Risk prediction model training method and device and risk prediction method and device

Abstract

The invention provides a training method of a business risk prediction model and a business risk prediction method and device.The training method of the business risk prediction model comprises the steps that first sample business data in a first historical period and second sample business data in a second historical period are acquired, the first historical period comprises a second historical period, and the period length of the first historical period is greater than that of the second historical period; according to the first sample service data and the second sample service data, respectively obtaining a first model corresponding to a future first period and a second model corresponding to a future second period; inputting second sample service data into the second model, and predicting resource risk prediction data of the second sample service data corresponding to a second period in the future; and training the first model according to resource risk prediction data of the first sample service data and the second sample service data corresponding to the second period in the future to obtain a service risk prediction model.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of computer technology, and specifically to a risk prediction model training method and device, and a risk prediction method and device. Background Art

[0002] Virtual resources refer to the use of Internet technology, information technology, etc. to transform real resources into resources that can be disseminated through the Internet, or to transform real resources into information resources that can flow freely. Currently, common virtual resources include video resources, software resources, tutorial resources, digital assets or resources, etc. Virtual resources are provided by resource providers, and virtual resource requesters can apply for virtual resources from resource providers through the Internet.

[0003] In the scenario where a virtual resource requester applies for virtual resources from a resource provider through the Internet, it is often the case that some virtual resource requesters fail to circulate or use virtual resources safely due to abnormal operations such as illegal operations, idle resources, and overdue resource return, such as low utilization, improper use, and resource gaps. Therefore, it is usually necessary to identify whether the virtual resource requester has abnormal risks and provide data support for whether matching resources can be allocated to the virtual resource requester. Summary of the invention

[0004] The embodiments of the present disclosure propose a risk prediction model training method and device, and a risk prediction method and device. When training the business risk prediction model, the business data performance of the business object in a longer historical period and the business data performance in a shorter historical period are fully considered, so that the prediction results of long-term business risks output by the business risk prediction model can more fully reflect the short-term and long-term performance of the business object, and can effectively improve the prediction accuracy and effect of the long-term business risks corresponding to the business object.

[0005] In a first aspect, the present disclosure provides a method for training a business risk prediction model, comprising:

[0006] Acquire first sample business data within a first historical period and second sample business data within a second historical period, wherein the first historical period includes the second historical period, and the period length of the first historical period is greater than the period length of the second historical period;

[0007] According to the first sample business data and the second sample business data, respectively obtain a first model corresponding to a first future period and a second model corresponding to a second future period;

[0008] Inputting the second sample business data into the second model, and predicting resource risk prediction data corresponding to the second sample business data in the second future period;

[0009] The first model is trained based on the resource risk prediction data of the second future period corresponding to the first sample business data and the second sample business data to obtain a business risk prediction model.

[0010] In some optional implementations, the second sample business data is one group or multiple groups, and different groups of second sample business data correspond to historical second periods with different period lengths.

[0011] In some optional implementations, obtaining, according to the first sample service data and the second sample service data, a first model corresponding to a first future period and a second model corresponding to a second future period, respectively, includes:

[0012] Obtaining a first model corresponding to the first future period according to the first sample business data;

[0013] According to each group of the second sample business data, a second model corresponding to each group of the second sample business data is obtained respectively.

[0014] In some optional implementations, the first sample service data includes first object data and first resource expiration data, and the second sample service data includes second object data and second resource expiration data.

[0015] In some optional implementations, obtaining, according to the first sample service data and the second sample service data, a first model corresponding to a first future period and a second model corresponding to a second future period, respectively, includes:

[0016] Taking the first object data as input and the first resource expiration data as output, training a first preset model to obtain a first model corresponding to the first future period;

[0017] Taking the second object data of the second sample business data of each historical second period as input and the second resource expiration data of the second sample business data of each historical second period as output, the second preset model corresponding to the second sample business data of each historical second period is trained to obtain the second model corresponding to each future second period.

[0018] In some optional implementations, training the first model to obtain a business risk prediction model based on the resource risk prediction data of the second future period corresponding to the first sample business data and the second sample business data includes:

[0019] Acquire one or more prediction data combinations from one or more resource risk prediction data output by each second model, wherein the prediction data combination includes resource risk prediction data output by at least one second model;

[0020] For each of the prediction data combinations, a first model is trained using the first sample business data and each resource risk prediction data in the prediction data combination to obtain a business risk prediction model corresponding to each prediction data combination.

[0021] In some optional embodiments, the method further comprises:

[0022] Obtaining effect parameters of the business risk prediction model corresponding to the one or more prediction data combinations;

[0023] A target business risk prediction model whose effect parameters meet preset conditions and a target second model corresponding to the target business prediction model are selected for business risk prediction, wherein the target business prediction model and the second sample business data used to obtain the target second model are the same group.

[0024] In a second aspect, the present disclosure provides a business risk prediction method, comprising:

[0025] Obtain business data of the forecast object;

[0026] Input the business data into the second model to obtain resource risk prediction data of the prediction object in the second future period output by the second model, wherein the second model is obtained based on the second sample business data in the second historical period;

[0027] Input the business data and the resource risk prediction data of the prediction object in the second period in the future into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future, wherein the target business risk prediction model is obtained based on the first sample business data in the historical first period and the second sample business data corresponding to the resource risk prediction data of the second period in the future.

[0028] In some optional implementations, the second model has one or more, and the second model is used to predict the resource risk prediction data of the future second period corresponding to the second sample data used by the second model, and the second sample business data used by different second models are obtained as different groups, and the period lengths of the historical second periods corresponding to the second sample business data of different groups are different, then:

[0029] There are one or more business risk prediction models, and different business risk prediction models are obtained based on the risk prediction data of each resource in different prediction data combinations and the first sample business data in the first historical period.

[0030] In some optional implementations, the target business risk prediction model is a model whose effect parameters in the business risk prediction model meet preset conditions.

[0031] The step of inputting the business data into the second model to obtain the resource risk prediction data of the prediction object in the second period in the future outputted by the second model comprises: inputting the business data into one or more target second models to obtain the target resource risk prediction data of the prediction object in the second period in the future outputted by each target second model, wherein the target business prediction model and the second sample business data used to obtain the target second model are the same group;

[0032] The step of inputting the business data and the resource risk prediction data of the prediction object in the second period in the future into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future includes: inputting the business data and the target resource risk prediction data output by each target second model into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future.

[0033] In a third aspect, the present disclosure provides a training device for a business risk prediction model, comprising:

[0034] A first acquisition unit is configured to acquire first sample business data within a first historical period and second sample business data within a second historical period, wherein the first historical period includes the second historical period, and the period length of the first historical period is greater than the period length of the second historical period;

[0035] A first training unit is configured to obtain a first model corresponding to a first future period and a second model corresponding to a second future period respectively according to the first sample business data and the second sample business data;

[0036] An input unit, configured to input the second sample business data into the second model, and predict resource risk prediction data corresponding to the second sample business data in the second future period;

[0037] The second training unit is configured to train the first model according to the resource risk prediction data of the first sample business data and the second sample business data corresponding to the second future period to obtain a business risk prediction model.

[0038] In some optional implementations, the second sample business data is one group or multiple groups, and different groups of second sample business data correspond to historical second periods with different period lengths.

[0039] In some optional implementations, the first training unit is configured to:

[0040] Obtaining a first model corresponding to the first future period according to the first sample business data;

[0041] According to each group of the second sample business data, a second model corresponding to each group of the second sample business data is obtained respectively.

[0042] In some optional implementations, the first sample service data includes first object data and first resource expiration data, and the second sample service data includes second object data and second resource expiration data.

[0043] In some optional implementations, the first training unit is configured to:

[0044] Taking the first object data as input and the first resource expiration data as output, training a first preset model to obtain a first model corresponding to the first future period;

[0045] Taking the second object data of the second sample business data of each historical second period as input and the second resource expiration data of the second sample business data of each historical second period as output, the second preset model corresponding to the second sample business data of each historical second period is trained to obtain the second model corresponding to each future second period.

[0046] In some optional implementations, the second training unit is configured to:

[0047] Acquire one or more prediction data combinations from one or more resource risk prediction data output by each second model, wherein the prediction data combination includes resource risk prediction data output by at least one second model;

[0048] For each of the prediction data combinations, a first model is trained using the first sample business data and each resource risk prediction data in the prediction data combination to obtain a business risk prediction model corresponding to each prediction data combination.

[0049] In some optional embodiments, the device further comprises:

[0050] A verification unit configured to obtain effect parameters of a business risk prediction model corresponding to the one or more prediction data combinations;

[0051] The selection unit is configured to select a target business risk prediction model whose effect parameters meet preset conditions and a target second model corresponding to the target business prediction model for business risk prediction, wherein the target business prediction model and the second sample business data used to obtain the target second model are the same group.

[0052] In a fourth aspect, the present disclosure provides a business risk prediction device, comprising:

[0053] A second input unit is configured to obtain business data of a prediction object;

[0054] A second model is configured to receive the business data and output resource risk prediction data of the prediction object in the second future period, wherein the second model is obtained based on second sample business data in the second historical period;

[0055] The target business risk prediction model is configured to receive the business data and the resource risk prediction data of the prediction object in the second period in the future to the target business risk prediction model, and obtain the resource risk prediction data of the prediction object in the first period in the future, wherein the target business risk prediction model is obtained based on the first sample business data in the historical first period and the second sample business data corresponding to the resource risk prediction data of the second period in the future.

[0056] In some optional implementations, the second model has one or more, and the second model is used to predict the resource risk prediction data of the future second period corresponding to the second sample data used by the second model, and the second sample business data used by different second models are obtained as different groups, and the period lengths of the historical second periods corresponding to the second sample business data of different groups are different, then:

[0057] There are one or more business risk prediction models, and different business risk prediction models are obtained based on the risk prediction data of each resource in different prediction data combinations and the first sample business data in the first historical period.

[0058] In some optional implementations, the target business risk prediction model is a model in which the effect parameters of the business risk prediction model meet preset conditions, the second model includes one or more target second models, and the target business prediction model and the second sample business data used to obtain the target second model are the same group;

[0059] The target second model is configured to receive the business data as input and output target resource risk prediction data of the prediction object in the second period in the future;

[0060] The target business risk prediction model is configured to receive the business data and the target resource risk prediction data output by each of the target second models, and output the resource risk prediction data of the prediction object in the first period in the future.

[0061] In a fifth aspect, the present disclosure provides an electronic device, including:

[0062] one or more processors;

[0063] a storage device having one or more programs stored thereon,

[0064] When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of the first aspect and the second aspect.

[0065] In a sixth aspect, the present disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by one or more processors, implements the method as described in any one of the first aspect and the second aspect.

[0066] When training the business risk prediction model, the disclosed embodiment fully considers the business data performance of the business object in a relatively long historical period (i.e., the first historical period) and the business data performance in a relatively short historical period (i.e., the second historical period), and combines the business data of the relatively long period (the first historical period) and the business data of the relatively short period (the second historical period) to train a business risk prediction model for predicting a relatively long period in the future. When predicting the business risk (i.e., long-term business risk) of the business object in the first future period, the business risk prediction model will combine the short-term resource risk prediction data of the business object in the second future period, so that the prediction result of the long-term business risk output by the business risk prediction model can more fully reflect the short-term and long-term performance of the business object, and can effectively improve the prediction accuracy and effect of the long-term business risk corresponding to the business object. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Other features, objects and advantages of the present disclosure will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the drawings:

[0068] Figure 1 is a system architecture diagram of an embodiment that can be applied thereto according to the present disclosure;

[0069] Figure 2 is a flowchart of an embodiment of a training method for a business risk prediction model according to the present disclosure;

[0070] Figure 3-1 is a flowchart of another embodiment of a method for training a business risk prediction model according to the present disclosure;

[0071] Figure 3-2 is a decomposition flow chart according to an embodiment of 305 in 3-1 of the present disclosure;

[0072] Figure 4 is a flowchart of another embodiment of a training method for a business risk prediction model according to the present disclosure;

[0073] Figure 5is a flow chart of an embodiment of a business risk prediction method according to the present disclosure;

[0074] Figure 6 is a module diagram of an embodiment of a training device for a business risk prediction model according to the present disclosure;

[0075] Figure 7 is a module diagram of an embodiment of a business risk prediction device according to the present disclosure;

[0076] Figure 8 It is a structural diagram of a computer system of an electronic device suitable for implementing an embodiment of the present disclosure. DETAILED DESCRIPTION

[0077] The present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0078] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0079] It should be noted that the information, data and signals involved in this disclosure are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions.

[0080] The inventors have found that in the field of virtual resources, for example, in the scenario where a virtual resource requester applies for virtual resources from a resource provider via the Internet, some virtual resource requesters often cause virtual resources to be unsafely circulated or unsafely used due to abnormal operations such as illegal operations, idle resources, and overdue resource return. Therefore, it is necessary to identify whether a virtual resource requester has abnormal risks and provide data support for whether matching resources can be allocated to the virtual resource requester.

[0081] In the field of virtual resources, it is generally recognized by those skilled in the art that the business data of virtual resource requesters in a relatively long historical period can be used to identify the resource usage risk of virtual resource requesters in a relatively long future period. For example, by using the resource return overdue data of virtual resource requesters in a relatively long historical period, it is possible to predict whether the virtual resource requesters have the risk of overdue resource return. However, the business data of virtual resource requesters in a relatively short period cannot objectively reflect the resource usage risk of virtual resource requesters due to the relatively short period. If the business data in a relatively short period is used to predict the business risk of virtual resource requests in a relatively long period, it will have an inaccurate and objective effect, and will have negative interference.

[0082] Therefore, in the field of virtual resources, the industry uses longer-term and shorter-term business data separately for business risk prediction in different scenarios, specifically:

[0083] 1. The business data of the virtual resource requester in a longer historical period is used to identify the business risks of the virtual resource requester in a longer future period, such as predicting the credit risk of the business object in the field of financial technology, and using the business data of the business object in a longer historical period (such as resource return overdue data) to predict the risk in a longer future period. The longer period is generally longer, such as 2 months or more.

[0084] 2. The business data of the virtual resource requester in a relatively short historical period is used to identify the business risks of the virtual resource requester in a relatively short future period. For example, in the field of financial technology, the risk of theft of business objects is predicted. The business data of business objects in a relatively short historical period (such as resource return overdue data) is used to predict the risks in a relatively short future period. The relatively short period is generally shorter, and its cycle length is often shorter than that of the relatively long period, such as 1 week, 2 weeks or 1 month.

[0085] However, the inventors found that: in the traditional scheme, the business data of longer periods and shorter periods are used separately for business risk prediction in different scenarios, and the relevant technology fails to fully utilize the data performance provided by the business data in longer periods and shorter periods when predicting business risks, which limits the enhanced effect of the business data in shorter periods in predicting business risks in longer periods. Because the business data of the virtual resource requester in a shorter period can reflect the business performance of the virtual resource requester in the initial stage of resource use, if the business data of the virtual resource requester in a longer period is used for risk prediction, it is combined with the business data of the virtual resource party in a shorter period for prediction, which can improve the prediction ability and effect of the scheme, that is, enhance the prediction ability of the resource use risk of the virtual resource requester in a longer period.

[0086] The following will describe in detail the business risk prediction model training method and device, and the business risk prediction method and device provided by the embodiments of the present disclosure.

[0087] Figure 1 An exemplary system architecture 100 is shown to which embodiments of the business risk prediction model training method, business risk prediction method, apparatus, terminal device, and storage medium of the present disclosure can be applied.

[0088] like Figure 1As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0089] The user can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various client applications can be installed on the terminal devices 101, 102, 103, such as business risk prediction applications, business risk prediction model training applications, voice interaction applications, video conferencing applications, short video social applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0090] Terminal devices 101, 102, 103 can be hardware or software. When terminal devices 101, 102, 103 are hardware, they can be various electronic devices with microphones and speakers, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III, Moving Picture Experts Compression Standard Audio Layer 3), MP4 players (Moving Picture Experts Group Audio Layer IV, Moving Picture Experts Compression Standard Audio Layer 4), portable computers and desktop computers, etc. When terminal devices 101, 102, 103 are software, they can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules (for example, obtaining the first sample business data in the first historical period and the second sample business data in the second historical period), or it can be implemented as a single software or software module. No specific limitation is made here.

[0091] The server 105 may be a server that provides various services, such as a background server that processes the first sample service data in the first historical period and the second sample service data in the second historical period obtained by the terminal devices 101, 102, and 103. The background server may perform corresponding processing based on the resource application behavior information sent by the terminal device.

[0092] In some cases, the training method of the business risk prediction model provided in the present disclosure can be jointly executed by the terminal devices 101, 102, 103 and the server 105. For example, the steps of "obtaining the first sample business data within the first historical period and the second sample business data within the second historical period" can be executed by the terminal devices 101, 102, 103, and "obtaining the first model corresponding to the first future period and the second model corresponding to the second future period according to the first sample business data and the second sample business data, respectively; inputting the second sample business data into the second model to predict the resource risk prediction data corresponding to the second future period of the second sample business data; training the first model according to the resource risk prediction data corresponding to the first sample business data and the second sample business data of the second future period to obtain the business risk prediction model" can be executed by the server 105. The present disclosure does not limit this. Accordingly, the training device of the business risk prediction model can also be respectively arranged in the terminal devices 101, 102, 103 and the server 105.

[0093] In some cases, the training method of the business risk prediction model provided in the present disclosure can be executed by the server 105. Accordingly, the training device of the business risk prediction model can also be set in the server 105. In this case, the system architecture 100 may not include the terminal devices 101, 102, and 103.

[0094] In some cases, the training method of the business risk prediction model provided in the present disclosure can be executed by the terminal devices 101, 102, and 103. Accordingly, the training device of the business risk prediction model can also be set in the terminal devices 101, 102, and 103. In this case, the system architecture 100 may not include the server 105.

[0095] In other cases, the business risk prediction method provided by the present disclosure can be jointly executed by the terminal devices 101, 102, 103 and the server 105. For example, the step of "obtaining the business data of the prediction object" can be executed by the terminal devices 101, 102, 103, and the step of "inputting the business data into the second model to obtain the resource risk prediction data of the prediction object in the second period in the future output by the second model; inputting the business data and the resource risk prediction data of the prediction object in the second period in the future into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future" can be executed by the server 105. The present disclosure does not limit this. Accordingly, the business risk prediction device can also be respectively arranged in the terminal devices 101, 102, 103 and the server 105.

[0096] In some cases, the business risk prediction method provided by the present disclosure may be executed by the server 105 , and accordingly, the business risk prediction device may also be set in the server 105 . In this case, the system architecture 100 may not include the terminal devices 101 , 102 , and 103 .

[0097] In some cases, the business risk prediction method provided by the present disclosure may be executed by the terminal devices 101 , 102 , 103 , and accordingly, the business risk prediction apparatus may also be disposed in the terminal devices 101 , 102 , 103 . In this case, the system architecture 100 may not include the server 105 .

[0098] It should be noted that the server 105 can be hardware or software. When the server 105 is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server 105 is software, it can be implemented as multiple software or software modules (for example, for providing distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.

[0099] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0100] See also Figure 2 , Figure 2 A flow chart showing an embodiment of a method for training a business risk prediction model provided by the present disclosure is shown. Figure 2 The training method of the business risk prediction model shown can be applied to Figure 1 The terminal device or server shown in. The process includes at least 202-208.

[0101] 202. Obtain first sample business data within a first historical period and second sample business data within a second historical period.

[0102] In the embodiments of the present disclosure, the first sample business data and the second sample business data can be obtained based on the acquired sample business data. In some examples, the sample business data within a preset period can be obtained. In some cases, the preset period is greater than the first historical period, that is, the first historical period is part of the preset period; in other cases, the preset period is the first historical period.

[0103] After obtaining the sample business data within the preset period, the first sample business data belonging to the first historical period can be obtained, and the second sample data belonging to the second historical period can be obtained. In some embodiments, the period length of the first historical period is greater than the period length of the second historical period. For example, the first historical period can be 2 months, and the second historical period can be 1 week, 2 weeks, or 1 month. In some cases, the first historical period includes the second historical period, that is, the second historical period is part of the first historical period. For example, the first historical period can be the past 2 months (such as 2021 / 10 / 1-2021 / 11 / 30), and the second historical period can be 1 week, 2 weeks, or 1 month in the past 2 months, that is, the second historical period can be:

[0104] 1 week: 2021 / 10 / 1-2021 / 10 / 7, 2021 / 10 / 8-2021 / 10 / 14, ... (the period moves by week) ..., 2021 / 11 / 19-2021 / 11 / 25;

[0105] 2 weeks: 2021 / 10 / 1-2021 / 10 / 14, 2021 / 10 / 15-2021 / 10 / 28, ... (the period is shifted by 2 weeks) ..., 2021 / 11 / 12-2021 / 11 / 25;

[0106] Or 1 month: 2021 / 10 / 1-2021 / 10 / 31, 2021 / 11 / 1-2021 / 11 / 30.

[0107] It should be understood that the historical second cycle may also be a cycle other than the historical first cycle, and its cycle length is less than the cycle length of the historical first cycle. For example, the historical first cycle may be 2 months (such as 2021 / 10 / 1-2021 / 11 / 30), and the historical second cycle may be 1 week (such as 2021 / 9 / 26-2021 / 9 / 31).

[0108] 204. Obtain a first model corresponding to a first future period and a second model corresponding to a second future period respectively according to the first sample business data and the second sample business data.

[0109] In the disclosed embodiment, the first preset model is trained with the first sample business data to obtain the first model corresponding to the first future period; the second preset model is trained with the second sample business data to obtain the second model corresponding to the second future period. The first model is used to predict the business risk of the business object in the first future period, such as predicting the resource usage risk of the business object in the next 2 months. The second model is used to predict the business risk of the business object in the second future period, such as predicting the resource usage risk of the business object in the next week. At this time, the first model is trained based on a longer period (such as the first historical period), and does not incorporate the data performance of a shorter period (such as the second historical period), that is, it is not trained based on the data features of a shorter period (such as the second historical period), and the data features of a shorter period (such as the second historical period) have no effect on the prediction results of the first model.

[0110] In some examples, the sample business data includes object data and resource data. The object data includes an identifier of a virtual resource requester, which is used to represent the virtual resource requester. The embodiment of the present disclosure also refers to the virtual resource requester as a business object. The virtual resource requester can be a user account or a corporate account. Resource data can include resource usage data, such as resource overdue data (such as resource usage overdue time, resource return overdue time), resource return value, resource transfer data, resource application data (such as resource application business type, resource application provider information, resource application intention information, etc.), etc. Among them, resource transfer data refers to data for transferring resources to other accounts, such as account IDs of other accounts, resource values ​​of transferred resources, etc. In the financial technology scenario, the resource application business type can be the type of virtual resource business applied for (such as the type of loan product), the resource application provider information can be the information of the financial institution, and the resource application intention information can be the verification of the loan intention. In the network scenario, the resource application business type can be a business type such as server resources, network bandwidth, etc., the resource application provider information can be the information of the resource supplier, and the resource application intention information can be the authenticity of the application intention or the level of the intention. The resource overtime data is the difference data between the transfer time and the agreed time when the virtual resource requester transfers the virtual resource to the resource provider. In one example, the resource overtime data may refer to the resource return overtime. The agreed time is the time agreed upon by the virtual resource requester and the resource provider to transfer the virtual resource. In some cases, the agreed time is the agreed time for the virtual resource requester to transfer the virtual resource to the resource provider. For example: the time agreed upon by the virtual resource requester and the resource provider to transfer the virtual resource is 2021 / 10 / 1. If the actual time when the virtual resource requester transfers the virtual resource to the resource provider is 2021 / 10 / 20, the difference data between the transfer time and the agreed time is 19 days, and the resource overtime data is 19 days, that is, the resource return overtime is 19 days.

[0111] It should be understood that if the virtual resource requester has not transferred the virtual resources, the resource overdue data will accumulate with the time that the virtual resource requester has not transferred the virtual resources. For example: the time agreed upon by the virtual resource requester and the resource provider to transfer the virtual resources is 2021 / 10 / 1. If the virtual resource requester has not transferred the virtual resources to the resource provider on 2021 / 10 / 20, the difference between the transfer time and the agreed time is 19 days, and the resource overdue data is 19 days; when the time reaches 2021 / 10 / 21, the difference between the transfer time and the agreed time is 20 days, and the resource overdue data is 20 days, that is, the resource return overdue time is 20 days.

[0112] Based on the above example, the process of acquiring the first model is: taking the first object data as input and the first resource overdue data as output, the first preset model is trained to obtain the first model corresponding to the first period in the future. The first model obtained is used to predict the resource overdue risk of the business object in the first period in the future, such as predicting the resource overdue risk of the business object in the next 2 months. The first model obtained is trained based on a longer period (such as the first historical period), and the data characteristics of the shorter period (such as the second historical period) have no effect on the prediction results of the first model.

[0113] The process of acquiring the second model is: taking the second object data as input and the second resource overdue data as output, training the second preset model to obtain the second model corresponding to the second period in the future. The obtained second model is used to predict the resource overdue risk of the business object in the second period in the future, such as predicting the resource overdue risk of the business object in the next week.

[0114] In the embodiments of the present disclosure, the first preset model and the second preset model may be the same model or different models. For example, the first preset model may be the LightGBM gradient boosting decision tree algorithm, or the XGBoost or RF (random forest) algorithm; the second preset model may be the LightGBM gradient boosting decision tree algorithm, or the XGBoost or RF (random forest) algorithm. The present disclosure does not limit the model type, structure, and algorithm of the first preset model and the second preset model.

[0115] 206. Input the second sample business data into the second model, and predict the resource risk prediction data of the second sample business data corresponding to the second period in the future.

[0116] The second model is used to predict the business risk of the business object in the second future period, such as predicting the resource usage risk of the business object in the next week. In the disclosed embodiment, the business risk can be a resource risk, and the resource risk can be a resource usage risk, such as resource overdue risk (such as resource usage overdue risk, resource return overdue risk), resource return insufficient risk, resource transfer abnormality risk, etc. Among them, resource transfer abnormality refers to the operation of transferring resources to an abnormal account.

[0117] By inputting the second sample business data into the second model, resource risk prediction data of the business object in the second future period can be obtained.

[0118] 208. Train a first model based on the resource risk prediction data of the second period in the future corresponding to the first sample business data and the second sample business data to obtain a business risk prediction model.

[0119] As can be seen from 208, when training the business risk prediction model, the disclosed embodiment fully considers the business data performance of the business object in the longer historical period (i.e., the first historical period) and the business data performance in the shorter historical period (i.e., the second historical period), and combines the business data of the longer period (the first historical period) and the business data of the shorter period (the second historical period) to train a business risk prediction model for predicting the longer future period. When predicting the business risk (i.e., long-term business risk) of the business object in the first future period, the business risk prediction model will combine the short-term resource risk prediction data of the business object in the second future period, so that the prediction result of the long-term business risk output by the business risk prediction model can more fully reflect the short-term and long-term performance of the business object, and can effectively improve the prediction accuracy and effect of the long-term business risk of the business object.

[0120] In order to better improve the prediction accuracy and effect of the long-term business risk of the business object, the embodiment of the present disclosure can also combine the short-term business data performance of multiple different historical second periods to train the business risk prediction model, so that when predicting the long-term business risk of the business object in the first period in the future, it can fully consider the business data performance of the business object in a longer period, as well as the business data performance of multiple different shorter periods. The more shorter periods considered during training, the more choices of business risk prediction models will be obtained in the end, and when predicting risks, you can choose the business risk prediction model that meets expectations or has the best prediction effect. Please refer to Figure 3-1 , showing a training method for a business risk prediction model provided by some embodiments, including at least 301-305.

[0121] 301. Obtain first sample business data in a historical first period and second sample business data in a historical second period, wherein the second sample business data is one or more groups, and different groups of second sample business data correspond to different period lengths of the historical second period.

[0122] In the embodiments of the present disclosure, the first sample business data and the second sample business data can be obtained based on the acquired sample business data. In some examples, the sample business data within a preset period can be obtained. In some cases, the preset period is greater than the first historical period, that is, the first historical period is part of the preset period; in other cases, the preset period is the first historical period.

[0123] After acquiring the sample business data within a preset period, the first sample business data belonging to the historical first period can be acquired, and the second sample data belonging to multiple historical second periods can be acquired, and the second sample data of one historical second period is a group.

[0124] In this embodiment, the period lengths of the historical second periods corresponding to different groups of second sample business data are different. For example, the historical first period can be the past 2 months, and the historical second period is 2, namely the past 1 week and the past 1 month. Then, the first sample business data obtained is the business data within the past 2 months, and the second sample business data obtained is 2 groups: business data within the past 1 week and business data within the past 1 month.

[0125] Here, the business data within the past week can be obtained in any of the following ways: ①②③: ① Obtain the business data of any week within the past two months; ② Divide the business data of the past two months into one-week periods to obtain 8 one-week business data; ③ Slide the business data of the past two months (assuming that February has 60 days in total) according to a one-day time window to obtain 54 one-week business data.

[0126] ① Get business data for any week in the past 2 months

[0127] For example, the past two months are 2021 / 10 / 1-2021 / 11 / 30, so the business data of any one week in the past can be selected as the second sample business data, such as selecting the business data of 2021 / 10 / 1-2021 / 10 / 7 as a set of second sample business data.

[0128] ② Split the business data of the past two months into one-week periods to obtain eight one-week business data periods.

[0129] For example, if the past two months are 2021 / 10 / 1-2021 / 11 / 30, you can choose to divide the two months by week and obtain business data of eight periods as a set of second sample business data, that is, the second sample business data includes: business data of the first week from 2021 / 10 / 1-2021 / 10 / 7, business data of the second week from 2021 / 10 / 8-2021 / 10 / 14, ... (the period shifts by week) ..., business data of the eighth week from 2021 / 11 / 19-2021 / 11 / 25.

[0130] ③ Slide the business data of the past two months (assuming February has 60 days) according to a 1-day time window to obtain 54 1-week business data

[0131] For example, the past two months are 2021 / 10 / 1-2021 / 11 / 30, then you can slide the time window with 1 day to obtain 55 periods of business data as a set of second sample business data, that is, the second sample business data includes: business data of the first week from 2021 / 10 / 1-2021 / 10 / 7, business data of the second week from 2021 / 10 / 2-2021 / 10 / 8, ... (shifted by 1 day time window) ..., business data of the eighth week from 2021 / 11 / 23-2021 / 11 / 29, 2021 / 11 / 24-2021 / 11 / 30.

[0132] Similarly, the business data within the past month is obtained in the same way as the business data within the past week. The period is set from one week to one month, which can be any of the following ④⑤⑥: ④ Obtain the business data of any month within the past two months; ⑤ Split the business data of the past two months by one month to obtain business data for two months; ⑥ Slide the business data of the past two months (assuming that February has 60 days in total) according to a 1-day time window to obtain business data for 31 months.

[0133] ④ Obtain business data for any month in the past two months

[0134] For example, the past two months are 2021 / 10 / 1-2021 / 11 / 30, then the business data of any one month in the past can be selected as the second sample business data, such as selecting the business data of 2021 / 10 / 1-2021 / 10 / 31 as a set of second sample business data.

[0135] ⑤ Split the business data of the past two months into one month to obtain the business data of two months.

[0136] For example, the past two months are 2021 / 10 / 1-2021 / 11 / 30, then you can choose to split the two months by month and obtain the business data of two periods as a set of second sample business data, that is, the second sample business data includes: business data of the first month from 2021 / 10 / 1-2021 / 10 / 31,... (the period shifts by month)..., business data of the second month from 2021 / 11 / 1-2021 / 11 / 30.

[0137] ⑥ Slide the business data of the past two months (assuming February has 60 days) according to a 1-day time window to obtain 31 days of business data in January

[0138] For example, the past two months are 2021 / 10 / 1-2021 / 11 / 30, then you can slide the time window of 1 day to obtain 32 periods of business data as a set of second sample business data, that is, the second sample business data includes: business data of the first month from 2021 / 10 / 1-2021 / 10 / 30, business data of the second month from 2021 / 10 / 2-2021 / 10 / 31, ... (shifted by a time window of 1 day) ..., business data of the 32nd month from 2021 / 10 / 31-2021 / 11 / 29, 2021 / 11 / 1-2021 / 11 / 30.

[0139] 302. Obtain a first model corresponding to a first future period according to the first sample business data.

[0140] In the disclosed embodiment, the first preset model is trained with the first sample business data to obtain the first model corresponding to the first period in the future. The first model is used to predict the business risk of the business object in the first period in the future, such as predicting the resource usage risk of the business object in the next two months. At this time, the first model is trained based on a longer period (such as the first historical period) and does not incorporate the data performance of a shorter period (such as the second historical period), that is, it is not trained based on the data features of a shorter period (such as the second historical period), and the data features of a shorter period (such as the second historical period) have no effect on the prediction results of the first model.

[0141] In some examples, the first sample business data includes first object data and first resource data. The first resource data may include first resource expiration data, such as resource usage expiration time, resource return expiration time, etc. The first object data includes an identifier of a virtual resource requester, which is used to represent the virtual resource requester. The disclosed embodiment also refers to the virtual resource requester as a business object. The virtual resource requester may be a user account or an enterprise account. For an explanation of the first resource expiration data, please refer to the description of the resource expiration data in 204 above, which will not be repeated here.

[0142] Based on the above example, the process of acquiring the first model is: taking the first object data as input and the first resource expiration data as output, training the first preset model to obtain the first model corresponding to the first period in the future.

[0143] 303. According to each group of second sample business data, obtain a second model corresponding to each group of second sample business data.

[0144] In the disclosed embodiment, the second preset model is trained with each group of second sample business data to obtain the second model corresponding to each group of second sample business data. Each group of second sample business data will obtain a corresponding second model, and the number of second models is the same as the number of groups of second sample business data. For example: 2 groups of second sample business data are obtained, which are the second sample business data within 1 week of history and the second sample business data within 1 month of history, then two second models can be obtained, namely: second model 1 obtained based on the second sample business data within 1 week of history, and second model 2 obtained based on the second sample business data within 1 month of history.

[0145] The length of the cycle predicted by the second model is the same as the historical cycle corresponding to the second sample business data used by the model, that is, the corresponding second model obtained based on a set of second sample business data is used to predict the business risk of the business object in the future second cycle corresponding to the set of second sample business data. For example: the second model 1 obtained based on the second sample business data in the past week is used to predict the resource usage risk of the business object in the next week; the second model 2 obtained based on the second sample business data in the past month is used to predict the resource usage risk of the business object in the next month.

[0146] In some examples, the second sample business data includes second object data and second resource data, and the second resource data may include second resource expiration data, such as resource usage expiration time, resource return expiration time, etc. The second object data includes an identifier of a virtual resource requester, which is used to represent the virtual resource requester. The embodiment of the present disclosure also refers to the virtual resource requester as a business object, and the virtual resource requester can be a user account or an enterprise account. The explanation of the second resource expiration data can refer to the description of the resource expiration data in the aforementioned 204, which will not be repeated here.

[0147] Based on the above example, the process of acquiring the second model is: taking the second object data of the second sample business data of each historical second period as input and the second resource overdue data of the second sample business data of each historical second period as output, the second preset model corresponding to the second sample business data of each historical second period is trained to obtain the second model corresponding to each future second period. Based on this training process, according to a group of second sample business data of each historical second period, the corresponding second model will be trained and obtained, and the number of second models is the same as the number of historical second periods. Correspondingly, the number of second models is the same as the number of groups of second sample business data. The corresponding second model obtained based on a group of second sample business data of a historical second period is used to predict the business risk of the business object in the future second period corresponding to the group of second sample business data.

[0148] 304. Input the corresponding group of second sample business data into each second model, and predict the resource risk prediction data of the second sample business data of each group corresponding to the second period in the future.

[0149] For a second model, the second sample business data of the corresponding group is a group of second sample business data used to obtain the second model, and the second model is used to predict the resource risk prediction data of the second period in the future corresponding to the second sample business data of the corresponding group. For a group of second sample business data, the period length of the corresponding second period in the future is consistent with the historical second period of the group of second sample business data.

[0150] For example, the second model 1 is obtained based on the second sample business data in the past week, then the second sample business data of the corresponding group of the second model 1 is: the second sample business data in the past week, and the second model 1 is used to predict the resource usage risk of the business object in the next week.

[0151] For another example, the second model 2 is obtained based on the second sample business data within the past month, then the second sample business data of the corresponding group of the second model 2 is: the second sample business data within the past month, and the second model 2 is used to predict the resource usage risk of the business object in the next month.

[0152] 305. Based on the first sample business data, use the resource risk prediction data of the second period in the future corresponding to each group of second sample business data to train the first model respectively, and obtain a business risk prediction model corresponding to each group of second sample business data.

[0153] After obtaining resource risk prediction data corresponding to the second period in the future for each group of second sample business data, the first sample business data and a group of resource risk prediction data corresponding to the second period in the future can be used as training data to train the first model and obtain the corresponding business risk prediction model. The number of business risk prediction models finally obtained is the same as the number of groups of second sample business data.

[0154] For example, if the first sample business data is business data within 2 months of history, and there are two sets of second sample business data, namely, the second sample business data within 1 week of history and the second sample business data within 1 month of history, then two second models can be obtained, namely: second model 1 obtained based on the second sample business data within 1 week of history, and second model 2 obtained based on the second sample business data within 1 month of history. Then the aforementioned 304 can obtain two resource risk prediction data: resource usage risk A of the business object output by the second model 1 in the next week, and resource usage risk B of the business object output by the second model 2 in the next month.

[0155] Then 305 will execute the following process:

[0156] Using the first sample business data and resource usage risk A to train the first model, a corresponding business risk prediction model 1 is obtained;

[0157] The first model is trained using the first sample business data and resource usage risk B to obtain the corresponding business risk prediction model 2.

[0158] Business risk prediction models 1 and 2 are both used to predict the resource usage risk of business objects in the next two months. The difference is that the periods of the second sample business data used by business risk prediction models 1 and 2 are different. That is, business risk prediction model 1 is based on the business data performance of the business object in the past two months and the business data performance of the business object in the past week, while business risk prediction model 2 is based on the business data performance of the business object in the past two months and the business data performance of the business object in the past month.

[0159] By combining the short-term business data performances of multiple different historical second periods, the business risk prediction model is trained. This allows the business data performance of the business object in a longer period and the business data performances of multiple different shorter periods to be fully considered when predicting the long-term business risk of the business object in the first period in the future. Ultimately, multiple business risk prediction models are obtained. When predicting risks, the business risk prediction model with the best prediction effect that meets expectations can be selected, which can maximize the enhancing effect of short-term business data in long-term business risk prediction.

[0160] Based on the aforementioned disclosed embodiments, the inventors have also proposed an optional improvement scheme: combining the long-term business data performance of the first historical period, and using different combinations of short-term business data performance of the second historical period, to train a corresponding business risk prediction model. This scheme can fully consider the business data performance of the business object in a longer period, as well as the business data performance of multiple different combinations of shorter periods, and ultimately obtain multiple business risk prediction models. When predicting risks, you can choose the business risk prediction model with the best prediction effect that meets expectations, which can maximize the enhancement effect of each combination of short-term business data in long-term business risk prediction. Figure 3-1 Example, in one implementation, 305 can also obtain one or more prediction data combinations including at least one resource risk prediction data from the one or more resource risk prediction data obtained by 304, and for each prediction data combination, use the first sample business data and each resource risk prediction data in the prediction data combination to train a first model to obtain a business risk prediction model corresponding to each prediction data combination.

[0161] Specifically, Figure 3-2 As shown, 305 includes 3051-3052.

[0162] 3051. Obtain one or more prediction data combinations from one or more resource risk prediction data output by each second model. The prediction data combination includes resource risk prediction data output by at least one second model.

[0163] In some examples, the prediction data combination is: a full combination of resource risk prediction data output by at least any second model.

[0164] For example, there are second model 1 and second model 2, where second model 1 is based on the second sample business data in the past week, and second model 2 is based on the second sample business data in the past month. Second model 1 outputs resource risk prediction data 1, and second model 2 outputs resource risk prediction data 2, then the prediction data combination includes:

[0165] {Resource risk prediction data 1},

[0166] {Resource risk prediction data 2},

[0167] {Resource risk prediction data 1, resource risk prediction data 2}.

[0168] For another example, there are second models 1, 2, and 3, and second model 1 outputs resource risk prediction data 1, second model 2 outputs resource risk prediction data 2, and second model 3 outputs resource risk prediction data 3. Then the prediction data combination includes:

[0169] {Resource risk prediction data 1},

[0170] {Resource risk prediction data 2},

[0171] {Resource risk prediction data 3},

[0172] {Resource risk prediction data 1, resource risk prediction data 2},

[0173] {Resource risk prediction data 2, resource risk prediction data 3},

[0174] {Resource risk prediction data 1, resource risk prediction data 3},

[0175] {Resource risk prediction data 1, resource risk prediction data 2, resource risk prediction data 3}.

[0176] 3052. For each prediction data combination, a first model is trained using the first sample business data and each resource risk prediction data in the prediction data combination to obtain a business risk prediction model corresponding to each prediction data combination.

[0177] After obtaining the resource risk prediction data of each prediction data combination, the first sample business data and a prediction data combination can be used as training data to train the first model and obtain the corresponding business risk prediction model. The number of business risk prediction models finally obtained is the same as the number of combinations of prediction data combinations.

[0178] For example, there are second model 1 and second model 2, where second model 1 is based on the second sample business data in the past week, and second model 2 is based on the second sample business data in the past month. Second model 1 outputs resource risk prediction data 1, and second model 2 outputs resource risk prediction data 2, then the prediction data combination includes:

[0179] Prediction data combination 1-{resource risk prediction data 1},

[0180] Prediction data combination 2-{resource risk prediction data 2},

[0181] Prediction data combination 3-{resource risk prediction data 1, resource risk prediction data 2};

[0182] Then 305 will execute the following process:

[0183] Using the first sample business data and the prediction data combination 1 to train the first model, to obtain the corresponding business risk prediction model 1;

[0184] Using the first sample business data and the prediction data combination 2 to train the first model, and obtain the corresponding business risk prediction model 2;

[0185] The first model is trained using the first sample business data and the prediction data combination 3 to obtain the corresponding business risk prediction model 3.

[0186] Business risk prediction models 1, 2, and 3 are all used to predict the resource usage risk of business objects in the next two months. The difference is that the period combinations of the second sample business data used in the predictions of business risk prediction models 1, 2, and 3 are different. That is, business risk prediction model 1 is based on the business data performance of the business object in the past two months and the business data performance of the business object in the past week, business risk prediction model 2 is based on the business data performance of the business object in the past two months and the business data performance of the business object in the past month, and business risk prediction model 3 is based on the business data performance of the business object in the past two months, the business data performance of the business object in the past week, and the business data performance of the business object in the past month.

[0187] By combining the short-term business data performance of the second period of history in multiple different combinations, the business risk prediction model is trained. When predicting the long-term business risk of the business object in the first period in the future, the business data performance of the business object in a longer period and the comprehensive performance of business data in multiple different shorter periods can be fully considered. Finally, multiple business risk prediction models are obtained. When predicting risks, both the long-term business data performance of the business object and the combination of different short-term business data performance can be considered. The final effect is more objective and accurate. This improved scheme can maximize the enhancement effect of different short-term business data among multiple short-term business data and the combination of short-term business data in long-term business risk prediction, and maximize the enhancement effect of short-term business data in long-term business risk prediction.

[0188] In some of the aforementioned embodiments, multiple business risk prediction models are finally obtained, and the embodiments of the present disclosure can also select the business risk prediction model whose prediction effect meets expectations or has the best prediction effect. Figure 4 As shown, the embodiment of the present disclosure further includes 401-402.

[0189] 401. Obtain effect parameters of one or more business risk prediction models.

[0190] In some embodiments, if the business risk prediction model is obtained based on training of a combination of prediction data, effect parameters of the business risk prediction model corresponding to one or more combinations of prediction data may be obtained.

[0191] In other embodiments, if the business risk prediction model is obtained by training based on each group of second sample business data, the effect parameters of the corresponding one or more business risk prediction models can be obtained.

[0192] 402. Select a target business risk prediction model whose effect parameters meet preset conditions and a target second model corresponding to the target business prediction model for business risk prediction, wherein the target business prediction model and the second sample business data used to obtain the target second model are the same group.

[0193] In some embodiments, the effect parameter of the business risk prediction model can be an AUC value or a KS value. We can select the business risk prediction model with the top N AUC values ​​or KS values ​​as the target business risk prediction model, and select the target second model corresponding to the target business prediction model, and use the target second model and the target business risk prediction model to predict the business risk of the business object in the first period in the future. Wherein, N is a positive integer. In one example, the business risk prediction model with the first KS value can be selected as the target business risk prediction model, and the target second model corresponding to the target business prediction model can be selected. In another example, any one of the top 2 KS values ​​can be selected as the target business risk prediction model, and the target second model corresponding to the target business prediction model can be selected. In other examples, any one of the top 2 ACU values ​​can be selected as the target business risk prediction model, and the target second model corresponding to the target business prediction model can be selected.

[0194] After verification testing, the business risk prediction model trained with the second sample business data has better prediction results than the business risk prediction model trained without the second sample business data. In a test example, the AUC value and KS value of the business risk prediction model trained with the second sample business data are 1% higher than the AUC value and KS value of the business risk prediction model trained without the second sample business data.

[0195] The following example is used to compare and illustrate that the solution provided by the embodiment of the present disclosure can improve the prediction ability and effect of the model.

[0196] (1) A first model, namely, business risk prediction model A, is obtained based on training of the first sample business data.

[0197] The business risk prediction model A is a traditional business method. The business risk prediction model A is used to predict the first sample business data, and its AUC value is 61.19% and the KS value is 16.74%.

[0198] (2) Based on two sets of second sample business data, two second models are trained. Among them, the second model 1 is obtained based on the second sample business data in the past week, and the second model 2 is obtained based on the second sample business data in the past month. The second model 1 outputs resource risk prediction data 1, and the second model 2 outputs resource risk prediction data 2. Then the prediction data combination includes:

[0199] Prediction data combination 1-{resource risk prediction data 1},

[0200] Prediction data combination 2-{resource risk prediction data 2},

[0201] Prediction data combination 3-{resource risk prediction data 1, resource risk prediction data 2};

[0202] The following process will be executed during training:

[0203] Using the first sample business data and the prediction data combination 1 to train the first model, to obtain the corresponding business risk prediction model 1;

[0204] Using the first sample business data and the prediction data combination 2 to train the first model, and obtain the corresponding business risk prediction model 2;

[0205] The first model is trained using the first sample business data and the prediction data combination 3 to obtain the corresponding business risk prediction model 3.

[0206] The above-mentioned business risk prediction models 1, 2, and 3 are used to predict the first sample business data respectively, and their AUC values ​​and KS values ​​compared with business risk prediction model A are shown in the following table.

[0207] Business risk prediction model AUC value KS value Business Risk Prediction Model A 61.19％ 16.74％ Business risk prediction model 1 62.60％ 18.33％ Business risk prediction model 2 62.88％ 18.84％ Business risk prediction model 3 63.27％ 19.56％

[0208] See also Figure 5 The embodiment of the present disclosure also provides a business risk prediction method, which is applied to the second model and the business risk prediction model obtained by training in the above embodiment, such as Figure 5 As shown, the business risk prediction method includes at least 501-503.

[0209] 501. Obtain business data of the prediction object.

[0210] In some embodiments, the business data of the object includes object data and resource data. The object data includes an identifier of the predicted object, which may be a user account or a corporate account. The resource data may include resource usage data, such as resource overtime data (such as resource usage overtime time, resource return overtime time), resource return value, resource transfer data, etc. Among them, the resource transfer data refers to data for transferring resources to other accounts, such as the account ID of other accounts, the resource value of the transferred resources, etc. The resource overtime data is the difference data between the transfer time and the agreed time of the virtual resource transferred from the predicted object to the resource provider. In one example, the resource overtime data may refer to the resource return overtime time. Among them, the agreed time is the time agreed upon by the predicted object and the resource provider to transfer the virtual resource. In some cases, the agreed time is the agreed time for the predicted object to transfer the virtual resource to the resource provider. For example: the time agreed upon by the forecast object and the resource provider to transfer virtual resources is 2021 / 10 / 1. If the actual time when the forecast object transfers the virtual resources to the resource provider is 2021 / 10 / 20, the difference between the transfer time and the agreed time is 19 days, and the resource overdue data is 19 days, that is, the resource return overdue time is 19 days.

[0211] It should be understood that if the forecast object has not transferred the virtual resources, the resource overdue data will accumulate with the time when the forecast object has not transferred the virtual resources. For example: the time agreed by the forecast object and the resource provider to transfer the virtual resources is 2021 / 10 / 1. If the forecast object has not transferred the virtual resources to the resource provider on 2021 / 10 / 20, the difference between the transfer time and the agreed time is 19 days, and the resource overdue data is 19 days; when the time reaches 2021 / 10 / 21, the difference between the transfer time and the agreed time is 20 days, and the resource overdue data is 20 days, that is, the resource return overdue time is 20 days.

[0212] 502. Input the business data into the second model to obtain the resource risk prediction data of the prediction object in the second period in the future output by the second model, wherein the second model is obtained based on the second sample business data in the second historical period.

[0213] The training process of the second model can be found in the previous Figure 1 -to Figure 4 The process shown will not be repeated here.

[0214] 503. Input the business data and the resource risk prediction data of the prediction object in the second period in the future into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future, wherein the target business risk prediction model is obtained based on the resource risk prediction data of the second period in the future corresponding to the first sample business data and the second sample business data in the first historical period.

[0215] For the training process of the target business risk prediction model, please refer to the previous Figure 1 -to Figure 4 The process shown in the business risk prediction model will not be repeated here.

[0216] When predicting business risks (such as long-term business risks), the business risk prediction model of the disclosed embodiment will combine the short-term resource risk prediction data of the business object in the second future period, and fully consider the business data performance of the prediction object in a longer historical period (i.e., the first historical period) and the business data performance in a shorter historical period (i.e., the second historical period) (such as resource risk prediction data). This makes the prediction results of long-term business risks output by the business risk prediction model more fully reflect the short-term and long-term performance of the business object, and can effectively improve the prediction accuracy and effect of the long-term business risk of the business object.

[0217] In some optional implementations, the second model has one or more, and the second model is used to predict the second sample data used by the second model corresponding to the resource risk prediction data of the second period in the future, and the second sample business data used by different second models are obtained as different groups, and the period lengths of the historical second period corresponding to different groups of second sample business data are different. Then: the business risk prediction model has one or more, and different business risk prediction models are obtained based on the resource risk prediction data in different prediction data combinations and the first sample business data in the historical first period. The specific description of the second model and the business risk prediction model can be as mentioned above Figure 3-1 , 3-2 The embodiments mentioned are shown in the figure and will not be described in detail here.

[0218] Combined with the above Figure 4 In the embodiment shown, in some optional implementations, the target business risk prediction model is a model whose effect parameters in the business risk prediction model meet the preset conditions, then:

[0219] Inputting business data into the second model to obtain resource risk prediction data of the prediction object output by the second model in the second period in the future, including: inputting business data into one or more target second models to obtain target resource risk prediction data of the prediction object output by each target second model in the second period in the future, wherein the target business prediction model and the second sample business data used by the target second model are the same group;

[0220] Input business data and resource risk prediction data of the prediction object in the second period in the future into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future, including: input business data and target resource risk prediction data output by each target second model into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future.

[0221] See also Figure 6 , as a response to the above Figure 2-4 The present disclosure provides an embodiment of a training device for a business risk prediction model. Figure 2-4 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices. In the embodiment of the present disclosure, each module of the training device of the business risk prediction model can be set in one electronic device, or can be separately set in different electronic devices, which is not limited by the embodiment of the present disclosure.

[0222] like Figure 6 As shown, an embodiment of the present disclosure provides a training device for a business risk prediction model, including: a first acquisition unit 601, a first training unit 602, an input unit 603 and a second training unit 604.

[0223] The first acquisition unit 601 is configured to acquire first sample business data within a first historical period and second sample business data within a second historical period, wherein the first historical period includes the second historical period and the period length of the first historical period is greater than the period length of the second historical period.

[0224] The first training unit 602 is configured to obtain a first model corresponding to a first future period and a second model corresponding to a second future period according to the first sample business data and the second sample business data, respectively.

[0225] The input unit 603 is configured to input second sample business data into the second model, and predict resource risk prediction data corresponding to the second sample business data in a second future period.

[0226] The second training unit 604 is configured to train the first model according to the first sample business data and the second sample business data corresponding to the resource risk prediction data of the second period in the future, so as to obtain a business risk prediction model.

[0227] In some optional implementations, the second sample business data is one group or multiple groups, and the period lengths of the historical second periods corresponding to different groups of second sample business data are different.

[0228] In some optional implementations, the first training unit 602 is configured to:

[0229] Obtaining a first model corresponding to a first future period according to the first sample business data;

[0230] According to each group of second sample business data, a second model corresponding to each group of second sample business data is obtained respectively.

[0231] In some optional implementations, the first sample business data includes first object data and first resource expiration data, and the second sample business data includes second object data and second resource expiration data.

[0232] In some optional implementations, the first training unit 602 is configured to:

[0233] Taking the first object data as input and the first resource expiration data as output, training the first preset model to obtain a first model corresponding to the first future period;

[0234] Taking the second object data of the second sample business data of each historical second period as input and the second resource expiration data of the second sample business data of each historical second period as output, the second preset model corresponding to the second sample business data of each historical second period is trained to obtain the second model corresponding to each future second period.

[0235] In some optional implementations, the second training unit 604 is configured to:

[0236] Obtain one or more prediction data combinations from one or more resource risk prediction data output by each second model, wherein the prediction data combination includes resource risk prediction data output by at least one second model;

[0237] For each prediction data combination, the first model is trained using the first sample business data and each resource risk prediction data in the prediction data combination to obtain a business risk prediction model corresponding to each prediction data combination.

[0238] In some optional implementations, the training device for the business risk prediction model further includes:

[0239] A verification unit configured to obtain effect parameters of a business risk prediction model corresponding to one or more prediction data combinations;

[0240] The selection unit is configured to select a target business risk prediction model whose effect parameters meet preset conditions and a target second model corresponding to the target business prediction model for business risk prediction, wherein the target business prediction model and the second sample business data used to obtain the target second model are the same group.

[0241] The implementation details and technical effects of each unit in the training device for the business risk prediction model provided by the embodiments of the present disclosure can be referred to the description of other embodiments in the present disclosure and will not be repeated here.

[0242] See also Figure 7 , as a response to the above Figure 5 The present disclosure provides an embodiment of a business risk prediction device. Figure 5 The method embodiment shown corresponds to Figure 7 As shown, the embodiment of the present disclosure provides a business risk prediction device, including: a second input unit 701, a second model 702, and a target business risk prediction model 703. In the embodiment of the present disclosure, each module of the business risk prediction device can be set in one electronic device, or can be separately set in different electronic devices, which is not limited by the embodiment of the present disclosure.

[0243] The second input unit 701 is configured to obtain business data of a prediction object;

[0244] The second model 702 is configured to receive business data and output resource risk prediction data of the prediction object in the second future period, wherein the second model is obtained based on second sample business data in the second historical period;

[0245] The target business risk prediction model 703 is configured to receive business data and resource risk prediction data of the prediction object in the second period in the future to the target business risk prediction model, and obtain the resource risk prediction data of the prediction object in the first period in the future, wherein the target business risk prediction model is based on the first sample business data and the second sample business data in the first historical period corresponding to the resource risk prediction data of the second period in the future.

[0246] In some optional implementations, the second model 702 has one or more, and the second model is used to predict the resource risk prediction data of the second period in the future corresponding to the second sample data used by the second model, and the second sample business data used by different second models are obtained as different groups, and the period lengths of the historical second periods corresponding to the second sample business data of different groups are different, then:

[0247] There are one or more business risk prediction models, and different business risk prediction models are obtained based on the risk prediction data of each resource in different prediction data combinations and the first sample business data in the first historical period.

[0248] In some optional implementations, the target business risk prediction model 703 is a model in which the effect parameters of the business risk prediction model meet the preset conditions, the second model 702 includes one or more target second models, and the target business prediction model 703 is obtained and the second sample business data used to obtain the target second model is the same group;

[0249] The target second model is configured to receive input business data and output target resource risk prediction data of the prediction object in the second period in the future;

[0250] The target business risk prediction model 703 is configured to receive business data and target resource risk prediction data output by each target second model, and output resource risk prediction data of the prediction object in the first period in the future.

[0251] The implementation details and technical effects of each unit in the business risk prediction device provided by the embodiments of the present disclosure can be referred to the description of other embodiments in the present disclosure, and will not be repeated here.

[0252] The present disclosure also provides an electronic device, including:

[0253] one or more processors;

[0254] a storage device having one or more programs stored thereon,

[0255] When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method as described in any one of the aforementioned embodiments.

[0256] An embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by one or more processors, implements the method as described in any one of the aforementioned embodiments.

[0257] Reference below Figure 8 , which shows a schematic diagram of the structure of a computer system 800 of a terminal device suitable for implementing the present disclosure. Figure 8 The computer system 800 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0258] like Figure 8 As shown, the computer system 800 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage device 808 to a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the computer system 800 are also stored. The processing device 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 804 is also connected to the bus 804.

[0259] Typically, the following devices may be connected to the I / O interface 804: input devices 806 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, etc.; output devices 807 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 808 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 809. The communication devices 809 may allow the computer system 800 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 4 The computer system 800 of an electronic device having various devices is shown, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0260] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device 809, or installed from a storage device 808, or installed from a ROM 802. When the computer program is executed by the processing device 801, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0261] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0262] The computer-readable medium may be included in the electronic device, or may exist independently without being installed in the electronic device.

[0263] The computer readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device implements the following Figure 2 The illustrated embodiment and its optional implementations illustrate a method for determining application information trends.

[0264] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0265] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0266] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware. The name of a unit does not limit the unit itself in some cases. For example, the acquisition unit may also be described as a "unit for resource application behavior information".

[0267] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

Claims

1. A training method for a business risk prediction model, characterized in that: The method comprises: Acquire first sample business data within a first historical period and second sample business data within a second historical period, wherein the first historical period includes the second historical period, and the period length of the first historical period is greater than the period length of the second historical period; According to the first sample business data and the second sample business data, respectively obtain a first model corresponding to a first future period and a second model corresponding to a second future period; Inputting the second sample business data into the second model, and predicting resource risk prediction data corresponding to the second sample business data in the second future period; The first model is trained based on the resource risk prediction data of the second future period corresponding to the first sample business data and the second sample business data to obtain a business risk prediction model.

2. The training method according to claim 1, characterized in that: The second sample business data is one group or multiple groups, and the period lengths of the historical second periods corresponding to different groups of second sample business data are different.

3. The training method according to claim 2, characterized in that: The obtaining, according to the first sample service data and the second sample service data, a first model corresponding to a first period in the future and a second model corresponding to a second period in the future respectively, comprises: Obtaining a first model corresponding to the first future period according to the first sample business data; According to each group of the second sample business data, a second model corresponding to each group of the second sample business data is obtained respectively.

4. The training method according to any one of claims 1 to 3, characterized in that: The first sample service data includes first object data and first resource expiration data, and the second sample service data includes second object data and second resource expiration data.

5. The training method according to claim 4, characterized in that: The obtaining, according to the first sample service data and the second sample service data, a first model corresponding to a first period in the future and a second model corresponding to a second period in the future respectively, comprises: Taking the first object data as input and the first resource expiration data as output, training a first preset model to obtain a first model corresponding to the first future period; Taking the second object data of the second sample business data of each historical second period as input and the second resource expiration data of the second sample business data of each historical second period as output, the second preset model corresponding to the second sample business data of each historical second period is trained to obtain the second model corresponding to each future second period.

6. The training method according to claim 3, characterized in that: The step of training the first model to obtain a business risk prediction model based on the resource risk prediction data of the second future period corresponding to the first sample business data and the second sample business data includes: Acquire one or more prediction data combinations from one or more resource risk prediction data output by each second model, wherein the prediction data combination includes resource risk prediction data output by at least one second model; For each of the prediction data combinations, a first model is trained using the first sample business data and each resource risk prediction data in the prediction data combination to obtain a business risk prediction model corresponding to each prediction data combination.

7. The training method according to claim 6, characterized in that: The method further comprises: Obtaining effect parameters of the business risk prediction model corresponding to the one or more prediction data combinations; A target business risk prediction model whose effect parameters meet preset conditions and a target second model corresponding to the target business prediction model are selected for business risk prediction, wherein the target business prediction model and the second sample business data used to obtain the target second model are the same group.

8. A business risk prediction method, characterized in that: The method comprises: Obtain business data of the forecast object; Input the business data into the second model to obtain resource risk prediction data of the prediction object in the second future period output by the second model, wherein the second model is obtained based on the second sample business data in the second historical period; Input the business data and the resource risk prediction data of the prediction object in the second period in the future into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future, wherein the target business risk prediction model is obtained based on the first sample business data in the historical first period and the second sample business data corresponding to the resource risk prediction data of the second period in the future.

9. The prediction method according to claim 8, characterized in that: The second model has one or more, and the second model is used to predict the resource risk prediction data of the second sample data used by the second model corresponding to the future second period, and the second sample business data used by different second models are obtained as different groups, and the period lengths of the historical second periods corresponding to the second sample business data of different groups are different, then: There are one or more business risk prediction models, and different business risk prediction models are obtained based on the risk prediction data of each resource in different prediction data combinations and the first sample business data in the first historical period.

10. The prediction method according to claim 9, characterized in that: The target business risk prediction model is a model whose effect parameters in the business risk prediction model meet the preset conditions. The step of inputting the business data into the second model to obtain the resource risk prediction data of the prediction object in the second period in the future outputted by the second model comprises: inputting the business data into one or more target second models to obtain the target resource risk prediction data of the prediction object in the second period in the future outputted by each target second model, wherein the target business prediction model and the second sample business data used to obtain the target second model are the same group; The step of inputting the business data and the resource risk prediction data of the prediction object in the second period in the future into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future includes: inputting the business data and the target resource risk prediction data output by each target second model into the target business risk prediction model to obtain the resource risk prediction data of the prediction object in the first period in the future.

11. A training device for a business risk prediction model, characterized in that: include: A first acquisition unit is configured to acquire first sample business data within a first historical period and second sample business data within a second historical period, wherein the first historical period includes the second historical period, and the period length of the first historical period is greater than the period length of the second historical period; A first training unit is configured to obtain a first model corresponding to a first future period and a second model corresponding to a second future period respectively according to the first sample business data and the second sample business data; An input unit, configured to input the second sample business data into the second model, and predict resource risk prediction data corresponding to the second sample business data in the second future period; The second training unit is configured to train the first model according to the resource risk prediction data of the first sample business data and the second sample business data corresponding to the second future period, so as to obtain a business risk prediction model.

12. A business risk prediction device, characterized in that: include: A second input unit is configured to obtain business data of a prediction object; A second model is configured to receive the business data and output resource risk prediction data of the prediction object in the second future period, wherein the second model is obtained based on second sample business data in the second historical period; The target business risk prediction model is configured to receive the business data and the resource risk prediction data of the prediction object in the second future period to the target business risk prediction model, and obtain the resource risk prediction data of the prediction object in the first future period, wherein the target business risk prediction model is based on the first sample business data in the historical first period and the second sample business data corresponding to the resource risk prediction data of the second future period.

13. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 10.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by one or more processors, the method according to any one of claims 1 to 10 is implemented.

Images