Digital asset risk prediction model construction method and device, equipment and storage medium

Abstract

The present disclosure provides a digital asset risk prediction model construction method, device and equipment and a storage medium. The method comprises: obtaining a modeling sample set, generating a trained digital asset risk prediction base model according to digital asset application data and real risk label values of each modeling sample; obtaining target prediction risk label values of each modeling sample and risk difference values between the real risk label values and the target prediction risk label values according to the trained digital asset risk prediction base model; generating a trained digital asset risk difference value prediction model according to digital asset correlation data of each modeling sample and the target prediction risk label values of each modeling sample and the risk difference values between the real risk label values and the target prediction risk label values; and generating a target digital asset risk prediction model according to the trained digital asset risk prediction base model and the trained digital asset risk difference value prediction model.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and specifically to a method, apparatus, device, and storage medium for constructing a digital asset risk prediction model. Background Technology

[0002] The main purpose of digital asset risk prediction models is to assess the credit risk of digital asset applicants so that digital asset service providers can make reasonable digital asset application decisions.

[0003] In related technologies, the technical solution for digital asset service providers to build digital asset risk prediction models is as follows: A basic model is typically built using the service provider's own data (internal data owned by the service provider), such as application data from various digital asset service providers. An external data model is also built using external data (data provided by third-party data providers that can be further used to assess the credit risk of digital asset applicants), such as the applicants' consumption records. Then, a logistic regression algorithm or a weighted summation method is used to fuse the basic model and the external data model to obtain the final digital asset risk prediction model.

[0004] However, the aforementioned basic model and external data model are constructed independently based on proprietary data and external data, respectively. External data serves as supplementary data to further assess the credit risk of digital asset applicants using proprietary data, but it cannot effectively compensate for the inaccurate predictions of the basic model, thus affecting the final performance of the digital asset risk prediction model.

[0005] Therefore, it is necessary to propose a method for constructing a digital asset risk prediction model to solve at least one of the above-mentioned technical problems. Summary of the Invention

[0006] This disclosure provides a method, apparatus, device, and storage medium for constructing a digital asset risk prediction model. When constructing a target digital asset risk prediction model, digital asset service providers can obtain the risk difference between the real risk label value and the target predicted risk label value for each modeling sample based on the real risk label value and the basic digital asset risk prediction model. Then, based on the risk difference and related digital asset data, a digital asset risk difference prediction model is constructed. Finally, the basic digital asset risk prediction model and the digital asset risk difference prediction model are merged to obtain the target digital asset risk prediction model. This allows the digital asset risk difference prediction model to specifically learn from the data that the basic digital asset risk prediction model cannot accurately predict, enabling it to better complement the basic model. This results in a more accurate prediction of the credit risk of digital asset applicants, maximizing the gain of the digital asset risk difference prediction model on top of the basic model and improving the accuracy of the target digital asset risk prediction model.

[0007] Firstly, this disclosure provides a method for constructing a digital asset risk prediction model, including:

[0008] Obtain a modeling sample set, wherein the modeling sample set includes digital asset application data, real risk label values ​​and digital asset association data for each modeling sample;

[0009] Based on the digital asset application data and the real risk label value of each modeling sample, a trained basic model for digital asset risk prediction is generated.

[0010] Based on the trained digital asset risk prediction model, the target predicted risk label value of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value are obtained.

[0011] Based on the digital asset association data of each modeling sample and the risk difference between the real risk label value and the target predicted risk label value, a trained digital asset risk difference prediction model is generated.

[0012] Based on the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model, a target digital asset risk prediction model is generated.

[0013] In some optional implementations, generating a trained digital asset risk prediction base model based on the digital asset application data and the real risk label values ​​of each of the modeling samples includes:

[0014] The modeling samples in the modeling sample set are divided into a training sample set and a test sample set;

[0015] The digital asset risk prediction basic model to be trained is trained based on the digital asset application data and the real risk label value of each training sample in the training sample set, and the digital asset risk prediction basic model to be tested is generated.

[0016] Based on the digital asset application data and the real risk label value of each test sample in the test sample set, the digital asset risk prediction basic model to be tested is tested to obtain the first test result of the digital asset risk prediction basic model to be tested.

[0017] When the first test result is determined to be qualified, the tested digital asset risk prediction basic model is determined as the trained digital asset risk prediction basic model.

[0018] In some optional implementations, training the digital asset risk prediction basic model to be trained based on the digital asset application data and the real risk label value of each training sample in the training sample set, to generate the digital asset risk prediction basic model to be tested, includes:

[0019] The digital asset risk prediction model to be trained is trained based on the digital asset application data and the real risk label value of each training sample in the training sample set, so as to obtain the first predicted risk label value of each training sample.

[0020] The first model parameters of the digital asset risk prediction base model to be trained are adjusted based on the difference between the first predicted risk label value and the actual risk label value of each training sample.

[0021] The digital asset risk prediction base model to be trained after adjusting the parameters of the first model is determined as the digital asset risk prediction base model to be tested, and the second predicted risk label value of each training sample output by the digital asset risk prediction base model to be tested is recorded.

[0022] In some optional implementations, the step of testing the digital asset risk prediction basic model to be tested based on the digital asset application data and the real risk label value of each test sample in the test sample set, and obtaining a first test result of the digital asset risk prediction basic model to be tested, includes:

[0023] Input the digital asset application data of each test sample in the test sample set into the digital asset risk prediction basic model to be tested, and output the second predicted risk label value of each test sample;

[0024] Based on the second predicted risk label value and the actual risk label value of each training sample, and based on the second predicted risk label value and the actual risk label value of each test sample, the first test result of the digital asset risk prediction basic model to be tested is obtained.

[0025] In some optional implementations, obtaining the target predicted risk label value and the risk difference between the actual risk label value and the target predicted risk label value for each modeled sample based on the trained digital asset risk prediction base model includes:

[0026] Based on the trained digital asset risk prediction model, the target predicted risk label value of each training sample and the target predicted risk label value of each test sample are obtained.

[0027] Calculate a first difference between the true risk label value and the target predicted risk label value for each of the training samples, and calculate a second difference between the true risk label value and the target predicted risk label value for each of the test samples;

[0028] The risk difference for each modeling sample is obtained based on the first difference and the second difference.

[0029] In some optional implementations, generating a trained digital asset risk difference prediction model based on the digital asset association data of each of the modeling samples and the risk difference between the actual risk label value and the target predicted risk label value includes:

[0030] The digital asset risk difference prediction model to be trained is trained based on the digital asset association data and risk difference of each training sample in the training sample set, and the digital asset risk difference prediction model to be tested is generated.

[0031] Based on the digital asset association data and risk difference of each test sample in the test sample set, the digital asset risk difference prediction model to be tested is tested to obtain the second test result of the digital asset risk difference prediction model to be tested.

[0032] When the second test result is determined to be qualified, the tested digital asset risk difference prediction model is determined as the trained digital asset risk difference prediction model.

[0033] In some optional implementations, training the digital asset risk difference prediction model to be trained based on the digital asset association data and the risk difference of each training sample in the training sample set, and generating the digital asset risk difference prediction model to be tested, includes:

[0034] The digital asset risk difference prediction model to be trained is trained based on the digital asset association data and risk difference of each training sample in the training sample set, to obtain the first predicted risk difference of each training sample.

[0035] The second model parameters of the digital asset risk difference prediction model to be trained are adjusted based on the difference between the first predicted risk difference and the risk difference of each training sample.

[0036] The digital asset risk difference prediction model to be trained after adjusting the second model parameters is determined as the digital asset risk difference prediction model to be tested, and the second predicted risk difference of each training sample output by the digital asset risk difference prediction model to be tested is recorded.

[0037] In some optional implementations, the step of testing the digital asset risk difference prediction model to be tested based on the digital asset association data and the risk difference of each test sample in the test sample set, and obtaining a second test result of the digital asset risk difference prediction model to be tested, includes:

[0038] Input the digital asset association data of each test sample in the test sample set into the digital asset risk difference prediction model to be tested, and output the second predicted risk difference of each test sample.

[0039] Based on the second predicted risk difference and the risk difference of each training sample, and based on the second predicted risk difference and the risk difference of each test sample, the second test result of the digital asset risk difference prediction model to be tested is obtained.

[0040] Secondly, this disclosure provides a method for predicting digital asset risks, including:

[0041] Obtain the digital asset application data and associated digital asset data of the target digital asset applicant;

[0042] The target digital asset application data and associated data of the target digital asset applicant are input into the target digital asset risk prediction model, and the target digital asset risk prediction value of the target digital asset applicant is output. The target digital asset risk prediction model includes a basic digital asset risk prediction model and a digital asset risk difference prediction model. The basic digital asset risk prediction model is used to obtain the target predicted risk label value of the target digital asset applicant based on the digital asset application data of the target digital asset applicant. The digital asset risk difference prediction model is used to obtain the target predicted risk difference value of the target digital asset applicant based on the associated data of the target digital asset applicant. The target digital asset risk prediction model is used to obtain the target predicted risk value of the target digital asset applicant based on the target predicted risk label value and the target predicted risk difference value. The basic digital asset risk prediction model, the digital asset risk difference prediction model, and the target digital asset risk prediction model are obtained through training as described in the first aspect of this disclosure.

[0043] Thirdly, this disclosure provides a device for constructing a digital asset risk prediction model, comprising:

[0044] The sample set acquisition unit is used to acquire the modeling sample set, wherein the modeling sample set includes digital asset application data, real risk label values ​​and digital asset association data of each modeling sample;

[0045] The first generation unit is used to generate a trained digital asset risk prediction basic model based on the digital asset application data and the real risk label value of each modeling sample.

[0046] The second generation unit is used to obtain the target predicted risk label value of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value based on the trained digital asset risk prediction basic model.

[0047] The third generation unit is used to generate a trained digital asset risk difference prediction model based on the digital asset association data of each modeling sample and the risk difference between the real risk label value and the target predicted risk label value.

[0048] The fourth generation unit is used to generate a target digital asset risk prediction model based on the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model.

[0049] Fourthly, this disclosure provides a digital asset risk prediction device, comprising:

[0050] The data acquisition unit is used to acquire the digital asset application data and digital asset-related data of the target digital asset applicant.

[0051] A risk prediction value output unit is used to input the digital asset application data and the associated digital asset data of the target digital asset applicant into a target digital asset risk prediction model, and output the target predicted risk value of the target digital asset applicant. The target digital asset risk prediction model includes a basic digital asset risk prediction model and a digital asset risk difference prediction model. The basic digital asset risk prediction model is used to obtain the target predicted risk label value of the target digital asset applicant based on the digital asset application data. The digital asset risk difference prediction model is used to obtain the target predicted risk difference value of the target digital asset applicant based on the associated digital asset data. The target digital asset risk prediction model is used to obtain the target predicted risk value of the target digital asset applicant based on the target predicted risk label value and the target predicted risk difference value. The basic digital asset risk prediction model, the digital asset risk difference prediction model, and the target digital asset risk prediction model are obtained through training using an embodiment of the second aspect of this disclosure.

[0052] Fifthly, this disclosure provides an electronic device, including:

[0053] One or more processors;

[0054] Storage device, on which one or more programs are stored,

[0055] When the above-described one or more programs are executed by the above-described one or more processors, the above-described one or more processors implement the method as described in any embodiment of the first or second aspect of this disclosure.

[0056] In a sixth aspect, this 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 described in any embodiment of the first or second aspect of this disclosure.

[0057] In a seventh aspect, this disclosure provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the method described in any embodiment of the first or second aspect of this disclosure.

[0058] The digital asset risk prediction model construction method, apparatus, device, and storage medium provided in the embodiments of this disclosure firstly acquire a modeling sample set, wherein the modeling sample set includes digital asset application data, real risk label values, and digital asset association data for each modeling sample. Based on the digital asset application data and real risk label values ​​of each modeling sample, a trained digital asset risk prediction base model is generated. Then, based on the trained digital asset risk prediction base model, the target predicted risk label value and the risk difference between the real risk label value and the target predicted risk label value for each modeling sample are obtained. Next, based on the digital asset association data of each modeling sample, the target predicted risk label value of each modeling sample, and the risk difference between the real risk label value and the target predicted risk label value, a trained digital asset risk difference prediction model is generated. Finally, based on the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model, a target digital asset risk prediction model is generated. This disclosure, in constructing a digital asset risk difference prediction model, obtains the risk difference between the actual risk label value and the target predicted risk label value for each modeling sample based on the actual risk label value and the basic digital asset risk prediction model. Then, based on the risk difference and related digital asset data, a digital asset risk difference prediction model is constructed. Finally, the basic digital asset risk prediction model and the digital asset risk difference prediction model are merged to obtain the target digital asset risk prediction model. This allows the digital asset risk difference prediction model to specifically learn from the data that the basic digital asset risk prediction model cannot accurately predict, enabling it to better complement the basic model. Consequently, the target digital asset risk prediction model more accurately predicts the credit risk of digital asset applicants, maximizing the gain of the digital asset risk difference prediction model on top of the basic model and improving the accuracy of the target digital asset risk prediction model. Attached Figure Description

[0059] Other features, objects, and advantages of this disclosure will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings. The drawings are for illustrative purposes only and are not intended to limit the invention. In the drawings:

[0060] Figure 1 This is a system architecture diagram in which an embodiment of the digital asset risk prediction model construction method disclosed herein can be applied;

[0061] Figure 2 This is a flowchart of an embodiment of the digital asset risk prediction model construction method according to the present disclosure;

[0062] Figure 3This is an exploded flowchart of one embodiment of step 202 of this disclosure;

[0063] Figure 4 This is an exploded flowchart of one embodiment of step 203 of this disclosure;

[0064] Figure 5 This is an exploded flowchart of one embodiment of step 204 of this disclosure;

[0065] Figure 6 This is an example of a probability density distribution of the target predicted risk label values ​​of each modeling sample output by the trained digital asset prediction base model.

[0066] Figure 7 It could be an example of a density distribution map of the target predicted risk difference of each modeling sample output by a trained digital asset risk prediction model;

[0067] Figure 8 This is a flowchart of one embodiment of the digital asset risk prediction method according to the present disclosure;

[0068] Figure 9 This is a schematic diagram of a structure of an embodiment of the digital asset risk assessment device according to the present disclosure;

[0069] Figure 10 This is a schematic diagram of a structure of an embodiment of the digital asset risk prediction device according to the present disclosure;

[0070] Figure 11 This is a schematic diagram of the structure of a computer system suitable for implementing embodiments of the present disclosure. Detailed Implementation

[0071] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0072] It should be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other. This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0073] Figure 1 An exemplary system architecture 100 is shown, in which embodiments of the digital asset risk prediction model construction method, apparatus, terminal device, and storage medium of this disclosure can be applied.

[0074] like Figure 1As shown, system architecture 100 may include terminal devices 101, 102, and 103, network 104, and server 105. Network 104 is used to provide communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various communication connection types, such as wired communication links, wireless communication links, etc.

[0075] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as applications for building digital asset risk prediction models, applications for digital asset risk assessment, 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.

[0076] Terminal devices 101, 102, and 103 can be either hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with microphones and speakers, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), portable computers, and desktop computers, etc. When terminal devices 101, 102, and 103 are software, they can be installed in the aforementioned electronic devices. They can be implemented as multiple software programs or software modules (e.g., acquiring modeling sample sets) or as a single software program or software module. No specific limitations are imposed here.

[0077] Server 105 can be a server that provides various services, such as a backend server that processes the modeling sample sets obtained from terminal devices 101, 102, and 103. The backend server can perform corresponding processing based on the modeling sample sets obtained from the terminal devices.

[0078] In some cases, the digital asset risk prediction model construction method provided in this disclosure can be jointly executed by terminal devices 101, 102, and 103 and server 105. For example, the step of "obtaining the modeling sample set" can be executed by terminal devices 101, 102, and 103, and the step of "generating a trained digital asset risk prediction basic model based on the digital asset application data and real risk label values ​​of each modeling sample" can be executed by server 105. This disclosure does not limit this. Correspondingly, the digital asset risk prediction model construction device can also be respectively set in terminal devices 101, 102, and 103 and server 105.

[0079] In some cases, the digital asset risk prediction model construction method provided in this disclosure can be executed by server 105. Accordingly, the digital asset risk prediction model construction device can also be set in server 105. In this case, the system architecture 100 may not include terminal devices 101, 102, and 103.

[0080] In some cases, the digital asset risk prediction model construction method provided in this disclosure can be executed by terminal devices 101, 102, and 103. Correspondingly, the digital asset risk prediction model construction device can also be set in terminal devices 101, 102, and 103. In this case, the system architecture 100 may not include server 105.

[0081] It should be noted that server 105 can be either hardware or software. When server 105 is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When server 105 is software, it can be implemented as multiple software programs or software modules (e.g., used to provide distributed services), or as a single software program or software module. No specific limitations are made here.

[0082] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0083] All information, data, and signals disclosed herein are authorized by the user or by all parties, and the collection, use, and processing of such data comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0084] Continue to refer to Figure 2 , Figure 2 A flowchart 200 is shown as an embodiment of a digital asset risk prediction model construction method according to the present disclosure. Figure 2 The digital asset risk prediction model construction method shown can be applied to Figure 1The terminal device or server shown. This process 200 includes at least the following steps 201-205.

[0085] Step 201: Obtain the modeling sample set.

[0086] In this embodiment, the modeling sample set is used to construct the basic model for digital asset risk prediction and the digital asset risk difference prediction model.

[0087] The modeling sample set includes digital asset application data, real risk label values, and digital asset association data for each modeling sample.

[0088] In this embodiment, digital asset application data may refer to data from a first data source. In some examples, digital asset application data may be internal data owned by a digital asset service provider.

[0089] For example, digital asset application data includes the number of times a user applies for digital assets at various digital asset institutions, the value of the digital assets, the value of overdue digital assets, the number of successful digital asset applications, and the number of failed digital asset applications.

[0090] The true risk label value can refer to the credit mark made by the digital asset service provider for the digital asset applicant based on the digital asset application data of the digital asset applicant, which is used to characterize the credit risk of the digital asset applicant.

[0091] The true risk label value can categorize the creditworthiness of digital asset applicants into good and bad credit.

[0092] The true risk label value can be either 0 or 1. When the true risk label value of a digital asset applicant is 0, the applicant is considered to have good credit and low credit risk (e.g., the applicant will return the digital assets within the agreed period). When the true risk label value of a digital asset applicant is 1, the applicant is considered to have bad credit and high credit risk (e.g., the applicant will not return the digital assets within the agreed period).

[0093] Digital asset-related data can refer to data originating from a second data source, which is different from the first data source. Digital asset-related data can be used to further assess the credit risk of digital asset application users, building upon existing digital asset application data. In some embodiments, the second data source is a third-party data provider. In some examples, at least some dimensions of the digital asset-related data are not part of the digital asset application data.

[0094] For example, digital asset-related data can include the user's spending data on third-party platforms, digital asset income data, and application installation and activity data. Digital asset application data includes the number of times a user has applied for digital assets at various digital asset institutions, the value of the digital assets, the value of overdue digital assets, the number of successful digital asset applications, and the number of failed digital asset applications. However, the user's spending data, digital asset income data, and application installation and activity data on third-party platforms are not considered part of the digital asset application data. Digital asset-related data can supplement the digital asset application data to further assess the credit risk of the user.

[0095] Step 202: Generate a trained basic model for digital asset risk prediction based on the digital asset application data and real risk label values ​​of each modeling sample.

[0096] In some alternative implementations, step 202 may specifically include, for example: Figure 3 Steps 2021-2024 are shown.

[0097] Step 2021: Divide each modeling sample in the modeling sample set into a training sample set and a test sample set.

[0098] In this embodiment, a portion of the modeling samples in the modeling sample set can be randomly selected as the training sample set, and the remaining portion of the modeling samples can be used as the test sample set.

[0099] For example, the modeling sample set contains digital asset application data, real risk label values, and digital asset association data of 10,000 digital asset applicants. 7,000 digital asset applicants' data (such as digital asset application data, real risk label values, and digital asset association data) can be randomly selected as the training sample set, and the remaining 3,000 digital asset applicants' data can be used as the test sample set.

[0100] In step 202, the digital asset application data and real risk label values ​​of each training sample in the training sample set can be used to train the digital asset risk prediction basic model to be trained, so that the digital asset risk prediction basic model to be trained can learn the mapping relationship between the input data and the target output.

[0101] The test sample set can be used to evaluate the performance of the underlying digital asset risk prediction model under test. Through the test sample set, we can understand whether the generalization ability of the underlying digital asset risk prediction model under test meets the expected standards.

[0102] Step 2022: Train the basic model for predicting digital asset risk to be trained based on the digital asset application data and real risk label values ​​of each training sample in the training sample set, and generate the basic model for predicting digital asset risk to be tested.

[0103] In this embodiment, the basic model for digital asset risk prediction can be any known binary classification model.

[0104] For example, a binary classification model could be the LightGBMClassifier algorithm.

[0105] In some alternative implementations, the underlying model for predicting the risk of the digital asset to be tested can be obtained through the following (A1)-(A3).

[0106] (A1) The digital asset risk prediction basic model to be trained is trained based on the digital asset application data and real risk label values ​​of each training sample in the training sample set, and the first predicted risk label value of each training sample is obtained.

[0107] In this embodiment, the digital asset application data of each training sample can be used as the input data of the digital asset risk prediction basic model to be trained, and the real risk label value of each training sample can be used as the learning target (target output) of the digital asset risk prediction basic model to be trained.

[0108] The digital asset application data of each training sample in the training sample set is input into the digital asset risk prediction basic model to be trained. The digital asset risk prediction basic model to be trained outputs the first predicted risk label value of each training sample, so that the digital asset risk prediction basic model to be trained can learn the mapping relationship between digital asset application data and real risk label value.

[0109] Here, the predicted risk label value output by the basic model for digital asset risk prediction can be expressed by the formula:

[0110]

[0111] Where X1 can represent the digital asset application data of each modeling sample, and f can represent the basic model for digital asset risk prediction. It can represent the predicted risk label value of the basic model for predicting digital asset risks.

[0112] In the digital asset risk prediction base model to be trained in (A1), X1 can represent the digital asset application data of each training sample, and f can represent the digital asset risk prediction base model to be trained. This can represent the first predicted risk label value of the digital asset risk prediction base model to be trained.

[0113] (A2) Adjust the first model parameters of the digital asset risk prediction base model to be trained based on the difference between the first predicted risk label value and the actual risk label value of each training sample.

[0114] (A1) After obtaining the first predicted risk label value of each training sample, the first model parameter of the digital asset risk prediction basic model to be trained can be adjusted according to the difference between the first predicted risk label value and the real risk label value of each training sample, so that the first predicted risk label value output by the digital asset risk prediction basic model to be trained can gradually approach the real risk label value of each training sample.

[0115] Understandably, the closer the first predicted risk label value of each training sample output by the digital asset risk prediction basic model is to the true risk label value, the better the training effect of the digital asset risk prediction basic model is, and the more accurately it can predict whether the credit risk of the digital asset applicant is good or bad.

[0116] In this embodiment, the first model parameters can be gradually adjusted in at least one cycle (A1) and (A2).

[0117] Specifically, the number of times the first model parameters are adjusted can be determined based on the number of training iterations of the preset digital asset risk prediction basic model or whether the difference between the first predicted risk label value and the actual risk label value of each training sample meets the first threshold range.

[0118] (A3) The digital asset risk prediction base model to be trained after the first model parameter adjustment is determined as the digital asset risk prediction base model to be tested, and the second predicted risk label value of each training sample output by the digital asset risk prediction base model to be tested is recorded.

[0119] In this embodiment, the digital asset risk prediction base model to be trained after the last adjustment of the first model parameters can be determined as the digital asset risk prediction base model to be tested.

[0120] Simultaneously, the second predicted risk label value of each training sample output by the basic model for predicting the risk of digital assets under test can be recorded.

[0121] In the basic model for predicting digital asset risks under test, X1 in the formula for the predicted risk label value output by the above basic model can represent the digital asset application data of each training sample, and f can represent the basic model for predicting digital asset risks under test. This can represent the second predicted risk label value of the underlying digital asset risk prediction model to be tested.

[0122] At this point, the training process of the basic model for digital asset risk prediction is complete. After training with training samples, the basic model for digital asset risk prediction to be tested may be overfitted, that is, the prediction of the second predicted risk label value on each training sample is more accurate (closer to the real risk label value), but the prediction of the second predicted risk label value for other digital asset applicants is inaccurate. In step 2023 of this embodiment, a test sample set can be used to test the basic model for digital asset risk prediction to be tested in order to determine whether the basic model for digital asset risk prediction to be tested is overfitted.

[0123] Step 2023: Based on the digital asset application data and real risk label values ​​of each test sample in the test sample set, test the basic model for digital asset risk prediction to be tested, and obtain the first test result of the digital asset risk prediction model to be tested.

[0124] In some optional implementations, digital asset application data from each test sample in the test sample set can be input into the digital asset risk prediction base model to be tested to obtain a second predicted risk label value for each test sample. Based on the second predicted risk label value and the actual risk label value of each training sample, and based on the second predicted risk label value and the actual risk label value of each test sample, a first test result of the digital asset risk prediction model to be tested is obtained.

[0125] The KS (Kolmogorov-Smirnov) value can be used to evaluate the ability of a tested digital asset risk prediction model to differentiate between digital asset applicants with good and bad credit.

[0126] Here, the first KS value of the training sample set can be obtained based on the second predicted risk label value and the real risk label value of each training sample, and the second KS value of the test sample set can be obtained based on the second predicted risk label value and the real risk label value of each test sample. Then, the first test result of the digital asset risk prediction basic model to be tested can be obtained based on the difference between the first KS value and the second KS value.

[0127] The first test result is used to characterize whether the discrimination ability of the digital asset risk prediction basic model under test is similar on the training sample set and the test sample set, thereby determining whether the digital asset risk prediction basic model under test is overfitting.

[0128] Step 2024: When the first test result is determined to be qualified, the tested digital asset risk prediction basic model is determined as the trained digital asset risk prediction basic model.

[0129] In some alternative implementations, for example, when the difference between the first KS value and the second KS value is less than the first KS threshold, the first test result is determined to be qualified, indicating that the basic model for predicting digital asset risk under test performs similarly on the training sample set and the test sample set, and that the basic model for predicting digital asset risk under test has good generalization ability and there is no overfitting.

[0130] Here, the first KS threshold can be 0.02, 0.03, 0.04, 0.05, 0.06 or 0.08, etc., and the maximum first KS threshold is no more than 0.1. No specific restrictions are imposed here.

[0131] Generally, the smaller the first KS threshold, the better the generalization ability of the underlying digital asset risk prediction model under test.

[0132] When the underlying digital asset risk prediction model under test has good generalization ability, the tested underlying digital asset risk prediction model is determined as the trained underlying digital asset risk prediction model.

[0133] In some alternative implementations, for example, when the difference between the first KS value and the second KS value is greater than or equal to the first KS threshold, the first test result is determined to be unqualified, indicating that the performance of the digital asset risk prediction basic model to be tested differs significantly on the training sample set and the test sample set, and it is determined that the digital asset risk prediction basic model to be tested is overfitting. At this time, new modeling samples can be used to further train the digital asset risk prediction basic model to be tested until the digital asset risk prediction basic model to be tested has a better generalization ability.

[0134] Based on steps 201-202 above, a trained basic model for digital asset risk prediction is obtained. This basic model is used to preliminarily predict the creditworthiness of digital asset applicants based on digital asset application data.

[0135] Step 203: Based on the trained digital asset risk prediction basic model, obtain the target predicted risk label value and the risk difference between the actual risk label value and the target predicted risk label value for each modeling sample.

[0136] It is understandable that digital asset application data is only part of the behavioral data of digital asset applicants. The results obtained by predicting the credit risk of digital asset applicants based solely on digital asset application data are not accurate enough. Therefore, step 203 can obtain the target predicted risk label value of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value of each modeling sample based on the trained digital asset risk prediction basic model.

[0137] In this embodiment, the target predicted risk label value can represent the probability that the trained digital asset risk prediction base model will predict each modeled sample as having bad credit.

[0138] For example, if the target predicted risk label value is 0.9, it is believed that there is a 90% probability that the credit risk of the digital asset applicant is bad.

[0139] The risk difference of the modeling samples can be the difference between the true risk label value and the target predicted risk label value of the modeling samples. The risk difference can be used to characterize the credit risk of digital asset applicants that the trained digital asset risk prediction basic model cannot predict. By establishing a digital asset risk difference prediction model based on the risk difference, it is possible to specifically learn the parts of data that the digital asset risk prediction basic model cannot predict. In some embodiments of this disclosure, "unable to predict" refers to the situation where the prediction accuracy is lower than a preset threshold (also known as the situation of inaccurate prediction). The preset threshold can be specifically set based on the actual scenario, and this disclosure does not impose specific limitations.

[0140] In some alternative implementations, step 203 may specifically include, for example: Figure 4 Steps 2031-2033 are shown.

[0141] Step 2031: Based on the trained digital asset risk prediction basic model, obtain the target predicted risk label value of each training sample and the target predicted risk label value of each test sample.

[0142] In this embodiment, the target predicted risk label value for each training sample and the target predicted risk label value for each test sample are obtained based on the trained digital asset risk prediction model. This can be achieved in two ways:

[0143] Method 1: After determining the basic model for digital asset risk prediction after training, input each training sample into the basic model for digital asset risk prediction after training, and output the target predicted risk label value of each training sample through the basic model for digital asset risk prediction after training, so as to obtain the target predicted risk label value of each training sample.

[0144] Each test sample is input into the trained digital asset risk prediction base model, and the trained digital asset risk prediction base model outputs the target predicted risk label value of each test sample to obtain the target predicted risk label value of each test sample.

[0145] In the trained digital asset risk prediction base model in step 2031, X1 can represent the digital asset application data of each modeling sample (training sample and test sample), and f can represent the trained digital asset risk prediction base model. It can represent the target predicted risk label value output by the trained digital asset risk prediction base model.

[0146] Method 2: After the first test result is qualified and the basic model for digital asset risk prediction after training is determined, the second predicted risk label value of each training sample obtained in (A3) is determined as the target predicted risk label value of each training sample.

[0147] The second predicted risk label value of each test sample obtained in step 2023 is determined as the target predicted risk label value of each test sample.

[0148] Step 2032: Calculate the first difference between the true risk label value of each training sample and the target predicted risk label value of the corresponding training sample, and calculate the second difference between the true risk label value of each test sample and the target predicted risk label value of the corresponding test sample.

[0149] In this embodiment, the first difference between the true risk label value of a training sample and the target predicted risk label value of the corresponding training sample can represent the error between the target predicted risk label value and the true risk label value predicted by the trained digital asset risk prediction basic model for each training sample (i.e., the difference between the true risk label value and the target predicted risk label value, also known as the residual). The second difference between the true risk label value of each test sample and the target predicted risk label value of the corresponding test sample can represent the error between the target predicted risk label value and the true risk label value predicted by the trained digital asset risk prediction basic model for each test sample.

[0150] In this embodiment, the true risk label values ​​are 0 and 1, and the target predicted risk label values ​​are in the range of 0-1. Therefore, the range of the first difference and the second difference can be between -1 and 1.

[0151] Step 2033: Obtain the risk difference for each modeling sample based on the first difference and the second difference.

[0152] The risk difference among the modeling samples can be expressed by the following formula:

[0153]

[0154] Where Δy can represent the risk difference of the modeled samples, and y can represent the true risk label value of the modeled samples. It can represent the target predicted risk label value of the modeled sample.

[0155] In one embodiment, a first difference can be determined as the risk difference between training samples, and a second difference can be determined as the risk difference between test samples.

[0156] Step 204: Based on the digital asset association data of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value of each modeling sample, generate the trained digital asset risk difference prediction model.

[0157] In some alternative implementations, step 204 may specifically include, for example: Figure 5 Steps 2041-2043 are shown.

[0158] Step 2041: Train the digital asset risk difference prediction model to be trained based on the digital asset association data and risk difference of each training sample in the training sample set, and generate the digital asset risk difference prediction model to be tested.

[0159] In this embodiment, the digital asset association data and risk difference of each training sample in the training sample set can be used to train the digital asset risk difference prediction model to be trained, so that the digital asset risk difference prediction model to be trained can learn the mapping relationship between the input data and the target output.

[0160] The digital asset association data and risk difference of each test sample in the test sample set can be used to evaluate the performance of the digital asset risk difference prediction model under test. Through the test sample set, we can understand whether the generalization ability of the tested digital asset risk difference prediction model meets the prediction standard.

[0161] In this embodiment, the digital asset risk difference prediction model can be any known regression model.

[0162] For example, the regression model could be the LightGBMRegressor algorithm.

[0163] Here, digital asset-related data can serve as supplementary data to digital asset application data, further used to assess the credit risk of digital asset applicants. The risk difference represents the portion of data that the trained digital asset risk prediction model cannot predict. Therefore, based on the digital asset-related data and risk differences of each modeling sample, a trained digital asset risk difference prediction model is generated. This trained model can also serve as a digital asset risk residual prediction model, enabling it to specifically learn from the portions of data that the basic digital asset risk prediction model cannot accurately predict. Thus, the digital asset risk difference prediction model can better complement the basic digital asset risk prediction model, further improving the accuracy of credit risk prediction for digital asset applicants. This maximizes the benefit of the digital asset risk difference prediction model to the basic digital asset risk prediction model.

[0164] In some alternative implementations, the digital asset risk difference prediction model to be tested can be obtained through the following (C1)-(C3).

[0165] (C1) The digital asset risk difference prediction model to be trained is trained based on the digital asset association data and risk difference of each training sample in the first training set, and the first predicted risk difference of each training sample is obtained.

[0166] In this embodiment, the digital asset association data of each training sample can be used as the input data of the digital asset risk difference prediction model to be trained, and the risk difference of each training sample can be used as the learning target (target output) of the digital asset risk difference prediction model to be trained.

[0167] The digital asset association data of each training sample in the training sample set is input into the digital asset risk difference prediction model to be trained, and the first predicted risk difference of each training sample is output by the digital asset risk prediction difference model to be trained.

[0168] This enables the digital asset risk difference prediction model to be trained to learn the mapping relationship between digital asset-related data and risk differences.

[0169] Here, the predicted risk difference output by the digital asset risk difference prediction model can be expressed by the formula:

[0170]

[0171] Where X2 can represent the digital asset association data of each modeling sample, and g can represent the digital asset risk difference prediction model. It can represent the predicted risk difference of a digital asset risk difference prediction model.

[0172] In the digital asset risk difference prediction model to be trained in C(1), X2 can represent the digital asset association data of each training sample, and g can represent the digital asset risk difference prediction model to be trained. This can represent the first predicted risk difference of the digital asset risk difference prediction model to be trained.

[0173] (C2) Adjust the second model parameters of the digital asset risk difference prediction model to be trained based on the difference between the first predicted risk difference and the risk difference for each training sample.

[0174] (C1) After obtaining the first predicted risk difference for each training sample, the second model parameters of the digital asset risk difference prediction model to be trained can be adjusted according to the difference between the first predicted risk difference and the risk difference for each training sample, so that the first predicted risk difference output by the digital asset risk difference prediction model to be trained can gradually approach the risk difference for each training sample.

[0175] Understandably, the closer the first predicted risk difference of each training sample output by the digital asset risk difference prediction model is to the actual risk difference, the better the training effect of the digital asset risk difference prediction model is, and the more effectively it can supplement the credit risk data that the trained digital asset risk prediction basic model cannot predict.

[0176] In this embodiment, the second model parameters can be gradually adjusted in at least one cycle (C1) and (C).

[0177] Specifically, the number of times the second model parameters are adjusted can be determined based on the number of training iterations of the preset digital asset risk difference prediction model or whether the difference between the first predicted risk difference and the risk difference of each training sample meets the second threshold range.

[0178] (C3) The digital asset risk difference prediction model to be trained after the second model parameter adjustment is determined as the digital asset risk difference prediction model to be tested, and the second predicted risk difference of each training sample output by the digital asset risk difference prediction model to be tested is recorded.

[0179] In this embodiment, the digital asset risk difference prediction model to be trained after the last second model parameter adjustment can be identified as the digital asset risk difference prediction model to be tested.

[0180] Simultaneously, the second predicted risk difference of each training sample output by the digital asset risk difference prediction model under test can be recorded.

[0181] In the digital asset risk difference prediction model to be tested, X2 in the predicted risk difference formula output by the above digital asset risk difference prediction model can represent the digital asset association data of each training sample, and g can represent the digital asset risk difference prediction model to be tested. This can represent the second predicted risk difference of the digital asset risk difference prediction model to be tested.

[0182] At this point, the training process of the digital asset risk difference prediction model is completed. After training with training samples, the digital asset risk difference prediction model to be tested may be overfitted, that is, the second predicted risk difference is more accurate (closer to the risk difference) for each training sample, but the second predicted risk difference for other digital asset applicants is not accurate. In step 2042 of this embodiment, a test sample set can be used to test the digital asset risk difference prediction model to be tested in order to determine whether the digital asset risk difference prediction model to be tested is overfitted.

[0183] Step 2042: Based on the digital asset association data and risk difference of each test sample in the test sample set, test the digital asset risk difference prediction model to be tested, and obtain the second test result of the digital asset risk difference prediction model to be tested.

[0184] In some optional implementations, the digital asset association data of each test sample in the test sample set can be input into the digital asset risk difference prediction model to be tested, and the second predicted risk difference of each test sample can be output. Based on the second predicted risk difference and risk difference of each training sample, and based on the second predicted risk difference and risk difference of each test sample, the second test result of the digital asset risk difference prediction model to be tested can be obtained.

[0185] Here, the third KS value of the training sample set can be obtained based on the second predicted risk difference and the risk difference of each training sample, and the fourth KS value of the test sample set can be obtained based on the second predicted risk difference and the risk difference of each test sample. Then, the second test result of the digital asset risk difference prediction model to be tested can be obtained based on the difference between the third KS value and the fourth KS value.

[0186] The second test result is used to characterize whether the performance of the digital asset risk difference prediction model under test is similar to that of the training sample set and the test sample set, thereby determining whether the digital asset risk difference prediction model under test is overfitting.

[0187] Step 2043: When the second test result is determined to be qualified, the tested digital asset risk difference prediction model is determined as the trained digital asset risk difference prediction model.

[0188] In some alternative implementations, for example, when the difference between the third KS value and the fourth KS value is less than the second KS threshold, the second test result is determined to be qualified, indicating that the digital asset risk difference prediction model to be tested performs similarly on the training sample set and the test sample set, and that the digital asset risk difference prediction model to be tested has good generalization ability and there is no overfitting.

[0189] Here, the second KS threshold is 0.02, 0.03, 0.04, 0.05, 0.06 or 0.08, etc., and the maximum second KS threshold is no more than 0.1. No specific restrictions are imposed here.

[0190] Generally, the smaller the first KS threshold, the better the generalization ability of the digital asset risk difference prediction model under test.

[0191] When the digital asset risk difference prediction model under test has good generalization ability, the tested digital asset risk difference prediction model is determined as the trained digital asset risk difference prediction model.

[0192] In some alternative implementations, for example, when the difference between the third KS value and the fourth KS value is greater than or equal to the second KS threshold, the second test result is determined to be unqualified, indicating that the performance of the digital asset risk difference prediction model to be tested differs significantly between the training sample set and the test sample set, and it is determined that the digital asset risk difference prediction model to be tested is overfitting. At this time, new modeling samples can be used to further train the digital asset risk difference prediction model to be tested until the digital asset risk difference prediction model to be tested has good generalization ability.

[0193] Based on steps 203-204 above, a trained digital asset risk difference prediction model is obtained. The trained digital asset risk difference prediction model can specifically learn from the data that the basic digital asset risk prediction model cannot predict accurately, thus playing a better supplementary role to the basic digital asset risk prediction model. This makes the target digital asset risk prediction model more accurate in predicting the credit risk of digital asset applicants, and can maximize the gain of the digital asset risk difference prediction model on the basic digital asset risk prediction model.

[0194] Step 205: Generate the target digital asset risk prediction model based on the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model.

[0195] In some alternative implementations, the trained digital asset risk prediction base model and the trained digital asset risk prediction model can be merged to generate the target digital asset risk prediction model.

[0196] For example, the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model can be added together to generate the target digital asset risk prediction model.

[0197] This embodiment of the disclosure uses a trained digital asset risk difference prediction model to predict credit data that the trained digital asset risk prediction base model cannot predict. This allows the trained digital asset risk difference prediction model to specifically learn the data that the digital asset risk prediction base model cannot predict accurately. Then, the target predicted risk label value (i.e., the credit data that the digital asset risk prediction base model can predict) predicted by the digital asset risk prediction base model and the target predicted risk difference value (i.e., the predicted credit data that the digital asset risk prediction base model cannot predict) output by the trained digital asset risk difference prediction model are accumulated. This allows the digital asset risk difference prediction model to better complement the digital asset risk prediction base model, thereby making the target digital asset risk prediction model more accurate in predicting the credit risk of digital asset applicants. This maximizes the gain of the digital asset risk difference prediction model on the digital asset risk prediction base model and improves the accuracy of the digital asset risk prediction model.

[0198] The target digital asset risk prediction model can be expressed by the following formula:

[0199]

[0200] Where f represents the trained digital asset risk prediction base model, and g represents the trained digital asset risk difference prediction model. This represents a risk prediction model for target digital assets.

[0201] In some optional implementations, depending on actual needs, based on the aforementioned digital asset risk difference prediction model, the target predicted risk difference for each training sample and the target predicted risk difference for each test sample can be obtained. The second digital asset association data for each modeling sample is used as input, and the target predicted risk difference is used as the learning objective (target output). This allows for the continued training of new digital asset risk difference prediction models. For example, each time, new digital asset association data for each modeling sample can be added as input, and the target predicted risk difference output by the previously trained digital asset risk difference prediction model can be used as the learning objective, resulting in at least one new trained digital asset risk difference prediction model. Then, each trained digital asset risk difference prediction model is fused with the trained digital asset risk prediction base model to generate the target digital asset risk prediction model.

[0202] For example, the various trained digital asset risk difference prediction models and the trained digital asset risk prediction base model can be added together to generate the target digital asset risk prediction model.

[0203] Understandably, by combining the trained digital asset risk difference prediction model and the basic digital asset risk prediction model, the target digital asset risk prediction model will be more accurate in predicting the credit risk of digital asset applicants.

[0204] like Figure 6 and Figure 7 As shown, Figure 6 It could be an example of a probability density distribution map of the target predicted risk label values ​​of each modeling sample output by a trained digital asset prediction base model; Figure 7 This could be an example of a density distribution map of the target predicted risk difference for each modeling sample output by a trained digital asset risk difference prediction model.

[0205] As the trained digital asset risk prediction base model and multiple trained digital asset risk difference prediction models are added together, the target digital asset risk prediction model will eventually predict the target risk value of each modeling sample to be 0 or 1. That is, the more accurately it can predict whether the credit risk of each modeling sample is good or bad, the more likely its density distribution map will eventually show a bimodal distribution that tends to be 0 and 1.

[0206] The digital asset risk prediction model construction method provided in this disclosure first obtains a modeling sample set, which includes digital asset application data, actual risk label values, and digital asset association data for each modeling sample. Based on the digital asset application data and actual risk label values ​​of each modeling sample, a trained digital asset risk prediction base model is generated. Then, based on the trained digital asset risk prediction base model, the risk difference between the actual risk label and the target predicted risk label value for each modeling sample is obtained. Next, based on the digital asset association data and risk difference values ​​of each modeling sample, a trained digital asset risk difference prediction model is generated. Finally, based on the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model, a target digital asset risk prediction model is generated. In constructing a digital asset risk difference prediction model, this disclosure obtains a risk difference based on the actual risk label value and the target predicted risk label value output by the basic digital asset risk prediction model. Then, a residual digital asset risk difference prediction model is constructed using the risk difference and associated digital asset data. This allows the digital asset risk difference prediction model to specifically learn from the data that the basic digital asset risk prediction model cannot predict accurately, thus better complementing the basic digital asset risk prediction model. Consequently, the target digital asset risk prediction model can more accurately predict the credit risk of digital asset applicants, maximizing the gain of the digital asset risk difference prediction model on the basic digital asset risk prediction model and improving the accuracy of the target digital asset risk prediction model.

[0207] Continue to refer to Figure 8 , Figure 8 A flowchart 800 is shown as an embodiment of a digital asset risk prediction method according to the present disclosure, the flowchart 800 including at least the following steps 801-802.

[0208] After constructing the target digital asset risk prediction model as described above, steps 801-802 can be the application process of the above digital asset risk prediction model.

[0209] The digital asset risk prediction base model, digital asset risk difference prediction model, and target digital asset risk prediction model in steps 801-802 can be obtained through training in steps 201-205.

[0210] Step 801: Obtain the digital asset application data and digital asset association data of the target digital asset applicant.

[0211] In this embodiment, the target digital asset applicant can refer to the digital asset applicant whose creditworthiness the target digital asset service provider wants to obtain.

[0212] The digital asset application data of the target digital asset applicant can refer to data from a primary data source. In some examples, the digital asset application data of the target digital asset applicant is internal data about the target digital asset applicant owned by the target digital asset service provider.

[0213] For example, the digital asset application data of the target digital asset applicant includes the number of times the target digital asset applicant applied for digital assets at various digital asset institutions, the value of digital assets, the value of overdue digital assets, the number of successful digital asset applications, and the number of failed digital asset applications.

[0214] The digital asset association data of the target digital asset applicant can refer to data originating from a second data source, which is different from the first data source. This data can be used to further assess the credit risk of the target digital asset applicant, building upon the digital asset application data. In some embodiments, the second data source is a third-party data provider. In some examples, at least some dimensions of the digital asset association data are not part of the digital asset application data.

[0215] For example, the digital asset-related data of a target digital asset applicant may include the applicant's consumption data, digital asset income data, and application installation and activity data on third-party platforms. Digital asset application data includes the number of times the target digital asset applicant has applied for digital assets at various digital asset institutions, the value of the digital assets, the value of overdue digital assets, the number of successful digital asset applications, and the number of failed digital asset applications. However, the consumption data, digital asset income data, and application installation and activity data on third-party platforms are not considered part of the digital asset application data. The digital asset-related data can serve as a supplement to the digital asset application data, further assessing the credit risk of the digital asset applicant.

[0216] Step 802: Input the digital asset application data and associated data of the target digital asset applicant into the target digital asset risk prediction model, and output the target predicted risk value of the target digital asset applicant.

[0217] In this embodiment, the target digital asset risk prediction model includes a basic digital asset risk prediction model and a digital asset risk difference prediction model. Here, the basic digital asset risk prediction model and the digital asset risk difference prediction model can be fused to obtain the target digital asset risk prediction model. The process of obtaining the target digital asset risk prediction model can be as described above. Figure 1-7 The method described in the illustrated embodiment.

[0218] Among them, the basic model for digital asset risk prediction is used to obtain the target predicted risk label value of the target digital asset applicant based on the digital asset application data of the target digital asset applicant.

[0219] The target predicted risk label value of the target digital asset applicant can be used to characterize the initial credit score of the target digital asset applicant.

[0220] The digital asset risk difference prediction model is used to obtain the target predicted risk difference of the target digital asset applicant based on the digital asset association data of the target digital asset applicant.

[0221] The target predicted risk difference of the target digital asset applicant can be used to characterize the predicted risk value of the credit risk of the target digital asset applicant that cannot be predicted by the basic model for digital asset risk prediction.

[0222] Here, the target predicted risk difference can supplement the target predicted risk label value to more accurately predict the target predicted risk value of the target digital asset applicant.

[0223] Here, you can first obtain the target predicted risk label value of the target digital asset applicant based on the basic digital asset risk prediction model, and then obtain the target predicted risk difference of the target digital asset applicant based on the digital asset risk difference prediction model. Alternatively, you can first obtain the target predicted risk difference of the target digital asset applicant based on the digital asset risk difference prediction model, and then obtain the target predicted risk label value of the target digital asset applicant based on the basic digital asset risk prediction model. There are no specific restrictions.

[0224] The target digital asset risk prediction model is used to obtain the target predicted risk value of the target digital asset applicant based on the target predicted risk label value and the target predicted risk difference.

[0225] In some examples, the target predicted risk label value and the target predicted risk difference can be combined (e.g., added) to obtain the target predicted risk value for the target digital asset applicant.

[0226] Thus, the digital asset risk difference prediction model can obtain the target predicted risk difference of the credit risk of the target digital asset applicant, which cannot be predicted by the basic digital asset risk prediction model. Then, it is fused (e.g., added) with the target predicted risk label value predicted by the basic digital asset risk prediction model to obtain the target predicted risk value of the target digital asset applicant. The target predicted risk difference can supplement the target predicted risk label value. In the process of predicting the target predicted risk value of the target digital asset applicant, the digital asset-related data can maximize its supplementary role to the digital asset application data, thereby increasing the accuracy of the final predicted target risk value.

[0227] Further reference Figure 9 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of a digital asset risk prediction model construction device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various terminal devices.

[0228] like Figure 9As shown, the digital asset risk prediction model construction device of this embodiment includes: a sample set acquisition unit 901, a first generation unit 902, a second generation unit 903, a third generation unit 904, and a fourth generation unit 905. The system includes four main components: a sample set acquisition unit 901, which acquires a modeling sample set including digital asset application data, real risk label values, and digital asset association data for each modeling sample; a first generation unit 902, which generates a trained digital asset risk prediction base model based on the digital asset application data and real risk label values ​​of each modeling sample; a second generation unit 903, which obtains the target predicted risk label value and the risk difference between the real risk label value and the target predicted risk label value for each modeling sample based on the trained digital asset risk prediction base model; a third generation unit 904, which generates a trained digital asset risk difference prediction model based on the digital asset association data and the risk difference between the real risk label value and the target predicted risk label value of each modeling sample; and a fourth generation unit 905, which generates a target digital asset risk prediction model based on the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model.

[0229] In this embodiment, the specific processing of the sample set acquisition unit 901, the first generation unit 902, the second generation unit 903, the third generation unit 904, and the fourth generation unit 905, and the resulting technical effects, can be found by referring to [the relevant documentation]. Figure 2 The relevant descriptions of steps 201 to 205 in the corresponding embodiments will not be repeated here.

[0230] In some alternative implementations, the first generation unit 902 described above may be further configured as follows:

[0231] The modeling samples in the modeling sample set are divided into a training sample set and a test sample set;

[0232] The basic model for predicting digital asset risk is trained based on the digital asset application data and real risk label values ​​of each training sample in the training sample set, and a basic model for predicting digital asset risk to be tested is generated.

[0233] Based on the digital asset application data and real risk label values ​​of each test sample in the test sample set, the basic model for digital asset risk prediction to be tested is tested to obtain the first test result of the basic model for digital asset risk prediction to be tested.

[0234] Once the first test result is deemed satisfactory, the tested digital asset risk prediction base model will be designated as the trained digital asset risk prediction base model.

[0235] In some alternative implementations, the first generation unit 902 described above may be further configured as follows:

[0236] The digital asset risk prediction model to be trained is trained based on the digital asset application data and real risk label values ​​of each training sample in the training sample set, and the first predicted risk label value of each training sample is obtained.

[0237] The first model parameters of the digital asset risk prediction base model to be trained are adjusted based on the difference between the first predicted risk label value and the actual risk label value of each training sample.

[0238] The digital asset risk prediction base model to be trained after adjusting the parameters of the first model is determined as the digital asset risk prediction base model to be tested, and the second predicted risk label value of each training sample output by the digital asset risk prediction base model to be tested is recorded.

[0239] In some alternative implementations, the first generation unit 902 described above may be further configured as follows:

[0240] Input the digital asset application data of each test sample in the test sample set into the basic model for predicting digital asset risk to be tested, and output the second predicted risk label value of each test sample.

[0241] Based on the second predicted risk label value and the actual risk label value of each training sample, and based on the second predicted risk label value and the actual risk label value of each test sample, the first test result of the basic model for predicting the risk of digital assets to be tested is obtained.

[0242] In some alternative implementations, the second generation unit 903 described above can be further configured as follows:

[0243] Based on the trained digital asset risk prediction model, the target predicted risk label value of each training sample and the target predicted risk label value of each test sample are obtained.

[0244] Calculate the first difference between the true risk label value and the target predicted risk label value for each training sample, and calculate the second difference between the true risk label value and the target predicted risk label value for each test sample;

[0245] The risk difference for each modeling sample is obtained based on the first and second differences.

[0246] In some alternative implementations, the third generation unit 904 described above can be further configured as follows:

[0247] The digital asset risk difference prediction model to be trained is trained based on the digital asset association data and risk difference of each training sample in the training sample set, and the digital asset risk difference prediction model to be tested is generated.

[0248] Based on the digital asset association data and risk difference of each test sample in the test sample set, the digital asset risk difference prediction model to be tested is tested to obtain the second test result of the digital asset risk difference prediction model to be tested.

[0249] When the second test result is deemed satisfactory, the tested digital asset risk difference prediction model is identified as the trained digital asset risk difference prediction model.

[0250] In some alternative implementations, the third generation unit 904 described above can be further configured as follows:

[0251] The digital asset risk difference prediction model to be trained is trained based on the digital asset association data and risk difference of each training sample in the training sample set, and the first predicted risk difference of each training sample is obtained.

[0252] The second model parameters of the digital asset risk difference prediction model to be trained are adjusted based on the difference between the first predicted risk difference and the risk difference for each training sample.

[0253] The digital asset risk difference prediction model to be trained after adjusting the second model parameters is determined as the digital asset risk difference prediction model to be tested, and the second predicted risk difference of each training sample output by the digital asset risk difference prediction model to be tested is recorded.

[0254] In some alternative implementations, the third generation unit 904 described above can be further configured as follows:

[0255] Input the digital asset association data of each test sample in the test sample set into the digital asset risk difference prediction model to be tested, and output the second predicted risk difference of each test sample.

[0256] Based on the second predicted risk difference and risk difference of each training sample, and based on the second predicted risk difference and risk difference of each test sample, the second test result of the digital asset risk difference prediction model to be tested is obtained.

[0257] It should be noted that the implementation details and technical effects of each unit in the digital asset risk prediction model construction device provided in the embodiments of this disclosure can be referred to the descriptions of other embodiments in this disclosure, and will not be repeated here.

[0258] Further reference Figure 10As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of a digital asset risk prediction device, which is similar to... Figure 8 Corresponding to the method embodiments shown, this device can be specifically applied to various terminal devices.

[0259] like Figure 10 As shown in the figure, a digital asset risk prediction device of this embodiment includes a data acquisition unit 1001 and a predicted risk value output unit 1002. The data acquisition unit 1001 is used to acquire the digital asset application data and related data of the target digital asset applicant. The predicted risk value output unit 1002 is used to input the digital asset application data and related data of the target digital asset applicant into the target digital asset risk prediction model and output the target predicted risk value of the target digital asset applicant. The target digital asset risk prediction model includes a basic digital asset risk prediction model and a digital asset risk difference prediction model. The basic digital asset risk prediction model is used to obtain the target predicted risk label value of the target digital asset applicant based on the digital asset application data of the target digital asset applicant. The digital asset risk difference prediction model is used to obtain the target predicted risk difference value of the target digital asset applicant based on the related data of the target digital asset applicant. The target digital asset risk prediction model is used to obtain the target predicted risk value of the target digital asset applicant based on the target predicted risk label value and the target predicted risk difference value. The basic digital asset risk prediction model, the digital asset risk difference prediction model, and the target digital asset risk prediction model are obtained through training in each step of the digital asset risk prediction model construction device.

[0260] In this embodiment, the specific processing of the data acquisition unit 1001, the predicted risk value output unit 1002, the second output unit 1003, and the third output unit 1004, and the resulting technical effects, can be found by referring to [the documentation / reference]. Figure 8 The relevant descriptions of steps 801 to 802 in the corresponding embodiments will not be repeated here.

[0261] It should be noted that the implementation details and technical effects of each unit in the digital asset risk prediction device provided in the embodiments of this disclosure can be referred to the descriptions of other embodiments in this disclosure, and will not be repeated here.

[0262] The following is for reference. Figure 11 It shows a schematic diagram of the structure of a computer system 1100 suitable for implementing the terminal device of this disclosure. Figure 11 The computer system 1100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0263] like Figure 4 As shown, the computer system 1100 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 1101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1102 or a program loaded from a storage device 1108 into a random access memory (RAM) 1103. The RAM 1103 also stores various programs and data required for the operation of the computer system 1100. The processing device 1101, ROM 1102, and RAM 1103 are interconnected via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0264] Typically, the following devices can be connected to I / O interface 1105: input devices 1106 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, etc.; output devices 1107 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1108 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1109. Communication device 1109 allows computer system 1100 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 11 A computer system 1100 with various electronic devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0265] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 1109, or installed from storage device 1108, or installed from ROM 1102. When the computer program is executed by processing device 1101, it performs the functions defined above in the methods of embodiments of this disclosure.

[0266] It should be noted that the computer-readable medium described above in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0267] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0268] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the following functions: Figure 2 The embodiments shown and their alternative implementations illustrate a method for constructing a digital asset risk prediction model.

[0269] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone 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 remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0270] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0271] The units described in the embodiments of this disclosure can be implemented in software or in hardware. The names of the units are not necessarily limiting in certain circumstances; for example, a sample set acquisition unit can also be described as "a unit for acquiring a modeling sample set".

[0272] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A method for constructing a digital asset risk prediction model, characterized in that, include: Obtain a modeling sample set, wherein the modeling sample set includes digital asset application data, real risk label values ​​and digital asset association data for each modeling sample; Based on the digital asset application data and the real risk label value of each of the modeling samples, a trained digital asset risk prediction basic model is generated. This generation of the trained digital asset risk prediction basic model includes: generating a digital asset risk prediction basic model to be tested based on the digital asset application data, the real risk label value, and the digital asset risk prediction basic model to be trained in each of the modeling samples; and determining the first test result of the digital asset risk prediction basic model to be tested as qualified based on each test sample in each of the modeling samples, thus identifying the tested digital asset risk prediction basic model as the trained digital asset risk prediction basic model. Based on the trained digital asset risk prediction model, the target predicted risk label value of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value are obtained. Based on the digital asset association data of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value of each modeling sample, a trained digital asset risk difference prediction model is generated. The step of generating the trained digital asset risk difference prediction model based on the digital asset association data of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value includes: training the digital asset risk difference prediction model to be trained based on the digital asset association data and the risk difference of each training sample in each modeling sample to generate a digital asset risk difference prediction model to be tested; when the second test result of the digital asset risk difference prediction model to be tested is determined to be qualified based on each test sample in each modeling sample, the tested digital asset risk difference prediction model is determined as the trained digital asset risk difference prediction model. Based on the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model, a target digital asset risk prediction model is generated.

2. The method according to claim 1, characterized in that, The step of generating a trained digital asset risk prediction base model based on the digital asset application data and the real risk label value of each modeling sample includes: The modeling samples in the modeling sample set are divided into a training sample set and a test sample set; The digital asset risk prediction basic model to be trained is trained based on the digital asset application data and the real risk label value of each training sample in the training sample set, and the digital asset risk prediction basic model to be tested is generated. Based on the digital asset application data and the real risk label value of each test sample in the test sample set, the digital asset risk prediction basic model to be tested is tested to obtain the first test result of the digital asset risk prediction basic model to be tested. When the first test result is determined to be qualified, the tested digital asset risk prediction basic model is determined as the trained digital asset risk prediction basic model.

3. The method according to claim 2, characterized in that, The step of training the digital asset risk prediction basic model to be trained based on the digital asset application data and the real risk label value of each training sample in the training sample set, and generating the digital asset risk prediction basic model to be tested, includes: The digital asset risk prediction model to be trained is trained based on the digital asset application data and the real risk label value of each training sample in the training sample set, so as to obtain the first predicted risk label value of each training sample. The first model parameters of the digital asset risk prediction base model to be trained are adjusted based on the difference between the first predicted risk label value and the actual risk label value of each training sample. The digital asset risk prediction base model to be trained after adjusting the parameters of the first model is determined as the digital asset risk prediction base model to be tested, and the second predicted risk label value of each training sample output by the digital asset risk prediction base model to be tested is recorded.

4. The method according to claim 3, characterized in that, The step involves testing the underlying digital asset risk prediction model based on the digital asset application data and the actual risk label value of each test sample in the test sample set, and obtaining a first test result for the underlying digital asset risk prediction model, including: Input the digital asset application data of each test sample in the test sample set into the digital asset risk prediction basic model to be tested, and output the second predicted risk label value of each test sample; Based on the second predicted risk label value and the actual risk label value of each training sample, and based on the second predicted risk label value and the actual risk label value of each test sample, the first test result of the digital asset risk prediction basic model to be tested is obtained.

5. The method according to any one of claims 1-4, characterized in that, The step of obtaining the target predicted risk label value and the risk difference between the actual risk label value and the target predicted risk label value for each modeled sample based on the trained digital asset risk prediction model includes: Based on the trained digital asset risk prediction model, the target predicted risk label value of each training sample and the target predicted risk label value of each test sample are obtained. Calculate a first difference between the true risk label value and the target predicted risk label value for each of the training samples, and calculate a second difference between the true risk label value and the target predicted risk label value for each of the test samples; The risk difference for each modeling sample is obtained based on the first difference and the second difference.

6. The method according to claim 5, characterized in that, The step of generating a trained digital asset risk difference prediction model based on the digital asset association data of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value includes: The digital asset risk difference prediction model to be trained is trained based on the digital asset association data and risk difference of each training sample in the training sample set, and the digital asset risk difference prediction model to be tested is generated. Based on the digital asset association data and risk difference of each test sample in the test sample set, the digital asset risk difference prediction model to be tested is tested to obtain the second test result of the digital asset risk difference prediction model to be tested. When the second test result is determined to be qualified, the tested digital asset risk difference prediction model is determined as the trained digital asset risk difference prediction model.

7. The method according to claim 6, characterized in that, The step of training the digital asset risk difference prediction model to be trained based on the digital asset association data and risk difference of each training sample in the training sample set, and generating the digital asset risk difference prediction model to be tested, includes: The digital asset risk difference prediction model to be trained is trained based on the digital asset association data and risk difference of each training sample in the training sample set, to obtain the first predicted risk difference of each training sample. The second model parameters of the digital asset risk difference prediction model to be trained are adjusted based on the difference between the first predicted risk difference and the risk difference of each training sample. The digital asset risk difference prediction model to be trained after adjusting the second model parameters is determined as the digital asset risk difference prediction model to be tested, and the second predicted risk difference of each training sample output by the digital asset risk difference prediction model to be tested is recorded.

8. The method according to claim 7, characterized in that, The step of testing the digital asset risk difference prediction model to be tested based on the digital asset association data and the risk difference of each test sample in the test sample set, and obtaining a second test result of the digital asset risk difference prediction model to be tested, includes: Input the digital asset association data of each test sample in the test sample set into the digital asset risk difference prediction model to be tested, and output the second predicted risk difference of each test sample. Based on the second predicted risk difference and the risk difference of each training sample, and based on the second predicted risk difference and the risk difference of each test sample, the second test result of the digital asset risk difference prediction model to be tested is obtained.

9. A method for predicting the risk of digital assets, characterized in that, The method includes: Obtain the digital asset application data and associated digital asset data of the target digital asset applicant; The target digital asset application data and associated data of the target digital asset applicant are input into the target digital asset risk prediction model, and the target digital asset risk prediction value of the target digital asset applicant is output. The target digital asset risk prediction model includes a basic digital asset risk prediction model and a digital asset risk difference prediction model. The basic digital asset risk prediction model is used to obtain the target predicted risk label value of the target digital asset applicant based on the digital asset application data of the target digital asset applicant. The digital asset risk difference prediction model is used to obtain the target predicted risk difference value of the target digital asset applicant based on the associated data of the target digital asset applicant. The target digital asset risk prediction model is used to obtain the target predicted risk value of the target digital asset applicant based on the target predicted risk label value and the target predicted risk difference value. The basic digital asset risk prediction model, the digital asset risk difference prediction model, and the target digital asset risk prediction model are trained using the method described in any one of claims 1-8.

10. A device for constructing a digital asset risk prediction model, comprising: The sample set acquisition unit is used to acquire the modeling sample set, wherein the modeling sample set includes digital asset application data, real risk label values ​​and digital asset association data of each modeling sample; The first generation unit is configured to generate a trained digital asset risk prediction basic model based on the digital asset application data and the real risk label value of each of the modeling samples. The first generation unit is further configured to: generate a digital asset risk prediction basic model to be tested based on the digital asset application data, the real risk label value, and the digital asset risk prediction basic model to be trained in each of the modeling samples; and when the first test result of the digital asset risk prediction basic model to be tested is determined to be qualified based on each of the test samples in each of the modeling samples, the tested digital asset risk prediction basic model is determined as the trained digital asset risk prediction basic model. The second generation unit is used to obtain the target predicted risk label value of each modeling sample and the risk difference between the actual risk label value and the target predicted risk label value based on the trained digital asset risk prediction basic model. The third generation unit is used to generate a trained digital asset risk difference prediction model based on the digital asset association data of each of the modeling samples and the risk difference between the real risk label value and the target predicted risk label value. The third generation unit is further used to: train the digital asset risk difference prediction model to be trained based on the digital asset association data and the risk difference of each training sample in each of the modeling samples, generating a digital asset risk difference prediction model to be tested; when the second test result of the digital asset risk difference prediction model to be tested is determined to be qualified based on each test sample in each of the modeling samples, the tested digital asset risk difference prediction model is determined as the trained digital asset risk difference prediction model. The fourth generation unit is used to generate a target digital asset risk prediction model based on the trained digital asset risk prediction base model and the trained digital asset risk difference prediction model.

11. A digital asset risk prediction device, characterized in that, include: The data acquisition unit is used to acquire the digital asset application data and digital asset-related data of the target digital asset applicant. The predicted risk value output unit is used to input the digital asset application data and the associated digital asset data of the target digital asset applicant into the target digital asset risk prediction model, and output the target predicted risk value of the target digital asset applicant. The target digital asset risk prediction model includes a basic digital asset risk prediction model and a digital asset risk difference prediction model. The basic digital asset risk prediction model is used to obtain the target predicted risk label value of the target digital asset applicant based on the digital asset application data of the target digital asset applicant. The digital asset risk difference prediction model is used to obtain the target predicted risk difference of the target digital asset applicant based on the associated digital asset data of the target digital asset applicant. The target digital asset risk prediction model is used to obtain the target predicted risk value of the target digital asset applicant user based on the target predicted risk label value and the target predicted risk difference; the digital asset risk prediction base model, the digital asset risk difference prediction model and the target digital asset risk prediction model are trained by the method described in any one of claims 1-8.

12. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When one or more programs are executed by one or more processors, the one or more processors cause the one or more processors to implement the method as described in any one of claims 1-8 or 9.

13. A computer-readable storage medium having a computer program stored thereon, wherein, When a computer program is executed by one or more processors, it implements the method as described in any one of claims 1-8 or 9.

14. A computer program product comprising a computer program / instructions that, when executed by a processor, implement the method as described in any one of claims 1-8 or 9.

Images