects that are not processed at a target time before the processing deadline is reached; acquiring an overdue probability distribution of the M predicted business objects; calibrating the overdue probability distribution to obtain a calibrated overdue probability distribution of the M predicted business objects; the calibrated overdue probability distribution contains calibrated overdue probabilities corresponding to the M predicted business objects respectively; and predicting overdue data of the object set based on the calibrated overdue probability distribution of the M predicted business objects. By using the application, the accuracy of the predicted overdue data can be improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more particularly to a business processing method, apparatus, program product, computer equipment, and medium. Background Technology

[0002] For some new "buy now, pay later" e-commerce models, users can place an order and settle the payment for goods using a password-free deduction method, and then settle the actual payment for the goods within a certain period. If the user does not settle the actual payment for the goods within the period (i.e., overdue settlement), the order to which the goods belong is considered a bad order.

[0003] In existing applications, the data on bad orders generated at different times in history is usually used to predict the number of bad orders in a product order created at a later time. However, when there is no pattern in the bad order data generated at different times in history, the existing application method cannot accurately predict the number of bad orders that will be generated at a later time. Summary of the Invention

[0004] This application provides a business processing method, apparatus, program product, computer equipment, and medium that can improve the accuracy of predicted overdue data.

[0005] This application provides a business processing method, which includes:

[0006] Get the object set; the object set contains N business objects, and the object set corresponds to the processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer;

[0007] Obtain M predicted business objects from the object set; M predicted business objects refer to business objects that have not been processed by the target time before the processing deadline; M is a positive integer and M is less than or equal to N;

[0008] Obtain the delinquency probability distribution of M predicted business objects; the delinquency probability distribution is formed based on the predicted delinquency probability corresponding to each of the M predicted business objects.

[0009] The overdue probability distribution is calibrated to obtain the calibrated overdue probability distribution of M predicted business objects; the calibrated overdue probability distribution contains the calibrated overdue probability corresponding to each of the M predicted business objects.

[0010] Overdue data of the prediction object set based on the calibration overdue probability distribution of M prediction business objects.

[0011] This application provides a business processing apparatus, which includes:

[0012] The first acquisition module is used to acquire an object set; the object set contains N business objects, and the object set corresponds to a processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer;

[0013] The second acquisition module is used to acquire M predicted business objects from the object set; the M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline; M is a positive integer and M is less than or equal to N.

[0014] The distribution acquisition module is used to obtain the overdue probability distribution of M predicted business objects; the overdue probability distribution is formed based on the predicted overdue probability corresponding to each of the M predicted business objects.

[0015] The distribution calibration module is used to calibrate the overdue probability distribution to obtain the calibrated overdue probability distribution of M predicted business objects; the calibrated overdue probability distribution contains the calibrated overdue probabilities corresponding to the M predicted business objects respectively;

[0016] The prediction module is used to predict overdue data of an object set based on the calibrated overdue probability distribution of M prediction business objects.

[0017] Optionally, the distribution acquisition module may obtain the overdue probability distribution of the M predicted business objects in the following ways:

[0018] Obtain the object features of each predicted business object and input the object features of each predicted business object into the prediction model;

[0019] The prediction model is invoked to determine the predicted delinquency probability for each predicted business object based on its object characteristics.

[0020] The overdue probability distribution is determined based on the predicted overdue probability corresponding to each predicted business object.

[0021] Optionally, any one of the M predicted business objects can be represented as the target business object; the distribution acquisition module obtains the object features of each predicted business object in the following ways:

[0022] Obtain multiple object association information for the target business object;

[0023] Generate information features corresponding to the object association information of each target business object;

[0024] The information features corresponding to each object's associated information are concatenated to obtain the object features of the target business object.

[0025] Optionally, the above-mentioned device is also used for:

[0026] Obtain the sample object set; the sample object set contains multiple sample business objects, each of which carries a corresponding object tag. The object tag carried by any sample business object is either an overdue tag or a non-overdue tag.

[0027] Obtain the sample object characteristics of each sample business object in the sample object set;

[0028] Obtain the initial prediction model and call the initial prediction model to predict the sample prediction delinquency probability of each sample business object based on the sample object characteristics of each sample business object;

[0029] The model parameters of the initial prediction model are updated based on the sample prediction overdue probability of each sample business object and the object label carried by each sample business object, and the initial prediction model after the model parameter update is determined as the prediction model.

[0030] Optionally, the distribution calibration module calibrates the overdue probability distribution to obtain the calibrated overdue probability distribution of M predicted business objects, including:

[0031] Input the overdue probability distribution into the calibration model;

[0032] The calibration model is invoked to calibrate the overdue probability distribution, resulting in a calibrated overdue probability distribution.

[0033] Optionally, the overdue probability distribution is obtained based on the prediction model; the prediction model is obtained by training an initial prediction model, which is trained based on a sample object set. The sample object set contains multiple sample business objects, and each sample business object carries a corresponding object label, which is an overdue label or a non-overdue label.

[0034] The above-mentioned device is also used for:

[0035] Obtain the predicted overdue probability of each sample business object predicted by the initial prediction model during the training process of the initial prediction model;

[0036] Obtain the initial calibration model and call the initial calibration model to calibrate the probability distribution formed by the predicted overdue probabilities of multiple sample business objects based on the object tags carried by each sample business object;

[0037] The initial calibration model, after updating the model parameters during the calibration process of the probability distribution formed by the predicted overdue probabilities of multiple sample business objects, is determined as the calibration model.

[0038] Optionally, the prediction module can predict overdue data of the object set based on the calibrated overdue probability distribution of M prediction business objects, including:

[0039] Calculate the expected overdue value of M predicted business objects based on the calibration overdue probability distribution, and determine the expected overdue value as the overdue amount of the predicted object set;

[0040] The overdue data of the object set is determined based on the overdue amount of the objects.

[0041] Optionally, the prediction module may determine the overdue data of the object set based on the amount of overdue objects in the following ways:

[0042] Among N business objects, the business objects that were revoked before the target time are identified as revoked business objects;

[0043] The object delinquency rate of the object set is determined based on the first number of N business objects, the second number of revoked business objects, and the object delinquency rate.

[0044] Use the object delinquency rate as delinquency data.

[0045] Optionally, the prediction module determines the object delinquency rate of the object set based on a first number of N business objects, a second number of revoked business objects, and the object delinquency rate, including:

[0046] Get the third number of N business objects that were cancelled within the time interval formed by the target time and the processing deadline;

[0047] The sum of the second and third quantities is determined as the object cancellation amount;

[0048] The difference between the first quantity and the number of object cancellations is determined as the total number of reference objects;

[0049] The ratio between the number of overdue objects and the total number of reference objects is determined as the object overdue rate.

[0050] Optionally, the object set is created within the target time period; the prediction module obtains a third number of business objects that are revoked within the time interval formed by the target time and the processing deadline from among N business objects, including:

[0051] Retrieve multiple historical time periods preceding the target time period; each historical time period corresponds to a set of historical objects that have been created.

[0052] Get the number of business objects that were revoked within the historical time interval indicated by the time interval corresponding to each historical period;

[0053] The average number of business objects that were revoked within each historical time interval is determined as the third quantity.

[0054] Optionally, the overdue data includes the object overdue rate of the object set; the above apparatus is also used for:

[0055] Obtain the relationship between the predicted object delinquency rate and the delinquency rate threshold;

[0056] If the delinquency rate of an object exceeds the delinquency rate threshold, risk control measures will be applied to the object set.

[0057] Optionally, the N business objects included in the object set refer to the N business orders created within the same natural day, and the processing of any business order refers to the resource transfer processing for that business order; the M predicted business objects are the M business orders among the N business orders that have not yet been processed before the target time.

[0058] If the delinquency rate of the objects exceeds the delinquency rate threshold, the above-mentioned device will perform risk control processing on the object set in the following ways:

[0059] Send order processing notification messages to the business devices corresponding to each of the M business orders.

[0060] This application provides a computer device, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the method of this application.

[0061] This application provides a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method described above.

[0062] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various alternative embodiments described above.

[0063] This application can obtain an object set; the object set contains N business objects, and the object set corresponds to a processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer; M predicted business objects are obtained from the object set; the M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline is reached; M is a positive integer and M is less than or equal to N; the overdue probability distribution of the M predicted business objects is obtained; the overdue probability distribution is formed based on the predicted overdue probabilities corresponding to the M predicted business objects respectively; the overdue probability distribution is calibrated to obtain the calibrated overdue probability distribution of the M predicted business objects; the calibrated overdue probability distribution contains the calibrated overdue probabilities corresponding to the M predicted business objects respectively; the overdue data of the object set is predicted based on the calibrated overdue probability distribution of the M predicted business objects. Therefore, the method proposed in this application can calibrate the overdue probability distribution formed by the predicted overdue probability of each business object in the object set. Subsequently, the overdue data of the object set can be predicted by the probability distribution formed by the calibrated overdue probabilities of each business object, thus improving the accuracy of the predicted overdue data of the object set. Furthermore, since the prediction can be made without relying on regular historical overdue data, it also expands the applicability of predicting overdue data of the object set. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application;

[0066] Figure 2 This is a schematic diagram illustrating a scenario for predicting overdue data provided in this application;

[0067] Figure 3 This is a flowchart illustrating a business processing method provided in this application;

[0068] Figure 4 This is a schematic diagram illustrating a data change scenario provided in this application;

[0069] Figure 5 This is a schematic diagram of a data prediction scenario provided in this application;

[0070] Figure 6 This is a schematic diagram of another data prediction scenario provided in this application;

[0071] Figure 7 This is a flowchart illustrating a prediction model training method provided in this application;

[0072] Figure 8 This is a flowchart illustrating a calibration model training method provided in this application;

[0073] Figure 9 This is a schematic diagram of a model training scenario provided in this application;

[0074] Figure 10 This is a schematic diagram of a model training process provided in this application;

[0075] Figures 11a-11b This is a schematic diagram of order data provided in this application;

[0076] Figure 12 This is a schematic diagram of the structure of a business processing device provided in this application;

[0077] Figure 13 This is a schematic diagram of the structure of a computer device provided in this application. Detailed Implementation

[0078] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0079] This application relates to technologies related to artificial intelligence (AI). AI is the theory, methods, technology, and application systems that utilize digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce new intelligent machines that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities.

[0080] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0081] This application primarily concerns machine learning within artificial intelligence. Machine learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory, among others. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0082] The machine learning discussed in this application mainly refers to how to train and apply predictive models and how to calibrate models. For details, please refer to the following: Figure 3 The description in the corresponding embodiment.

[0083] This application also relates to blockchain-related technologies. Blockchain is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and cryptographic algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include a blockchain underlying platform, a platform product service layer, and an application service layer. A blockchain consists of a series of blocks sequentially generated in chronological order. Once a new block is added to the blockchain, it cannot be removed. Each block records the data submitted by nodes in the blockchain system. In this application, predicted overdue data can be uploaded to the blockchain to ensure its immutability.

[0084] First, it should be noted that this application may display a prompt interface or pop-up window before and during the collection of user-related data (such as business data such as the business objects created by the user and whether the business objects have been processed). This prompt interface or pop-up window is used to inform the user that their relevant data is being collected. This application will only begin the steps of collecting user-related data after receiving confirmation from the user regarding the prompt interface or pop-up window; otherwise (i.e., without receiving confirmation from the user), the steps of collecting user-related data will end, meaning no user-related data will be collected. In other words, all user data collected in this application is collected with the user's consent and authorization, and the collection, use, and processing of relevant user data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0085] Please see Figure 1 , Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application. Figure 1 As shown, the network architecture may include server 200 and a cluster of terminal devices. The cluster of terminal devices may include one or more terminal devices; the number of terminal devices is not limited here. Figure 1 As shown, the multiple terminal devices may specifically include terminal device 100a, terminal device 101a, terminal device 102a, ..., terminal device 103a; as Figure 1 As shown, terminal devices 100a, 101a, 102a, ..., 103a can all connect to server 200 via the network, so that each terminal device can interact with server 200 via the network.

[0086] like Figure 1 The server 200 shown can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Terminal devices can be smart terminals such as smartphones, tablets, laptops, desktop computers, and smart TVs.

[0087] Please participate as well. Figure 2 , Figure 2This is a schematic diagram illustrating a scenario for predicting overdue data provided in this application. First, it should be noted that the aforementioned terminal devices (such as terminal devices 100a, 101a, 102a, ..., 103a) can be terminal devices held by individual users. Each terminal device can include an application capable of creating business objects, such as a shopping application. Therefore, the created business object can be an order for goods. Thus, each user can create a business object (such as placing an order) through their terminal device. The created business object can initially be used for payment without a password, and the actual payment can be made later (e.g., before the processing deadline). After each terminal device creates a business object, the server 200 can also obtain the business objects created by each terminal device. Therefore, the server 200 can predict the overdue data of the created business objects before the processing deadline, as described in the following description.

[0088] like Figure 2 As shown, a set of objects can be created within a target time period (such as any day). This set of objects contains several business objects created within the target time period. If these business objects are not processed before the processing deadline (such as an order not being actually paid), they are considered to be overdue. Therefore, it can be understood that the processing deadline is the absolute time limit for processing business objects (such as within 30 days of creating the business object).

[0089] Server 200 can predict the overdue data of the object set in advance at a target time before the processing deadline. At this time, server 200 can take the business objects in the object set that have not been processed (such as those that have not been revoked or paid) before the target time as the predicted business objects (as shown in box 101b). Then, server 200 can call the prediction model to predict the overdue probability of each predicted business object (which can be called the predicted overdue probability, as shown in box 102b). The predicted overdue probability indicates the probability that the predicted business object will be overdue.

[0090] Furthermore, server 200 can call the calibration model to calibrate the predicted delinquency probability of each predicted business object, thereby obtaining the calibrated delinquency probability of the predicted business object (as shown in box 103b). Through the calibrated delinquency probability of each predicted business object, the delinquency data of the object set can be obtained (as shown in box 104b). This delinquency data can be the number of business objects in the predicted object set that will delinquent (which can be called the object delinquency amount) or the delinquency rate (which can be called the object delinquency rate). The specific process of predicting the delinquency data of the object set can be further explained in the following... Figure 3 The relevant descriptions in the corresponding embodiments.

[0091] Using the method provided in this application, the overdue probability of each business object can be predicted (e.g., the overdue probability can be predicted), and then the overdue probability of each business object can be calibrated to be closer to the true overdue probability (e.g., the overdue probability can be calibrated). Then, the overdue data of the object set can be accurately predicted by the overdue probability of each business object.

[0092] Please see Figure 3 , Figure 3 This is a flowchart illustrating a business processing method provided in this application. The execution entity in this application embodiment can be a single computer device or a cluster of multiple computer devices. The computer device can be a server or a terminal device. Therefore, the execution entity in this application embodiment can be a server, a terminal device, or a combination of both. Here, we will use a server as an example to illustrate the process. Figure 3 As shown, the method may include:

[0093] Step S101: Obtain the object set; the object set contains N business objects, and the object set corresponds to the processing deadline; when the processing deadline is reached, if there are business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer.

[0094] Optionally, the server can obtain an object set containing N business objects, where N is a positive integer. The specific value of N is determined based on the actual application scenario and is not limited thereto. All business objects in this object set correspond to the same processing deadline. If there are still business objects that have not been processed by the processing deadline, they can be considered overdue. That is, business objects in the object set that have not been processed before the processing deadline are determined to be overdue.

[0095] Optionally, this application can be applied to any business scenario that requires the creation of a business object first, and the business object needs to be processed within a specified period after its creation (such as before the processing deadline). Specifically, it can be applied to the prediction of the overdue amount or overdue rate of overdue business objects in such a business scenario.

[0096] Optionally, this application can be applied to product ordering scenarios. In this business scenario, the N business objects contained in the object set can be N orders (which can be referred to as business orders) of products created on the same natural day (which can be any day). Processing business orders can refer to resource transfer processing or cancellation processing for business orders. Resource transfer processing can refer to settlement or payment for business orders (e.g., the transferred resources can be RMB), and cancellation processing can refer to canceling the created business orders (i.e., canceling the order). In this case, the processing deadline for the object set is the deadline for resource transfer processing (e.g., settlement, i.e., payment) of the orders. For example, if the business orders in the object set were created on June 1st, the processing deadline is within 30 days from the date the business orders were created (i.e., the order was placed), so the processing deadline could be the end date of June 30th. Therefore, if a business order in the object set is not processed before the processing deadline, it indicates that the business order is overdue (i.e., overdue payment), and overdue payment business orders can be considered bad debts. Therefore, this application can be applied to the prediction of bad debt rate or bad debt amount in product ordering scenarios.

[0097] Optionally, this application can also be applied to lending scenarios. In this business scenario, the N business objects contained in the object set can be N loan orders created on the same natural day. Processing the loan orders can refer to the resource transfer processing of the loan orders, which can be the repayment processing of the loan orders. In this case, the processing deadline for the object set is the deadline for the resource transfer processing (i.e., repayment) of the loan orders. For example, if the loan orders in the object set were created on August 1, the processing deadline is 30 days from the date the loan order (i.e., the loan) was created. Therefore, the processing deadline could be the end date of August 30. Thus, if a loan order in the object set is not processed before the processing deadline, it indicates that the loan order is overdue (i.e., overdue repayment). Overdue loan orders can also be considered bad debts. Therefore, this application can also be applied to predict the overdue amount or overdue rate of repayments in lending scenarios.

[0098] Step S102: Obtain M predicted business objects from the object set; M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline; M is a positive integer and M is less than or equal to N.

[0099] Optionally, the server can obtain M prediction business objects from the object set, where M is a positive integer, less than or equal to N, and the specific value of M is determined according to the actual application scenario. The M prediction business objects belong to N business objects.

[0100] Specifically, business objects that have not been processed by the target time before the processing deadline among N business objects can be considered as predicted business objects. As the name suggests, predicted business objects are those that need to be predicted. The target time is later than the time when the object set was created, but earlier than the processing deadline. The target time can be understood as the prediction time. At the target time, there are actually no overdue business objects in the object set. By predicting the overdue data of potentially overdue business objects in the object set in advance at the target time, risk control-related processing can be carried out in advance for the object set, that is, preparatory work (i.e., response work).

[0101] For example, if the object set is created on September 1st and the processing deadline is September 30th, then the target time can be the completion time on September 5th (or another time). In this case, the goal of predicting the overdue data of the object set 25 days in advance can be achieved.

[0102] If a business object can be revoked (e.g., order cancellation), then M predicted business objects refer to the M business orders among the N business objects that have not been settled or paid for or have not been revoked before the target time.

[0103] Please see Figure 4 , Figure 4 This is a schematic diagram illustrating a data change scenario provided in this application. For example... Figure 4 As shown, the object set can be a collection of business objects created on day x. A business object can be an order. Therefore, the object set can contain N orders.

[0104] The target time could be the end time of day x+5 (i.e., the 5th day after the creation of the object set). Therefore, by the end time of day x+5, the N orders would consist of three parts: orders that were not paid or cancelled before the 5th day, orders that were paid before the 5th day, and orders that were cancelled before the 5th day. The first part of these three parts, "orders that were not paid or cancelled before the 5th day," can be used as the aforementioned M predicted business objects.

[0105] The processing deadline for the object set can be the completion time of day x+30 (i.e., 30 days after the object set is created). Therefore, at the completion time of day x+30, the first part at the completion time of day x+5 can be further subdivided into three parts. These three parts can include bad debt orders (i.e., overdue orders, i.e., orders not processed before the processing deadline) between day 5 and day 30 (i.e., day x+5 to day x-30), orders successfully paid within day 5 to day 30, and orders cancelled within day 5 to day 30. Therefore, it can be understood that predicting the above M predicted business objects is equivalent to predicting the number or bad debt rate of bad debt orders between day 5 and day 30.

[0106] Step S103: Obtain the overdue probability distribution of M predicted business objects; the overdue probability distribution is formed based on the predicted overdue probability corresponding to each of the M predicted business objects.

[0107] Optionally, the server can obtain a prediction model, which can be a pre-trained model used to predict the delinquency probability of business objects. This prediction model can be an XGB model (a tree model) or other models (such as a deep neural network model). The training process of this prediction model can be found below. Figure 7 The relevant descriptions in the corresponding embodiments.

[0108] The server can obtain the object features of each of the M predicted business objects and input the object features of each predicted business object into the prediction model. Then, by calling the prediction model, the probability of each predicted business object being overdue can be predicted based on the object features of each predicted business object. This probability can be called the predicted overdue probability. Each predicted business object has a corresponding predicted overdue probability, which indicates the probability that the predicted business object will be overdue as predicted by the prediction model.

[0109] Then, by using the predicted overdue probability corresponding to each predicted business object, an overdue probability distribution of M predicted business objects can be formed. This overdue probability distribution can be a Bernoulli distribution, which contains the predicted overdue probability corresponding to each predicted business object.

[0110] In this context, any one of the M predicted business objects can be represented as the target business object. Since the principle of obtaining the object characteristics of each predicted business object is the same, the following description will be used as an example to illustrate the process of obtaining the object characteristics of the target business object.

[0111] The server can obtain multiple object association information for the target business object. This multiple object association information can refer to several pieces of information related to predicting whether the target business object is overdue. For example, if the target business object is a business order, the multiple object association information for the target business object can include information such as the order placement time, the user information of the user who placed the order, the product information of the ordered product, and the order amount.

[0112] Therefore, the server can obtain the information features (which can be embedded features) corresponding to each object association information by embedding multiple object association information of the target business object. One object association information can correspond to one information feature, which can be a feature vector.

[0113] Furthermore, the server can perform feature concatenation processing on the information features corresponding to the association information of each object to obtain the object features of the target business object.

[0114] Step S104: The overdue probability distribution is calibrated to obtain the calibrated overdue probability distribution of M predicted business objects; the calibrated overdue probability distribution includes the calibrated overdue probability corresponding to each of the M predicted business objects.

[0115] Optionally, the server can obtain a calibration model, which can be a pre-trained model capable of calibrating the overdue probability distribution. This calibration model can be a Platt model (a type of calibration model) or other models. The training process of this calibration model can be found below. Figure 8 The relevant descriptions in the corresponding embodiments.

[0116] The server can input the obtained overdue probability distribution into the calibration model. For example, the predicted overdue probability of each predicted business object can be input into the calibration model. Then, the server can call the calibration model to calibrate the overdue probability distribution and obtain the calibrated overdue probability distribution, which can be called the calibrated overdue probability distribution. The calibrated overdue probability distribution can also be a Bernoulli distribution.

[0117] The calibration overdue probability distribution includes the overdue probability after calibrating the predicted overdue probability of each predicted business object. This overdue probability can be called the calibration overdue probability. Each predicted business object corresponds to a calibration overdue probability. The calibration overdue probability of a predicted business object indicates the probability that the predicted business object will be overdue after calibration.

[0118] Step S105: Based on the calibration overdue probability distribution of M predicted business objects, predict the overdue data of the object set.

[0119] Optionally, the server can predict the overdue data of the object set by using the calibrated overdue probability distribution of M prediction business objects. Please refer to the following description:

[0120] The server can calculate the expected overdue value of M predicted business objects based on the calibration overdue probability distribution. This expected overdue value can be equal to the sum of the calibration overdue probabilities corresponding to the M predicted business objects in the calibration overdue probability distribution. This expected overdue value is also the expected overdue value of the M business objects.

[0121] Therefore, the server can use this expected overdue value as the predicted overdue amount of the object set, which can be called the object overdue amount. This object overdue amount is the number of business objects that may be overdue in the object set predicted through the above process. The object overdue amount N1 is as shown in the following formula (1):

[0122]

[0123] Where N1 represents the predicted amount of overdue payments, p i Let i represent the predicted overdue probability of the i-th predicted business object among the M predicted business objects in the object set. The value of i ranges from 1 to M. In formula (1), "1" means the probability of overdue result is 1.

[0124] Please see Figure 5 , Figure 5 This is a schematic diagram illustrating a data prediction scenario provided in this application. For example... Figure 5 As shown, the object set can be a collection of business objects created within the target time period. The server can obtain M business objects participating in the prediction from the object set. These M business objects can be called M predicted business objects. These M predicted business objects are business objects in the object set that have not been processed (such as those that have not undergone resource transfer or cancellation processing) before the target time.

[0125] The server can use a joint prediction model and a calibration model to predict the number of business objects (i.e., object overdue amount) that will expire in the time interval between the target time and the processing deadline among M predicted business objects. The predicted object overdue amount of the M predicted business objects is also the predicted object overdue amount of the object set.

[0126] Please see also Figure 6 , Figure 6 This is a schematic diagram illustrating another data prediction scenario provided in this application. For example... Figure 6 As shown, the server can input the object features corresponding to M predicted business objects into the prediction model, and then use the prediction model to predict the predicted delinquency probability of each predicted business object.

[0127] The server can then input the predicted overdue probability of each predicted business object (which can also be understood as the probability distribution formed by the predicted overdue probabilities of each predicted business object, i.e., the overdue probability distribution mentioned above) into the calibration model. Through this calibration model, the input overdue probability distribution can be calibrated to obtain the calibrated overdue probability distribution, which includes the calibrated overdue probability of each predicted business object. The server can then calculate the expected overdue value of the calibrated overdue probability distribution (the calculation method is shown in the formula (1) above) and use the expected overdue value as the overdue amount of the predicted object set.

[0128] The server can directly use the predicted overdue amount of the object as the overdue data of the predicted object set. Alternatively, the server can also calculate the object overdue rate of the object set based on the overdue amount of the object, and then use the object overdue rate as the overdue data of the predicted object set. Please refer to the following description.

[0129] The server can obtain N business objects that were cancelled before the target time. These business objects can be called cancelled business objects. Then, the server can calculate the overdue rate of the object set by the number of N business objects (equal to N, which can be called the first quantity), the number of cancelled business objects (actually known, which can be called the second quantity), and the object overdue amount. This overdue rate can be called the object overdue rate. There are two ways to calculate the object overdue rate, as described below.

[0130] The first method for calculating the object delinquency rate: Since the number of business objects canceled between the target time and the processing deadline is usually very small, the business objects canceled within this time interval (i.e., the time interval formed by the target time and the processing deadline) can be ignored. Therefore, the server can use the difference between the first quantity and the second quantity as the total number of reference objects. Furthermore, the server can use the ratio of the predicted delinquent object quantity to this total number of reference objects as the predicted object delinquency rate. Originally, the object delinquency rate Y is as shown in the following formula (2):

[0131]

[0132] And when N + < <N-N - At that time, N + This can be ignored, and the object delinquency rate Y is as shown in the following formula (3):

[0133]

[0134] Where Y represents the object delinquency rate, N1 represents the predicted object delinquency amount, and N represents the total number of business objects created in the object set (i.e., the first quantity).- N represents the number of business objects in the object set that were revoked before the target time (i.e., the second number mentioned above). + This indicates the number of business objects that were cancelled within the time interval between the target time and the processing deadline.

[0135] The second way to calculate the overdue rate of objects: This method does not ignore business objects that are cancelled within the time interval consisting of the target time and the processing deadline. Optionally, the number of business objects cancelled within the corresponding historical time interval can be used to estimate the number of business objects cancelled within the time interval (this number can be estimated and can be called the third number), as described below.

[0136] The aforementioned object set can be created within the target time period (which can be any natural day). Therefore, the server can obtain multiple historical time periods for the target time period. A historical time period can be any day before the target time period, and different historical time periods can be different days. A historical time period corresponds to a historical object set, and the historical object set corresponding to a historical time period can be created within that historical time period. A historical object set can also contain several business objects created in the past. The actual overdue business objects in the historical object sets corresponding to each historical time period are all known (i.e., have already occurred).

[0137] Therefore, the server can obtain the number of business objects (which can be called the object count) that were revoked within the historical time interval corresponding to each historical period (i.e., the time interval formed by the target time and the processing deadline) within the aforementioned time interval. Furthermore, the server can use the average number of revoked business objects within the historical time interval corresponding to each historical period (i.e., the average value) as the predicted number of business objects to be revoked within the time interval, i.e., the third quantity mentioned above. The relationship between a historical period and its corresponding historical time interval is similar to the relationship between the target period and the aforementioned time interval.

[0138] For example, if the target time period for creating an object set is June 1st, the processing deadline for the object set is June 30th, and the target completion time is June 5th, then the aforementioned time interval would be the period from the completion time of June 5th to the completion time of June 30th. Similarly, if a historical time period for creating a historical object set is April 1st, and the processing deadline for that historical object set is April 30th, then the historical time interval corresponding to the aforementioned time interval would be the period from the completion time of April 5th (i.e., the 5th day after the creation of the historical object set) to the completion time of April 30th.

[0139] Therefore, the server can calculate the sum of the second and third quantities as the object cancellation amount, and then calculate the difference between the first quantity and the object cancellation amount as the reference total quantity. Subsequently, the server can use the ratio of the predicted object delinquency amount to the reference total quantity as the predicted object delinquency rate.

[0140] Through the above process, the overdue data of the object set can be predicted. This overdue data may include the number of overdue objects and / or the overdue rate of the object set.

[0141] Furthermore, the server can also obtain the relationship between the predicted object delinquency rate and the delinquency rate threshold (which can be set according to the actual application scenario). The delinquency rate threshold can be the maximum delinquency rate that can be tolerated. If the object delinquency rate is greater than the delinquency rate threshold, the server can perform risk control processing on the object set.

[0142] Optionally, N business objects can refer to N business orders created within the same natural day, and M predicted business objects can refer to M business orders among the N business orders that have not yet been processed before the target time. Processing any business order can refer to the resource transfer processing or cancellation processing of that business order. If the object delinquency rate is greater than the delinquency rate threshold, the risk control processing of the object set can include: the server can send order processing prompt information to the business devices corresponding to the M business orders respectively, so as to remind the user to which the corresponding business device belongs to process the business order through the order processing prompt information. The business device corresponding to a business order can refer to the device that created the business order, such as the terminal device held by the user who created the business order.

[0143] Optionally, if the object delinquency rate is greater than the delinquency rate threshold, the risk control processing of the object set may also include: the server can establish voice communication connections with the business devices corresponding to the M business orders respectively, and use the established voice communication connections with the business devices to provide voice prompts to the users of the business devices to process the business orders.

[0144] Using the method provided in this application, a bottom-up approach (e.g., predicting the predicted overdue probability of each business object) and an top-down approach (e.g., predicting the overdue amount or overdue rate on a certain day) can be adopted to predict the overdue rate of business objects. Instead of directly using the predicted overdue probability to obtain the number of overdue business objects in the object set, the probability distribution formed by the predicted overdue probabilities of each business object (e.g., the overdue probability distribution) is calibrated to a probability distribution that is closer to the true distribution (e.g., the calibrated probability distribution). The expected value of the overdue business objects can be calculated using the calibrated probability distribution to predict the number of overdue business objects in the object set.

[0145] Furthermore, this application is also applicable to business scenarios where there is no pattern to the changes in overdue data on a historical timeline. In other words, this application can deduce the overdue data of business objects in the target time period without using the changing patterns of historical overdue data. Therefore, this application is applicable to a wide range of scenarios.

[0146] This application can obtain an object set; the object set contains N business objects, and the object set corresponds to a processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer; M predicted business objects are obtained from the object set; the M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline is reached; M is a positive integer and M is less than or equal to N; the overdue probability distribution of the M predicted business objects is obtained; the overdue probability distribution is formed based on the predicted overdue probabilities corresponding to the M predicted business objects respectively; the overdue probability distribution is calibrated to obtain the calibrated overdue probability distribution of the M predicted business objects; the calibrated overdue probability distribution contains the calibrated overdue probabilities corresponding to the M predicted business objects respectively; the overdue data of the object set is predicted based on the calibrated overdue probability distribution of the M predicted business objects. Therefore, the method proposed in this application can calibrate the overdue probability distribution formed by the predicted overdue probability of each business object in the object set. Subsequently, the overdue data of the object set can be predicted by the probability distribution formed by the calibrated overdue probabilities of each business object, thus improving the accuracy of the predicted overdue data of the object set. Furthermore, since the prediction can be made without relying on regular historical overdue data, it also expands the applicability of predicting overdue data of the object set.

[0147] Based on the above description, the specific process of training the prediction model is described here. Please refer to [link / reference]. Figure 7 , Figure 7 This is a flowchart illustrating a prediction model training method provided in this application. The execution entity in the embodiments of this application can be the same as described above. Figure 3 The execution entities in the corresponding embodiments are the same, such as Figure 7As shown, the method may include:

[0148] Step S201: Obtain the sample object set; the sample object set contains multiple sample business objects, each of which carries a corresponding object tag. The object tag carried by any sample business object is either an overdue tag or a non-overdue tag.

[0149] Optionally, the server can obtain a sample object set, which may contain multiple business objects. These business objects can be referred to as sample business objects. Each sample business object can carry a corresponding object tag. The object tag carried by any sample business object can be an overdue tag (represented by a tag of 1) or a non-overdue tag (represented by a tag of 0). The overdue tag indicates that the corresponding sample business object is actually overdue, while the non-overdue tag indicates that the corresponding sample business object is not actually overdue. This sample object set can be a collection created within a certain period prior to the creation of the aforementioned object set (e.g., one or two months prior), and the actual overdue or non-overdue results in this sample object set have already occurred (i.e., are known).

[0150] Step S202: Obtain the sample object features of each sample business object in the sample object set;

[0151] Optionally, the server can also obtain the object features corresponding to each sample business object in the sample object set. These object features can be referred to as sample object features. The process of obtaining the object features of each sample business object is the same as the process of obtaining the object features of the target business object described above; please refer to the above for details. Figure 3 The relevant descriptions in the corresponding embodiments will not be repeated here.

[0152] Step S203: Obtain the initial prediction model and call the initial prediction model to predict the sample prediction overdue probability of each sample business object based on the sample object characteristics of each sample business object;

[0153] Optionally, the server can obtain an initial prediction model, which can be an XGB model (a tree model) and is the model to be trained. The server can input the sample object features of each sample business object into the initial prediction model. The server can call the initial prediction model to predict the predicted delinquency probability of each sample business object based on its sample object features. This predicted delinquency probability is the probability that the sample business object will be delinquent, as predicted by the initial prediction model.

[0154] Step S204: Update the model parameters of the initial prediction model based on the sample prediction overdue probability of each sample business object and the object label carried by each sample business object, and determine the initial prediction model after updating the model parameters as the prediction model.

[0155] Optionally, the server can update the model parameters of the initial prediction model based on the predicted overdue probability of each sample business object and the object labels carried by each sample business object. During the update, the model parameters of the initial prediction model can be updated to make the predicted overdue probability of sample business objects carrying overdue labels higher, while making the predicted overdue probability of sample business objects carrying non-overdue labels lower. By using the above sample object set, the initial prediction model can be continuously trained iteratively. The trained initial prediction model (e.g., trained to convergence or trained to a certain number of training times threshold) can then be used as the final prediction model to predict the overdue probability of business objects.

[0156] Optionally, the initial prediction model can be a three-class classification model or a binary classification model. If the initial prediction model is a three-class classification model, the three classifications can include successfully processed categories, withdrawn categories, and overdue categories. The server can call the initial prediction model to predict the probability of each sample business object belonging to each of the three categories based on the sample object characteristics of each sample business object. The probability of a sample business object belonging to the overdue category is the sample predicted overdue probability of the sample business object (i.e., the probability of not being processed overdue). The probability of a sample business object belonging to the successfully processed category is the probability that the sample business object will be processed before the processing deadline (such as performing resource transfer processing). The probability of a sample business object belonging to the withdrawn category is the probability that the sample business object will be withdrawn before the processing deadline. The sum of the probabilities of a business object in each of the three categories can be equal to 1.

[0157] If the initial prediction model is a binary classification model, the binary classification can include a category that combines the categories of successfully processed and withdrawn, as well as an overdue category. Similarly, the server can also call the initial prediction model to predict the probability of each sample business object belonging to each category in the binary classification based on the sample object characteristics of each sample business object. The probability of a sample business object belonging to the overdue category is the sample predicted overdue probability of the sample business object (i.e., the probability of overdue and unprocessed). The sum of the probabilities of a business object in each binary classification can be equal to 1.

[0158] Alternatively, the initial prediction model of this application can also be a multi-classification model (the specific number of categories can be determined according to the actual application scenario). The target of classification (such as whether the prediction target is overdue) can be treated as one category, while other multiple categories can be merged into another category for prediction.

[0159] This training method allows for the prediction of revoked business objects during training. In subsequent actual predictions, revoked business objects will not be included in the calculation of overdue data. This means that the interference caused by revoked (i.e. canceled) business objects can be removed from all business objects created in the object set, making the calculated overdue data more accurate.

[0160] It's important to note that the predictive model can be iteratively updated over time using the most recent set of sample objects. For example, in March, predictive model 1 trained on the sample object set from January or February (i.e., the set of objects created in January or February that can be used as samples) can be used. In May, predictive model 2 trained on the sample object set from March or April (i.e., the set of objects created in March or April that can be used as samples) can be used. In July, predictive model 3 trained on the sample object set from May or June (i.e., the set of objects created in May or June that can be used as samples) can be used, and so on. By continuously iterating and updating the predictive model using the latest data, the model becomes more suitable for current predictions, and the prediction results (such as the predicted probability of delinquency) become more accurate and realistic. This also avoids situations where merchants (such as the merchants to whom the business object belongs) are unaware of data changes.

[0161] Based on the above description, the specific process of training and obtaining the calibration model is described here. Please refer to [link to documentation]. Figure 8 , Figure 8 This is a flowchart illustrating a calibration model training method provided in this application. The execution entity in the embodiments of this application can be the same as described above. Figure 3 The execution entities in the corresponding embodiments are the same, such as Figure 8 As shown, the method may include:

[0162] Step S301: Obtain the sample prediction overdue probability corresponding to each sample business object predicted by the initial prediction model during the training process of the initial prediction model;

[0163] Optional, the initial calibration model and the above Figure 7In the corresponding embodiment, the initial prediction model can be trained simultaneously, but the training of the initial calibration model and the initial prediction model are independent of each other. Only when training the initial calibration model is the prediction result of the prediction model (such as the predicted overdue probability of each sample business object) needed to be used.

[0164] Therefore, the server can obtain the predicted overdue probability of each sample business object (i.e., the probability that each sample business object will be overdue) predicted by the initial prediction model during the training of the initial prediction model, and can use the predicted overdue probability of the sample as data for training the initial calibration model.

[0165] Step S302: Obtain the initial calibration model and call the initial calibration model to calibrate the probability distribution formed by the predicted overdue probability of multiple sample business objects based on the object label carried by each sample business object.

[0166] Optionally, the predicted overdue probability of each of the above sample business objects can also carry a corresponding object label, which can be an overdue label or a non-overdue label. The server can obtain the initial calibration model, which can be a Platt model (a type of calibration model), or an ordinal-preserving regression model, etc., depending on the actual scenario, and there are no restrictions on this.

[0167] The server can also input the predicted overdue probability (carrying the corresponding label) of each sample business object into the initial calibration model. The initial calibration model is then invoked to calibrate the probability distribution formed by the predicted overdue probabilities of multiple sample business objects based on the object label carried by each sample business object. During the calibration process, the model parameters of the initial calibration model can be adjusted so that the predicted overdue probability of samples carrying overdue labels can be larger after calibration, and the predicted overdue probability of samples carrying non-overdue labels can be smaller after calibration.

[0168] Step S303: The initial calibration model after updating the model parameters during the calibration process of the probability distribution formed by the predicted overdue probabilities of multiple sample business objects is determined as the calibration model.

[0169] Optionally, the server can use the initial calibration model, which is completed during the calibration process of the probability distribution formed by the predicted overdue probabilities of multiple sample business objects, as the final prediction model used to calibrate the probability distribution formed by the predicted overdue probabilities of several business objects (such as the overdue probability distribution mentioned above).

[0170] Optionally, when continuously iterating and updating the prediction model using the sample object set closest to the current time, the prediction results of the prediction model during the iterative update process can also be used to iteratively update the calibration model to obtain a calibration model that is more suitable for the current prediction, making the calibration results of the calibration model more accurate.

[0171] Optionally, if the initial prediction model is a three-class classification model, the successfully processed categories and the revoked categories can be merged. The merged category corresponds to the same object label (e.g., label 0, indicating not overdue), while the overdue category is classified separately. If the initial prediction model is a binary classification model, there is no need to merge; the two categories can be retained. Therefore, it can be understood that the trained calibration model can be a binary logistic regression (LR) model, used to calibrate the probability that business objects will be overdue (e.g., predicting the overdue probability, which is the probability of predicting that the object label is label 1 (i.e., overdue label)).

[0172] Please see Figure 9 , Figure 9 This is a schematic diagram illustrating a model training scenario provided in this application. For example... Figure 9 As shown, the server can input the sample object features (carrying corresponding object labels) of each sample business object into the initial prediction model. By calling this initial prediction model, the predicted delinquency probability of each sample business object can be predicted based on its sample object features. The prediction loss (i.e., classification loss) can be obtained by comparing the predicted delinquency probability of each sample business object with its corresponding object label. This prediction loss is then backpropagated to the initial prediction model to correct its model parameters (e.g., to minimize the prediction loss). Finally, the initial prediction model with corrected model parameters can be used as the trained prediction model. It should be noted that the initial prediction model and the prediction model are the same model; the initial prediction model is the model that has not yet been fully trained, while the prediction model is the model that has been fully trained and can be applied.

[0173] Furthermore, such as Figure 9As shown, the predicted delinquency probabilities (carrying corresponding object labels) of each sample business object obtained from the initial prediction model can be input into the initial calibration model. This initial calibration model calibrates the predicted delinquency probabilities of each sample business object, yielding the calibrated predicted delinquency probabilities. The calibration loss can be obtained by comparing the calibrated predicted delinquency probabilities with their corresponding object labels. This calibration loss is then backpropagated to the initial calibration model to correct its parameters (e.g., to minimize the calibration loss). Finally, the initial calibration model with corrected parameters can be used as the trained calibration model. It should be noted that the initial calibration model and the calibration model are the same model; the initial calibration model is the untrained model, while the calibration model is the trained and ready-to-use model.

[0174] Please see Figure 10 , Figure 10 This is a schematic diagram of a model training process provided in this application. For example... Figure 10 As shown, s1: Order-level feature derivation refers to generating features for each order (e.g., each sample business object) (which can be called derived features, such as sample object features). s2: Using the derived features of each order, the classification model (i.e., the initial prediction model mentioned above) can be called to predict the overdue probability, thus training the classification model. s3: Furthermore, the overdue probability of the orders predicted by the classification model (e.g., the predicted overdue probability of the samples) can be used to train the calibration model (which is now the initial calibration model). (During training, the calibration probability (e.g., the predicted overdue probability of the calibrated samples) can be predicted). The classification model and calibration model can also use the latest historical data as samples for rolling iterative updates during training. s4: The trained prediction model and calibration model can then predict the bad debt rate (e.g., the overdue rate of the objects) of the object set.

[0175] It is known that, in the actual result verification experiment, the expected value of the overdue rate is obtained by calculating the overdue rate of each sample business object in the sample object set obtained by the calibration model (as described above). This expected value is close to or equal to the number of actual overdue business objects in the sample object set.

[0176] By continuously using the most recent data to iteratively update and train the prediction model (which can be called the classification model) and the calibration model, the accuracy of the obtained prediction model and calibration model can be improved. See Table 1 below for the experimental results:

[0177]

[0178] Table 1

[0179] Specifically, Method A can refer to using April data (such as business objects created in April) as samples (such as a sample object set) to train a predictive model (i.e., a classification model), and using April data as samples to train a calibration model; Method B can refer to using April data (such as business objects created in April) as samples to train a predictive model (i.e., a classification model), and using July data (such as business objects created in July) as samples to train a calibration model; Method C can refer to using July data (such as business objects created in July) as samples to train a predictive model (i.e., a classification model), and using July data as samples to train a calibration model.

[0180] As shown in Table 1 above, when the prediction and calibration models are trained using method C (i.e., using the most recent data as samples to train the prediction and calibration models), the order bad debt rate (which could be the overdue rate of the aforementioned objects) and the amount bad debt rate of business objects (such as orders created daily) created each day within a certain month will be very small.

[0181] For example, the mean error of the daily order bad debt rate predicted using the prediction model and calibration model trained using method C is 4.62%, and the mean error of the daily amount bad debt rate predicted using the prediction model and calibration model trained using method C is 2.62%. Similarly, the mean error of the daily order bad debt rate predicted using the prediction model and calibration model trained using method C is 4.79%, and the mean error of the daily amount bad debt rate predicted using the prediction model and calibration model trained using method C is 4.40%. It is evident that the prediction error of the order bad debt rate and the calibration bad debt rate obtained using method C are the smallest.

[0182] Furthermore, the mean error of the daily order bad debt rate predicted in September using the prediction model trained with Method A and the calibration model is 14.15%, and the mean error of the daily amount bad debt rate predicted in September using the prediction model trained with Method A and the calibration model is 13.31%. The mean error of the daily order bad debt rate predicted in October using the prediction model trained with Method A and the calibration model is 11.48%, and the mean error of the daily amount bad debt rate predicted in October using the prediction model trained with Method A and the calibration model is 10.91%.

[0183] The mean error of the daily order bad debt rate predicted by the prediction model and calibration model trained using Method B in September is 13.36%, and the mean error of the daily amount bad debt rate predicted using the prediction model and calibration model trained using Method B in September is 12.58%. The mean error of the daily order bad debt rate predicted using the prediction model and calibration model trained using Method B in October is 10.70%, and the mean error of the daily amount bad debt rate predicted using the prediction model and calibration model trained using Method B in October is 10.19%.

[0184] Please see Figures 11a-11b , Figures 11a-11b This is a graphical representation of order data provided in this application. For example... Figure 11a As shown, Figure 11a The figure shows the curves of daily order bad debt rate changes in September and October obtained using method A, and the curves of daily order bad debt rate changes in September and October obtained using method B.

[0185] The curves showing the daily order bad debt rate changes in September, October, and September are presented, along with the curves showing the actual daily order bad debt rate changes in September and October, obtained using method C. It is evident that the curve showing the order bad debt rate changes using method C is the closest to the actual order bad debt rate changes.

[0186] Similarly, such as Figure 11b As shown, Figure 11b The image shows the daily bad debt rate changes in September and October using Method A, Method B, and Method C, as well as the actual daily bad debt rate changes in September and October. It is evident that the bad debt rate change curve obtained using Method C is the closest to the actual bad debt rate change curve.

[0187] The above curves further demonstrate that by adopting the method provided in this application and using the latest data as samples to train the prediction model and calibration model, a more accurate prediction model and calibration model can be trained. In turn, a more accurate prediction of overdue data for the object set can be achieved through the more accurate prediction model and calibration model.

[0188] The formula for calculating the monthly average order bad debt rate prediction error is shown in the following formula (4):

[0189]

[0190] Where t represents the number of days in the predicted month, and j represents the j-th day of that month. This represents the order delinquency rate for day j predicted by the prediction model and the calibration model (e.g., the delinquency rate of the aforementioned object, i.e., the predicted bad debt rate). This represents the actual delinquency rate (i.e., the actual bad debt rate) on day j.

[0191] The formula for calculating the monthly average bad debt rate prediction error is shown in the following formula (5):

[0192]

[0193] Where t represents the number of days in the predicted month, j represents the j-th day of that month, and p s S represents the predicted calibration overdue probability of the s-th order created on day j. j This represents the number of orders created on day j that have not yet been processed before the predicted time (e.g., the 5th day after the order is created, similar to the meaning of the target time mentioned above), where s represents S. j The s-th order among the corresponding orders, the order amount in formula (5) here is the amount of the s-th order, N j (Similar to the meaning of the first quantity N above) represents the total number of orders created on day j. This represents the number of orders created on day j that were cancelled (i.e., cancelled) before the predicted time; it can be seen that in formula (5) This represents the predicted bad debt rate for day j. This represents the actual amount of bad debt on day j.

[0194] The experimental data above show that, by using the method provided in this application, data that is closer to the present can be used as samples to train the prediction model and calibration model. Subsequently, the prediction model and calibration model can be used to accurately predict overdue data.

[0195] Please see Figure 12 , Figure 12This is a schematic diagram of the structure of a service processing apparatus provided in this application. The service processing apparatus can be a computer program (including program code) running on a computer device; for example, the service processing apparatus is application software. This service processing apparatus can be used to execute corresponding steps in the methods provided in the embodiments of this application. Figure 12 As shown, the service processing device 1 may include:

[0196] The first acquisition module 11 is used to acquire an object set; the object set contains N business objects, and the object set corresponds to a processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer;

[0197] The second acquisition module 12 is used to acquire M predicted business objects from the object set; the M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline; M is a positive integer and M is less than or equal to N.

[0198] The distribution acquisition module 13 is used to acquire the overdue probability distribution of M predicted business objects; the overdue probability distribution is formed based on the predicted overdue probability corresponding to each of the M predicted business objects.

[0199] The distribution calibration module 14 is used to calibrate the overdue probability distribution to obtain the calibrated overdue probability distribution of M predicted business objects; the calibrated overdue probability distribution contains the calibrated overdue probabilities corresponding to the M predicted business objects respectively.

[0200] Prediction module 15 is used to predict overdue data of the object set based on the calibration overdue probability distribution of M prediction business objects.

[0201] Optionally, the distribution acquisition module 13 may acquire the overdue probability distribution of the M predicted business objects in the following ways:

[0202] Obtain the object features of each predicted business object and input the object features of each predicted business object into the prediction model;

[0203] The prediction model is invoked to determine the predicted delinquency probability for each predicted business object based on its object characteristics.

[0204] The overdue probability distribution is determined based on the predicted overdue probability corresponding to each predicted business object.

[0205] Optionally, any one of the M predicted business objects can be represented as the target business object; the distribution acquisition module 13 acquires the object features of each predicted business object in the following ways:

[0206] Obtain multiple object association information for the target business object;

[0207] Generate information features corresponding to the object association information of each target business object;

[0208] The information features corresponding to each object's associated information are concatenated to obtain the object features of the target business object.

[0209] Optionally, the above-mentioned device 1 is also used for:

[0210] Obtain the sample object set; the sample object set contains multiple sample business objects, each of which carries a corresponding object tag. The object tag carried by any sample business object is either an overdue tag or a non-overdue tag.

[0211] Obtain the sample object characteristics of each sample business object in the sample object set;

[0212] Obtain the initial prediction model and call the initial prediction model to predict the sample prediction delinquency probability of each sample business object based on the sample object characteristics of each sample business object;

[0213] The model parameters of the initial prediction model are updated based on the sample prediction overdue probability of each sample business object and the object label carried by each sample business object, and the initial prediction model after the model parameter update is determined as the prediction model.

[0214] Optionally, the distribution calibration module 14 calibrates the overdue probability distribution to obtain the calibrated overdue probability distribution of M predicted business objects, including:

[0215] Input the overdue probability distribution into the calibration model;

[0216] The calibration model is invoked to calibrate the overdue probability distribution, resulting in a calibrated overdue probability distribution.

[0217] Optionally, the overdue probability distribution is obtained based on the prediction model; the prediction model is obtained by training an initial prediction model, which is trained based on a sample object set. The sample object set contains multiple sample business objects, and each sample business object carries a corresponding object label, which is an overdue label or a non-overdue label.

[0218] The aforementioned device 1 is also used for:

[0219] Obtain the predicted overdue probability of each sample business object predicted by the initial prediction model during the training process of the initial prediction model;

[0220] Obtain the initial calibration model and call the initial calibration model to calibrate the probability distribution formed by the predicted overdue probabilities of multiple sample business objects based on the object tags carried by each sample business object;

[0221] The initial calibration model, after updating the model parameters during the calibration process of the probability distribution formed by the predicted overdue probabilities of multiple sample business objects, is determined as the calibration model.

[0222] Optionally, the prediction module 15 can predict the overdue data of the object set based on the calibrated overdue probability distribution of M prediction business objects, including:

[0223] Calculate the expected overdue value of M predicted business objects based on the calibration overdue probability distribution, and determine the expected overdue value as the overdue amount of the predicted object set;

[0224] The overdue data of the object set is determined based on the overdue amount of the objects.

[0225] Optionally, the prediction module 15 may determine the overdue data of the object set based on the amount of overdue objects in the following ways:

[0226] Among N business objects, the business objects that were revoked before the target time are identified as revoked business objects;

[0227] The object delinquency rate of the object set is determined based on the first number of N business objects, the second number of revoked business objects, and the object delinquency rate.

[0228] Use the object delinquency rate as delinquency data.

[0229] Optionally, the prediction module 15 determines the object delinquency rate of the object set based on the first number of N business objects, the second number of revoked business objects, and the object delinquency amount, including:

[0230] Get the third number of N business objects that were cancelled within the time interval formed by the target time and the processing deadline;

[0231] The sum of the second and third quantities is determined as the object cancellation amount;

[0232] The difference between the first quantity and the number of object cancellations is determined as the total number of reference objects;

[0233] The ratio between the number of overdue objects and the total number of reference objects is determined as the object overdue rate.

[0234] Optionally, the object set is created within the target time period; the prediction module 15 obtains the third number of business objects that are revoked within the time interval formed by the target time and the processing deadline from among the N business objects in the following ways:

[0235] Retrieve multiple historical time periods preceding the target time period; each historical time period corresponds to a set of historical objects that have been created.

[0236] Get the number of business objects that were revoked within the historical time interval indicated by the time interval corresponding to each historical period;

[0237] The average number of business objects that were revoked within each historical time interval is determined as the third quantity.

[0238] Optionally, the overdue data includes the object overdue rate of the object set; the above-mentioned device 1 is also used for:

[0239] Obtain the relationship between the predicted object delinquency rate and the delinquency rate threshold;

[0240] If the delinquency rate of an object exceeds the delinquency rate threshold, risk control measures will be applied to the object set.

[0241] Optionally, the N business objects included in the object set refer to the N business orders created within the same natural day, and the processing of any business order refers to the resource transfer processing for that business order; the M predicted business objects are the M business orders among the N business orders that have not yet been processed before the target time.

[0242] If the delinquency rate of the objects exceeds the delinquency rate threshold, the above-mentioned device 1 performs risk control processing on the object set in the following ways:

[0243] Send order processing notification messages to the business devices corresponding to each of the M business orders.

[0244] According to one embodiment of this application, Figure 3 The steps involved in the business processing method shown can be derived from... Figure 12 The various modules in the business processing unit 1 shown are used to execute this process. For example, Figure 3 Step S101 shown can be performed by Figure 12 The first acquisition module 11 in the process is executed. Figure 3 Step S102 shown can be performed by Figure 12 The second acquisition module 12 in the middle is used to execute; Figure 3 Step S103 shown can be performed by Figure 12 The distribution acquisition module 13 in the middle is used to execute, Figure 3 Step S104 shown can be derived from Figure 12 The distribution calibration module 14 in the middle is used to perform the calibration. Figure 3 Step S105 shown can be performed by Figure 12 The prediction module 15 in the middle is used to perform this.

[0245] This application can obtain an object set; the object set contains N business objects, and the object set corresponds to a processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer; M predicted business objects are obtained from the object set; the M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline is reached; M is a positive integer and M is less than or equal to N; the overdue probability distribution of the M predicted business objects is obtained; the overdue probability distribution is formed based on the predicted overdue probabilities corresponding to the M predicted business objects respectively; the overdue probability distribution is calibrated to obtain the calibrated overdue probability distribution of the M predicted business objects; the calibrated overdue probability distribution contains the calibrated overdue probabilities corresponding to the M predicted business objects respectively; the overdue data of the object set is predicted based on the calibrated overdue probability distribution of the M predicted business objects. Therefore, the device proposed in this application can calibrate the overdue probability distribution formed by the predicted overdue probability of each business object in the object set. Subsequently, the overdue data of the object set can be predicted by the probability distribution formed by the calibrated overdue probabilities of each business object, thus improving the accuracy of the predicted overdue data of the object set. Furthermore, since the prediction can also be made without relying on regular historical overdue data, the applicability of predicting overdue data of the object set is also expanded.

[0246] According to one embodiment of this application, Figure 12 The modules in the business processing device 1 shown can be individually or entirely combined into one or more units, or some of these units can be further divided into multiple functionally smaller sub-units to achieve the same operation without affecting the technical effects of the embodiments of this application. The above modules are based on logical functional division. In practical applications, the function of one module can be implemented by multiple units, or the function of multiple modules can be implemented by one unit. In other embodiments of this application, the business processing device 1 may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0247] According to one embodiment of this application, a general-purpose computer device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM), can perform operations such as... Figure 3 The computer program (including program code) for each step involved in the corresponding method shown, to construct such... Figure 12 The present invention describes a business processing apparatus 1 and a business processing method for implementing embodiments of this application. The computer program described above may be recorded on, for example, a computer-readable recording medium, loaded onto the aforementioned computing device via the same medium, and executed therein.

[0248] Please see Figure 13 , Figure 13 This is a schematic diagram of the structure of a computer device provided in this application. For example... Figure 13 As shown, the computer device 1000 may include a processor 1001, a network interface 1004, and a memory 1005. Furthermore, the computer device 1000 may also include a user interface 1003 and at least one communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as at least one disk storage device. Optionally, the memory 1005 may also be at least one storage device located remotely from the aforementioned processor 1001. Figure 13 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a device control application program.

[0249] exist Figure 13 In the computer device 1000 shown, the network interface 1004 provides network communication functionality; the user interface 1003 is mainly used to provide an input interface for the user; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:

[0250] Get the object set; the object set contains N business objects, and the object set corresponds to the processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer;

[0251] Obtain M predicted business objects from the object set; M predicted business objects refer to business objects that have not been processed by the target time before the processing deadline; M is a positive integer and M is less than or equal to N;

[0252] Obtain the delinquency probability distribution of M predicted business objects; the delinquency probability distribution is formed based on the predicted delinquency probability corresponding to each of the M predicted business objects.

[0253] The overdue probability distribution is calibrated to obtain the calibrated overdue probability distribution of M predicted business objects; the calibrated overdue probability distribution contains the calibrated overdue probability corresponding to each of the M predicted business objects.

[0254] Overdue data of the prediction object set based on the calibration overdue probability distribution of M prediction business objects.

[0255] Optionally, the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve:

[0256] Obtain the object features of each predicted business object and input the object features of each predicted business object into the prediction model;

[0257] The prediction model is invoked to determine the predicted delinquency probability for each predicted business object based on its object characteristics.

[0258] The overdue probability distribution is determined based on the predicted overdue probability corresponding to each predicted business object.

[0259] Optionally, any one of the M predicted service objects can be represented as the target service object; the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve:

[0260] Obtain multiple object association information for the target business object;

[0261] Generate information features corresponding to the object association information of each target business object;

[0262] The information features corresponding to each object's associated information are concatenated to obtain the object features of the target business object.

[0263] Optionally, the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve:

[0264] Obtain the sample object set; the sample object set contains multiple sample business objects, each of which carries a corresponding object tag. The object tag carried by any sample business object is either an overdue tag or a non-overdue tag.

[0265] Obtain the sample object characteristics of each sample business object in the sample object set;

[0266] Obtain the initial prediction model and call the initial prediction model to predict the sample prediction delinquency probability of each sample business object based on the sample object characteristics of each sample business object;

[0267] The model parameters of the initial prediction model are updated based on the sample prediction overdue probability of each sample business object and the object label carried by each sample business object, and the initial prediction model after the model parameter update is determined as the prediction model.

[0268] Optionally, the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve:

[0269] Input the overdue probability distribution into the calibration model;

[0270] The calibration model is invoked to calibrate the overdue probability distribution, resulting in a calibrated overdue probability distribution.

[0271] Optionally, the overdue probability distribution is obtained based on the prediction model; the prediction model is obtained by training an initial prediction model, which is trained based on a sample object set. The sample object set contains multiple sample business objects, and each sample business object carries a corresponding object label, which is an overdue label or a non-overdue label.

[0272] Processor 1001 can also be used to call device control applications stored in memory 1005 to achieve:

[0273] Obtain the predicted overdue probability of each sample business object predicted by the initial prediction model during the training process of the initial prediction model;

[0274] Obtain the initial calibration model and call the initial calibration model to calibrate the probability distribution formed by the predicted overdue probabilities of multiple sample business objects based on the object tags carried by each sample business object;

[0275] The initial calibration model, after updating the model parameters during the calibration process of the probability distribution formed by the predicted overdue probabilities of multiple sample business objects, is determined as the calibration model.

[0276] Optionally, the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve:

[0277] Calculate the expected overdue value of M predicted business objects based on the calibration overdue probability distribution, and determine the expected overdue value as the overdue amount of the predicted object set;

[0278] The overdue data of the object set is determined based on the overdue amount of the objects.

[0279] Optionally, the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve:

[0280] Among N business objects, the business objects that were revoked before the target time are identified as revoked business objects;

[0281] The object delinquency rate of the object set is determined based on the first number of N business objects, the second number of revoked business objects, and the object delinquency rate.

[0282] Use the object delinquency rate as delinquency data.

[0283] Optionally, the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve:

[0284] Get the third number of N business objects that were cancelled within the time interval formed by the target time and the processing deadline;

[0285] The sum of the second and third quantities is determined as the object cancellation amount;

[0286] The difference between the first quantity and the number of object cancellations is determined as the total number of reference objects;

[0287] The ratio between the number of overdue objects and the total number of reference objects is determined as the object overdue rate.

[0288] Optionally, the object set is created within the target time period;

[0289] Processor 1001 can also be used to call device control applications stored in memory 1005 to achieve:

[0290] Retrieve multiple historical time periods preceding the target time period; each historical time period corresponds to a set of historical objects that have been created.

[0291] Get the number of business objects that were revoked within the historical time interval indicated by the time interval corresponding to each historical period;

[0292] The average number of business objects that were revoked within each historical time interval is determined as the third quantity.

[0293] Optionally, overdue data includes the object overdue rate of the object set;

[0294] Processor 1001 can also be used to call device control applications stored in memory 1005 to achieve:

[0295] Obtain the relationship between the predicted object delinquency rate and the delinquency rate threshold;

[0296] If the delinquency rate of an object exceeds the delinquency rate threshold, risk control measures will be applied to the object set.

[0297] Optionally, the N business objects included in the object set refer to the N business orders created within the same natural day, and the processing of any business order refers to the resource transfer processing for that business order; the M predicted business objects are the M business orders among the N business orders that have not yet been processed before the target time.

[0298] Optionally, if the object delinquency rate exceeds the delinquency rate threshold, the processor 1001 can also be used to call the device control application stored in the memory 1005 to achieve:

[0299] Send order processing notification messages to the business devices corresponding to each of the M business orders.

[0300] It should be understood that the computer device 1000 described in the embodiments of this application can execute the foregoing text. Figure 3 The description of the above business processing method in the corresponding embodiments can also be executed as described above. Figure 12 The description of the above-mentioned service processing device 1 in the corresponding embodiments will not be repeated here.

[0301] This application can obtain an object set; the object set contains N business objects, and the object set corresponds to a processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, the unprocessed business objects are determined to be overdue; N is a positive integer; M predicted business objects are obtained from the object set; the M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline is reached; M is a positive integer and M is less than or equal to N; the overdue probability distribution of the M predicted business objects is obtained; the overdue probability distribution is formed based on the predicted overdue probabilities corresponding to the M predicted business objects respectively; the overdue probability distribution is calibrated to obtain the calibrated overdue probability distribution of the M predicted business objects; the calibrated overdue probability distribution contains the calibrated overdue probabilities corresponding to the M predicted business objects respectively; the overdue data of the object set is predicted based on the calibrated overdue probability distribution of the M predicted business objects. Therefore, the method proposed in this application can calibrate the overdue probability distribution formed by the predicted overdue probability of each business object in the object set. Subsequently, the overdue data of the object set can be predicted by the probability distribution formed by the calibrated overdue probabilities of each business object, thus improving the accuracy of the predicted overdue data of the object set. Furthermore, since the prediction can be made without relying on regular historical overdue data, it also expands the applicability of predicting overdue data of the object set.

[0302] Furthermore, it should be noted that this application also provides a computer-readable storage medium storing a computer program executed by the aforementioned business processing device 1. The computer program includes program instructions, which, when executed by a processor, enable the execution of the aforementioned... Figure 3 The description of the business processing method in the corresponding embodiments is already provided and will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the computer storage medium embodiments related to this application, please refer to the description of the method embodiments of this application.

[0303] As an example, the above program instructions can be deployed and executed on a single computer device, or deployed and executed on multiple computer devices located in one location, or executed on multiple computer devices distributed across multiple locations and interconnected via a communication network. Multiple computer devices distributed across multiple locations and interconnected via a communication network can form a blockchain network.

[0304] The aforementioned computer-readable storage medium can be an internal storage unit of the business processing apparatus provided in any of the foregoing embodiments or the computer device described above, such as a hard disk or memory of the computer device. The computer-readable storage medium can also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., provided on the computer device. Furthermore, the computer-readable storage medium may include both internal storage units and external storage devices of the computer device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0305] This application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the aforementioned... Figure 3 The description of the above-described business processing method in the corresponding embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the computer-readable storage medium embodiments related to this application, please refer to the description of the method embodiments of this application.

[0306] The terms "first," "second," etc., in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the term "comprising," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other step units inherent to these processes, methods, apparatuses, products, or devices.

[0307] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0308] The methods and related apparatuses provided in this application are described with reference to the method flowcharts and / or structural diagrams provided in this application. Specifically, each block of the method flowchart and / or structural diagram, as well as combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to create a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the process. Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 A process or multiple processes and / or structures illustrate the steps of the functions specified in one or more boxes.

[0309] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A business processing method, characterized in that, The method includes: Obtain an object set; the object set contains N business objects, and the object set corresponds to a processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, then the unprocessed business objects are determined to be overdue; N is a positive integer; M predicted business objects are obtained from the object set; the M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline is reached; M is a positive integer and M is less than or equal to N; Obtain the delinquency probability distribution of the M predicted business objects; the delinquency probability distribution is formed based on the predicted delinquency probability corresponding to each of the M predicted business objects. The overdue probability distribution is calibrated to obtain the calibrated overdue probability distribution of the M predicted business objects; the calibrated overdue probability distribution includes the calibration overdue probability corresponding to each of the M predicted business objects; The overdue data of the object set is predicted based on the calibration overdue probability distribution of the M predicted business objects; the overdue data is determined based on the overdue amount of the objects, the overdue amount of the objects is the expected overdue value, and the expected overdue value is calculated for the M predicted business objects based on the calibration overdue probability distribution; wherein, the process of determining the overdue data of the object set based on the overdue amount of the objects includes: Among the N business objects, those that were revoked before the target time are identified as revoked business objects; The object delinquency rate of the object set is determined based on the first number of the N business objects, the second number of the revoked business objects, and the object delinquency amount. The overdue rate of the object is used as the overdue data.

2. The method according to claim 1, characterized in that, The step of obtaining the overdue probability distribution of the M predicted business objects includes: Obtain the object features of each predicted business object, and input the object features of each predicted business object into the prediction model; The prediction model is invoked to determine the predicted delinquency probability for each predicted business object based on the object characteristics of each predicted business object. The overdue probability distribution is determined based on the predicted overdue probability corresponding to each predicted business object.

3. The method according to claim 2, characterized in that, Any one of the M predicted business objects is represented as the target business object; obtaining the object features of each predicted business object includes: Obtain multiple object association information of the target business object; Generate information features corresponding to each object association information of the target business object; The information features corresponding to each object's associated information are concatenated to obtain the object features of the target business object.

4. The method according to claim 3, characterized in that, The method further includes: Obtain a sample object set; the sample object set contains multiple sample business objects, each of which carries a corresponding object tag, and the object tag carried by any sample business object is either an overdue tag or a non-overdue tag; Obtain the sample object characteristics of each sample business object in the sample object set; Obtain the initial prediction model, and call the initial prediction model to predict the sample prediction overdue probability of each sample business object based on the sample object characteristics of each sample business object; The model parameters of the initial prediction model are updated based on the sample prediction overdue probability of each sample business object and the object label carried by each sample business object, and the initial prediction model after the model parameter update is determined as the prediction model.

5. The method according to claim 1, characterized in that, The calibration of the overdue probability distribution to obtain the calibrated overdue probability distribution of the M predicted business objects includes: Input the overdue probability distribution into the calibration model; The calibration model is invoked to calibrate the overdue probability distribution, thereby obtaining the calibration overdue probability distribution.

6. The method according to claim 5, characterized in that, The overdue probability distribution is obtained based on a prediction model; the prediction model is obtained by training an initial prediction model, which is trained based on a sample object set, which contains multiple sample business objects, each of which carries a corresponding object label, which is an overdue label or a non-overdue label. The method further includes: Obtain the sample prediction overdue probability corresponding to each sample business object predicted by the initial prediction model during the training process of the initial prediction model; Obtain the initial calibration model, and call the initial calibration model to calibrate the probability distribution formed by the predicted overdue probabilities of the samples corresponding to the multiple sample business objects based on the object tags carried by each sample business object; The initial calibration model, after updating the model parameters during the calibration process of the probability distribution formed by the predicted overdue probabilities of the samples corresponding to the multiple sample business objects, is determined as the calibration model.

7. The method according to claim 1, characterized in that, The step of determining the object delinquency rate of the object set based on the first number of the N business objects, the second number of the revoked business objects, and the object delinquency amount includes: Obtain the third number of business objects that are revoked within the time interval formed by the target time and the processing deadline among the N business objects; The sum of the second quantity and the third quantity is determined as the object cancellation amount; The difference between the first quantity and the object cancellation quantity is determined as the total number of reference objects; The ratio between the overdue amount of the object and the total number of reference objects is determined as the object overdue rate.

8. The method according to claim 7, characterized in that, The object set is created within the target time period; obtaining the third number of business objects that are revoked within the time interval formed by the target time and the processing deadline among the N business objects includes: Obtain multiple historical time periods preceding the target time period; each historical time period corresponds to a set of historical objects that have been created. Obtain the number of historical objects that were revoked within the historical time interval indicated by the time interval for each historical period; The third quantity is determined by the average number of business objects that were revoked within each historical time interval.

9. The method according to claim 1, characterized in that, The overdue data includes the object overdue rate of the object set; the method further includes: Obtain the relationship between the predicted overdue rate of the object and the overdue rate threshold; If the delinquency rate of the object is greater than the delinquency rate threshold, then risk control processing is performed on the object set.

10. The method according to claim 9, characterized in that, The N business objects included in the object set refer to the N business orders created within the same natural day, and the processing of any business order refers to the resource transfer processing for that business order. The M predicted business objects are the M business orders that have not been processed before the target time out of the N business orders; If the delinquency rate of the objects exceeds the delinquency rate threshold, the risk control processing of the object set includes: Send order processing prompts to the business devices corresponding to the M business orders respectively.

11. A business processing apparatus, characterized in that, The device includes: The first acquisition module is used to acquire an object set; the object set contains N business objects, and the object set corresponds to a processing deadline; when the processing deadline is reached, if there are any business objects in the object set that have not been processed, then the unprocessed business objects are determined to be overdue; N is a positive integer; The second acquisition module is used to acquire M predicted business objects from the object set; the M predicted business objects refer to business objects that have not been processed at the target time before the processing deadline is reached; M is a positive integer and M is less than or equal to N; The distribution acquisition module is used to acquire the overdue probability distribution of the M predicted business objects; the overdue probability distribution is formed based on the predicted overdue probability corresponding to each of the M predicted business objects. The distribution calibration module is used to calibrate the overdue probability distribution to obtain the calibrated overdue probability distribution of the M predicted business objects; the calibrated overdue probability distribution includes the calibration overdue probability corresponding to each of the M predicted business objects. The prediction module is used to predict the overdue data of the object set based on the calibration overdue probability distribution of the M predicted business objects; the overdue data is determined based on the overdue amount of the objects, the overdue amount of the objects is the expected overdue value, and the expected overdue value is calculated for the M predicted business objects based on the calibration overdue probability distribution; wherein, the process by which the prediction module determines the overdue data of the object set based on the overdue amount of the objects includes: Among the N business objects, those that were revoked before the target time are identified as revoked business objects; The object delinquency rate of the object set is determined based on the first number of the N business objects, the second number of the revoked business objects, and the object delinquency amount. The overdue rate of the object is used as the overdue data.

12. A computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1-10.

13. A computer device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method according to any one of claims 1-10.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1-10.

Images