Credit and loan amount assessment method, device, equipment and storage medium

By using cross-encryption algorithms to encrypt user IDs between multiple industries platforms and perform matching and classification training, the problem of data fusion difficulties in multi-industry data cooperation is solved, cross-platform credit risk and loan quota assessment is realized, and more Accurate evaluation results.

CN113724058BActive Publication Date: 2024-11-08CHINA MOBILE COMM LTD RES INST +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010437989.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-05-21
Publication Date
2024-11-08
Estimated Expiration
2040-05-21

AI Technical Summary

Technical Problem

In the multi-industry data cooperation, existing technology cannot achieve deep data fusion due to the restrictions on big data security regulations, resulting in the lack of cross-platform joint assessment of credit risk and loan quota assessment, and the data value cannot be fully tapped.

Method used

By using a cross-encryption algorithm between the first platform and the second platform, and matching it based on the encrypted user ID, a cross-platform user data set is generated, and classified training is performed to obtain a credit risk and loan quota evaluation model. Realize joint assessment of user credit risk and loan quota.

Benefits of technology

It has achieved the security and compliance integration of multi-industry data, provided more scientific and accurate credit risk and loan quota assessment results, and supported users to provide corresponding credit products and services.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113724058B_ABST
    Figure CN113724058B_ABST
Patent Text Reader

Abstract

The present invention discloses a credit and loan amount assessment method, device, equipment and storage medium. Among them, the credit assessment model training method includes: obtaining a first user data set corresponding to multiple users of a first platform; encrypting the user ID in the first user data set based on the cross encryption algorithm of the first platform and the second platform to obtain the encrypted user ID of each user; obtaining a second user data set corresponding to multiple users of the second platform; matching the first user data set and the second user data set based on the encrypted user ID, and generating a third user data set based on the first platform attribute data and label data of the matching user; performing classification training on the third user data set to obtain a model for credit risk assessment of the user. The embodiment of the present invention can perform joint modeling based on the third user data set to achieve user credit risk assessment based on multi-industry data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a credit and loan limit assessment method, apparatus, device and storage medium. Background Art

[0002] To meet market demand, a growing number of large enterprises are developing online lending businesses. This type of business often requires the use of third-party platforms to jointly assess users' credit risk based on data from both parties, thereby providing them with appropriate loan products and services. As big data security regulations become increasingly stringent, prohibiting companies from exporting sensitive user data outside of their data centers, multi-industry collaboration and joint modeling are becoming increasingly challenging.

[0003] The existing cooperation technology solutions are mainly based on the scoring of a single party's own data, and the partners screen customers based on the scoring data. The characteristics of this cooperation mechanism are complete connection, lack of deep data integration, and inability to fully tap the data value of multi-industry collaboration. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a credit and loan limit assessment method, apparatus, device and storage medium, aiming to assess user credit risk based on multi-industry data.

[0005] The technical solution of the embodiment of the present invention is achieved as follows:

[0006] An embodiment of the present invention provides a credit assessment model training method, comprising:

[0007] Obtaining a first user data set corresponding to multiple users of a first platform, the first user data set including: a user identity ID of each user and corresponding first platform attribute data, the first platform attribute data being user attribute data collected by a mobile communication operator;

[0008] Encrypting the user IDs in the first user data set based on the cross-encryption algorithm of the first platform and the second platform to obtain an encrypted user ID of each user;

[0009] Obtaining a second user data set corresponding to multiple users of the second platform, the second user data set including: non-sensitive data of each user; the non-sensitive data including: an encrypted user ID generated by encrypting the user ID based on a cross-encryption algorithm of the first platform and the second platform, and marking data for identifying whether the user has a financial default;

[0010] matching the first user data set and the second user data set based on the encrypted user ID, and generating a third user data set based on the first platform attribute data and the tag data of the matched users;

[0011] Classification training is performed on the third user data set to obtain a model for credit risk assessment of users.

[0012] An embodiment of the present invention further provides a credit assessment method, comprising:

[0013] Obtaining the first platform attribute data of the user to be evaluated;

[0014] Based on the first platform attribute data of the user to be evaluated and the model for performing credit risk assessment on the user generated by the credit assessment model training method according to the embodiment of the present invention, the credit risk result of the user to be evaluated is evaluated.

[0015] An embodiment of the present invention further provides a loan amount assessment model training method, comprising:

[0016] Obtaining a fourth user data set corresponding to multiple users of the second platform, the fourth user data set including: a user ID of each user, corresponding second platform attribute data, and marker data for identifying whether the user has a financial default, the second platform attribute data being user attribute data collected by a financial institution;

[0017] Encrypting the user IDs in the fourth user data set based on the cross-encryption algorithm of the first platform and the second platform to obtain an encrypted user ID of each user;

[0018] Obtaining a fifth user data set corresponding to multiple users of the first platform, the fifth user data set comprising: non-sensitive data of each user; the non-sensitive data comprising: an encrypted user ID generated by encrypting the user ID using a cross-encryption algorithm of the first and second platforms, and a default risk probability and credit score obtained by evaluating the user's attribute data on the first platform;

[0019] matching the fourth user dataset and the fifth user dataset based on the encrypted user ID, and generating a sixth user dataset based on the second platform attribute data, tag data, default risk probability, and credit score of the matched users;

[0020] A model for evaluating a user's loan amount is generated based on the sixth user dataset.

[0021] An embodiment of the present invention further provides a loan amount assessment method, comprising:

[0022] Obtaining the second platform attribute data of the user to be evaluated and the default risk probability and credit score corresponding to the user to be evaluated;

[0023] The model for evaluating a user's loan amount generated by the loan amount evaluation model training method described in an embodiment of the present invention is used to evaluate the user to be evaluated, and a loan amount evaluation result is obtained.

[0024] An embodiment of the present invention further provides a credit assessment model training device, comprising:

[0025] A first acquisition module is configured to acquire a first user data set corresponding to a plurality of users of a first platform, wherein the first user data set includes: a user ID of each user and corresponding first platform attribute data, wherein the first platform attribute data is user attribute data collected by a mobile communication operator;

[0026] a first encryption module, configured to encrypt the user IDs in the first user data set based on a cross-encryption algorithm of the first platform and the second platform to obtain an encrypted user ID of each user;

[0027] a second acquisition module, configured to acquire a second user data set corresponding to a plurality of users of the second platform, the second user data set comprising: non-sensitive data of each user; the non-sensitive data comprising: an encrypted user ID generated by encrypting the user ID of the user using a cross-encryption algorithm of the first platform and the second platform, and marking data for identifying whether the user has a financial default;

[0028] a first matching module, configured to match the first user data set and the second user data set based on the encrypted user ID, and generate a third user data set based on the first platform attribute data and the tag data of the matched users;

[0029] The first model generation module is used to perform classification training on the third user data set to obtain a model for credit risk assessment of users.

[0030] An embodiment of the present invention further provides a credit evaluation device, comprising:

[0031] A third acquisition module is used to obtain the first platform attribute data of the user to be evaluated;

[0032] The first evaluation module is configured to evaluate the credit risk result of the user to be evaluated based on the model for performing credit risk evaluation on the user generated by the credit evaluation model training device according to the embodiment of the present invention.

[0033] The embodiment of the present invention further provides a loan amount assessment model training device, comprising:

[0034] a fourth acquisition module, configured to acquire a fourth user data set corresponding to a plurality of users of the second platform, the fourth user data set comprising: a user ID of each user, corresponding second platform attribute data, and marker data for identifying whether the user has committed a financial default, the second platform attribute data being user attribute data collected by a financial institution;

[0035] a second encryption module, configured to encrypt the user IDs in the fourth user data set based on a cross encryption algorithm of the first platform and the second platform to obtain an encrypted user ID of each user;

[0036] a fifth acquisition module, configured to acquire a fifth user data set corresponding to multiple users on the first platform, the fifth user data set comprising: non-sensitive data of each user; the non-sensitive data comprising: an encrypted user ID generated by encrypting the user ID using a cross-encryption algorithm between the first and second platforms, and a default risk probability and credit score obtained by evaluating the user's attribute data on the first platform;

[0037] a second matching module, configured to match the fourth user dataset and the fifth user dataset based on the encrypted user ID, and generate a sixth user dataset based on the second platform attribute data, the tag data, the default risk probability, and the credit score of the matched users;

[0038] The second model generation module is used to generate a model for evaluating the user's loan amount based on the sixth user data set.

[0039] An embodiment of the present invention further provides a loan amount assessment device, comprising:

[0040] A sixth acquisition module, configured to acquire the second platform attribute data of the user to be evaluated and the default risk probability and credit score corresponding to the user to be evaluated;

[0041] The second evaluation module is used to evaluate the user to be evaluated using the model for evaluating the user's loan amount generated by the loan amount evaluation model training device according to an embodiment of the present invention to obtain a loan amount evaluation result.

[0042] An embodiment of the present invention also provides a credit assessment model training device, comprising: a processor and a memory for storing a computer program that can be run on the processor, wherein the processor is configured to execute the steps of the credit assessment model training method described in the embodiment of the present invention when running the computer program.

[0043] An embodiment of the present invention further provides a credit assessment device, comprising: a processor and a memory for storing a computer program that can be run on the processor, wherein the processor is configured to execute the steps of the credit assessment method described in the embodiment of the present invention when running the computer program.

[0044] An embodiment of the present invention also provides a loan amount assessment model training device, comprising: a processor and a memory for storing a computer program that can be run on the processor, wherein the processor is used to execute the steps of the loan amount assessment model training method described in the embodiment of the present invention when running the computer program.

[0045] An embodiment of the present invention further provides a loan amount assessment device, comprising: a processor and a memory for storing a computer program that can be run on the processor, wherein the processor is configured to execute the steps of the loan amount assessment method described in the embodiment of the present invention when running the computer program.

[0046] An embodiment of the present invention further provides a credit loan system, comprising: the credit assessment device described in the embodiment of the present invention and the loan amount assessment device described in the embodiment of the present invention, wherein the credit assessment device and the loan amount assessment device are communicatively connected.

[0047] An embodiment of the present invention further provides a storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method described in any embodiment of the present invention are implemented.

[0048] The technical solution provided by the embodiment of the present invention encrypts the user ID on the first platform side and the user ID on the second platform side based on the cross-encryption algorithm of the first platform and the second platform, matches the first user data set of the first platform and the second user data set of the second platform based on the encrypted user ID, thereby generating a third user data set after the integration of user data across platforms, and performs classification training based on the third user data set to obtain a model for credit risk assessment of users, which can realize joint assessment of user credit risk; in addition, the fourth user data set of the second platform and the fifth user data set of the first platform are matched based on the encrypted user ID to generate a sixth user data set after the integration of user data across platforms, and generates a model for assessing the user's loan amount based on the sixth user data set, which can realize joint assessment of user loan amounts and provide corresponding credit loan products and services. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Schematic diagram of the process of the credit assessment model training method according to an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of the principle of risk identification classifier training according to an embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram of the principle of performing default risk probability assessment based on a screening classification model according to an embodiment of the present invention;

[0052] Figure 4 Schematic diagram of a credit assessment method according to an embodiment of the present invention;

[0053] Figure 5 Schematic diagram of the process of the loan amount assessment model training method according to an embodiment of the present invention;

[0054] Figure 6 Schematic diagram of the process of loan amount assessment method according to an embodiment of the present invention;

[0055] Figure 7 A flowchart of a credit risk and loan limit assessment model training method according to an application embodiment of the present invention;

[0056] Figure 8 Schematic diagram of the composition of the first platform-side user data set Ua in an embodiment of the present invention;

[0057] Figure 9 Schematic diagram of the composition of the second platform-side user data set Ub in an embodiment of the present invention;

[0058] Figure 10 Schematic diagram of the process of matching and fusing two users in an embodiment of the present invention;

[0059] Figure 11 Schematic diagram of the composition of the matching user set Uab in an embodiment of the present invention;

[0060] Figure 12 Schematic diagram of the construction of the Rn-Ry hyperbola according to an embodiment of the present invention;

[0061] Figure 13 This is a schematic diagram of the structure of a credit assessment model training device according to an embodiment of the present invention;

[0062] Figure 14 This is a schematic diagram of the structure of a credit evaluation device according to an embodiment of the present invention;

[0063] Figure 15 This is a schematic diagram of the structure of a loan amount assessment model training device according to an embodiment of the present invention;

[0064] Figure 16 This is a schematic diagram of the structure of a loan amount assessment device according to an embodiment of the present invention;

[0065] Figure 17 This is a schematic diagram of the structure of a credit assessment model training device according to an embodiment of the present invention;

[0066] Figure 18 This is a schematic diagram of the structure of a credit assessment device according to an embodiment of the present invention;

[0067] Figure 19 This is a schematic diagram of the structure of a loan amount assessment model training device according to an embodiment of the present invention;

[0068] Figure 20 This is a schematic diagram of the structure of a loan amount assessment device according to an embodiment of the present invention;

[0069] Figure 21 Schematic diagram of the structure of the credit loan system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0070] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0071] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0072] In related technologies, due to restrictions on big data security regulations, different platforms are prohibited from exporting their sensitive data, leading to significant difficulties in business collaboration and joint modeling across different industries. Each platform often uses its own user data for scoring, while partners screen customers based on scoring data from third-party platforms. This collaborative mechanism is characterized by a complete cascade and a lack of deep data integration, which prevents the full realization of the data value of multi-industry collaboration.

[0073] Based on this, in various embodiments of the present invention, user data between different industries are securely and compliantly integrated, thereby realizing a joint assessment of user credit risk and loan amount based on user data from multiple industries, and then providing users with corresponding credit products and services based on the joint assessment results.

[0074] The embodiment of the present invention provides a credit assessment model training method, which is applied to a first platform, such as Figure 1 As shown, the credit assessment model training method includes:

[0075] Step 101: Acquire a first user data set corresponding to a plurality of users of a first platform, wherein the first user data set includes: a user ID of each user and corresponding first platform attribute data, wherein the first platform attribute data is user attribute data collected by a mobile communication operator;

[0076] Here, the first platform can be a mobile communication operator platform (also known as a telecommunications operator platform). The first platform obtains user data from different storage nodes within the platform to form a first user data set. For example, the original user data from different storage nodes under the first platform can be integrated into a unified user ID that uniquely identifies the user identity and corresponding first platform attribute data based on ID-mapping (user profiling) technology. The first platform attribute data may include: identity data, communication data, consumption data, location data, online behavior data and other indicators, and the above indicator set is represented by Xa. The first platform attribute data is user attribute data collected by the mobile communication operator and is used to reflect user credit behavior and risk. For example, characteristics such as the authenticity of mobile phone numbers and ID card numbers in identity data; characteristics such as loan issuance, overdue payments, and collection in communication data; consumption data reflects the user's consumption capacity; characteristics such as user stability reflected by changes in permanent locations in location data; and characteristics such as search keywords and financial APP usage behavior in online behavior data. The first user data set Ua is represented as Ua:[ID,Xa].

[0077] It should be noted that the number of users in the first user data set can be reasonably selected according to needs.

[0078] Step 102: Encrypt the user IDs in the first user data set based on the cross-encryption algorithm of the first platform and the second platform to obtain the encrypted user ID of each user;

[0079] Here, the second platform may be a financial institution platform. The first platform and the second platform each have their own encryption algorithms. The user IDs in the first user data set are encrypted based on the cross-encryption algorithms of the first platform and the second platform to obtain the encrypted user IDs of each user, including:

[0080] The first platform uses its own encryption algorithm to perform a first encryption process on the user ID of each user in the first user data set;

[0081] The first platform uses the encryption algorithm of the second platform to perform a second encryption process on the user ID after the first encryption process to obtain the encrypted user ID of each user.

[0082] Using the cross-encryption algorithm described above allows for serial encryption of user IDs, ensuring they are not leaked. It should be noted that user IDs on both the first and second platforms use a unified format, such as the user's mobile phone number or ID card number. The encryption algorithms used on both platforms should be unique and irreversible, such as the MD5 encryption algorithm.

[0083] In this way, the first user data set can be converted into Ua:[ID',Xa].

[0084] Step 103: Obtain a second user data set corresponding to multiple users of the second platform, where the second user data set includes: non-sensitive data of each user; the non-sensitive data includes: an encrypted user ID generated by encrypting the user ID using a cross-encryption algorithm of the first platform and the second platform, and tag data used to identify whether the user has financial default;

[0085] Here, the second user data set can be represented as [ID', Yb], where ID' is the encrypted user ID generated by the second platform using the cross-encryption algorithm of the first and second platforms. Yb is the tag data used to identify whether the user has financial defaults. Yb includes at least one of the following: whether the user is overdue, delinquent, or lost, and each tag data is marked by the second platform. Because ID' and Yb in the second user data set are non-sensitive data, the first platform obtains non-sensitive data transmitted by the second platform. This ensures that data transmission meets the requirements of big data security regulations and ensures that sensitive data does not leave the data center.

[0086] It should be noted that the number of users in the second user data set can be reasonably selected according to needs.

[0087] Step 104: Match the first user dataset and the second user dataset based on the encrypted user ID, and generate a third user dataset based on the first platform attribute data and the tag data of the matching users;

[0088] Here, [ID', Xa] and [ID', Yb] can be matched based on ID' to filter out a set of users that are shared by both. Xa and Yb of the user sets are then associated to generate a third user dataset, which can be represented as [ID', Xa, Yb]. The matching step involves determining users that are shared by both the first and second user datasets.

[0089] Step 105: Perform classification training on the third user data set to obtain a model for credit risk assessment of users.

[0090] Here, the third user data set is used as the modeling sample set, and through classification training, a model for credit risk assessment of users is obtained, thereby achieving the purpose of joint modeling based on the user data of the first platform and the second platform.

[0091] In some embodiments, performing classification training on the third user dataset to obtain a model for performing credit risk assessment on the user includes:

[0092] Performing classification training on the third user data set based on multiple classifiers using the first platform attribute data as an independent variable and the labeled data as a dependent variable to obtain a classification model;

[0093] Select a classification model based on classification evaluation indicators;

[0094] A first model for default risk probability assessment is generated based on the selected classification model.

[0095] In practical applications, the third user dataset [ID', Xa, Yb] is also preprocessed. Preprocessing includes data cleaning, sample balancing, feature discretization, feature selection, and feature dimensionality reduction, obtaining the preprocessed dataset Uab: [ID', Xa', Yb'].

[0096] In practical applications, classifiers include but are not limited to logistic regression, support vector machine, Bayesian classifier, decision tree, random forest, GBDT, neural network, etc. Xa' is used as the independent variable and Yb' is used as the dependent variable, and they are input into each classifier for training. Through training, the classification model (also called classifier) ​​corresponding to each classifier is obtained. For example, Figure 2 Classifiers M1-M7 are shown.

[0097] After obtaining classifiers M1-M7, the classification effect evaluation indicators are evaluated based on each classification model to obtain classification effect evaluation indicators C1-C7, and a classification model is selected based on the classification effect evaluation indicators of each classification model, and then a first model for default risk probability assessment is generated based on the selected classification model.

[0098] Here, the classification effect evaluation indicators include but are not limited to: accuracy, precision, recall, specificity, F1 score, and G-means.

[0099] The calculation method of each classification effect evaluation index is shown in Table 1 below.

[0100]

[0101] Table 1

[0102] Among them, TP, TN, FP, and FN are defined as follows: TP is the number of users who are actually "defaulting" and predicted to be "defaulting", TN is the number of users who are actually "not defaulting" and predicted to be "not defaulting", FP is the number of users who are actually "not defaulting" but predicted to be "defaulting", and FN is the number of users who are actually "defaulting" and predicted to be "not defaulting".

[0103] Compare the classification performance evaluation indicators (such as C1-C7) of each classification model and select the classification model with better performance from each classification model. For example, in the C1-C7 indicator set, if the values ​​of C1, C5, and C6 are all high, then the three classifiers M1, M5, and M6 are selected.

[0104] In an embodiment of the present invention, generating a first model for default risk probability assessment includes:

[0105] Based on the screened classification models, a classifier integration model is constructed to obtain the first model for default risk probability assessment.

[0106] Here, fusion can be performed through methods such as average value, weighted average, maximum value, etc. to obtain the first model after parallel fusion.

[0107] like Figure 3 As shown, the pre-processed data set [ID', Xa'] can be input into the first model to obtain the default risk probability Pa corresponding to each user.

[0108] like Figure 3 As shown, the recognition results corresponding to classifier M1 (default risk probability Py1, non-default risk probability Pn1), the recognition results corresponding to classifier M5 (default risk probability Py5, non-default risk probability Pn5), and the recognition results corresponding to classifier M6 (default risk probability Py6, non-default risk probability Pn6) are input into the "multi-classifier integrated result fusion device". The fusion device can fuse the results through methods such as average value, weighted average, and maximum value to obtain the user's final default risk probability Pa.

[0109] In some embodiments, the credit assessment model training method further includes:

[0110] A second model is constructed for converting the default risk probability into a credit score.

[0111] In actual applications, if the credit risk assessment result is the default risk probability Pa, it is difficult for users to understand. In order to facilitate explanation to users and more intuitively display the user's credit status, the default risk probability Pa can be converted into a credit score Sa through a second model (i.e., probability score converter). The second model can be selected from the following methods:

[0112] Sa=ln((1-Pa) / Pa)*F+C

[0113] Among them, 1-Pa is the probability that it is a high-quality user, Pa is the probability that it is a risky user, F is the factor coefficient of the linear transformation, and C is the offset constant that maps the score to a specific interval.

[0114] In this way, the first platform can generate a matching and fused third user dataset based on the first user dataset on the first platform side and the second user dataset of non-sensitive data on the second platform, and perform classification training on the third user dataset to obtain a model for credit risk assessment of users. Since the user data of the first platform and the second platform are integrated, the credit risk assessment is more scientific and accurate.

[0115] The embodiment of the present invention also provides a credit evaluation method, which is applied to the first platform, such as Figure 4 As shown, the credit assessment method includes:

[0116] Step 401: Obtain first platform attribute data of the user to be evaluated;

[0117] Step 402 : Evaluate the credit risk result of the user to be evaluated based on the first platform attribute data of the user to be evaluated and a model for performing credit risk evaluation on the user generated by a credit evaluation model training method.

[0118] Here, the model used for credit risk assessment is generated by the credit assessment model training method of the aforementioned embodiment.

[0119] In an embodiment of the present invention, the credit risk assessment result of the user to be assessed includes at least one of the following:

[0120] Assessing the default risk probability of the user to be assessed;

[0121] Evaluate the credit score of the user to be evaluated.

[0122] Specifically, the first platform attribute data of the user to be evaluated can be input into the aforementioned first model to obtain the user's default risk probability Pa. Furthermore, the default risk probability Pa can be converted into a credit score Sa based on the aforementioned second model.

[0123] Since the model for user credit risk assessment generated by the credit assessment model training method of the embodiment of the present invention integrates the user data of the first platform and the second platform, and fully taps the value of user data from different platforms, the credit assessment method of the embodiment of the present invention obtains a more scientific and accurate assessment result.

[0124] In order to evaluate the loan amount of the user, the embodiment of the present invention also provides a loan amount evaluation model training method, which is applied to the second platform, such as Figure 5 As shown, the loan amount assessment model training method includes:

[0125] Step 501: Acquire a fourth user dataset corresponding to multiple users of a second platform. The fourth user dataset includes: a user ID of each user, corresponding second platform attribute data, and tag data for identifying whether the user has a financial default. The second platform attribute data is user attribute data collected by a financial institution.

[0126] Here, the second platform can be a financial institution platform, which obtains user data from different storage nodes within the platform to form a fourth user data set. For example, the original user data from different storage nodes under the second platform can be integrated into a unified user ID that uniquely identifies the user identity and the corresponding second platform attribute data based on ID-mapping technology. The second platform attribute data may include: identity data, deposit and loan data, bank flow data, financial management data, etc., and the above indicator set is represented by Xb; the marker data used to identify whether the user has financial default refers to label data such as whether it is overdue, whether it is in arrears, whether it has been lost, etc., and the above indicator set is represented by Yb. The fourth data set Ub is expressed as Ub:[ID,Xb,Yb].

[0127] Step 502: Encrypt the user IDs in the fourth user data set based on the cross encryption algorithm of the first platform and the second platform to obtain the encrypted user ID of each user;

[0128] Here, the first platform may be a mobile communications operator platform. The first platform and the second platform each have their own encryption algorithms. The user IDs in the fourth user data set are encrypted based on the cross-encryption algorithms of the first platform and the second platform to obtain the encrypted user IDs of each user, including:

[0129] The second platform uses its own encryption algorithm to perform a first encryption process on the user ID of each user in the fourth user data set;

[0130] The second platform uses the encryption algorithm of the first platform to perform a second encryption process on the user ID after the first encryption process to obtain the encrypted user ID of each user.

[0131] Using the cross-encryption algorithm described above allows for serial encryption of user IDs, ensuring they are not leaked. It should be noted that user IDs on both the first and second platforms use a unified format, such as the user's mobile phone number or ID card number. The encryption algorithms used on both platforms should be unique and irreversible, such as the MD5 encryption algorithm.

[0132] In this way, the fourth user data set can be converted into Ub:[ID',Xb,Yb].

[0133] Step 503: Acquire a fifth user data set corresponding to multiple users on the first platform. The fifth user data set includes: non-sensitive data of each user; the non-sensitive data includes: an encrypted user ID generated by encrypting the user ID using a cross-encryption algorithm between the first and second platforms, and a default risk probability and credit score obtained by evaluating the user's attribute data on the first platform.

[0134] Here, the fifth user data set can be expressed as [ID', Pa, Sa]. Since ID' is the encrypted user ID, Pa is the user's default risk probability assessed by the first platform, and Sa is the user's credit score assessed by the first platform, the fifth user data set is all non-sensitive data on the first platform. This fifth user data set meets the security requirements for sensitive data transmission, thereby ensuring that data transmission meets the requirements of big data security regulations and that sensitive data does not leave the computer room.

[0135] Step 504: Match the fourth user dataset and the fifth user dataset based on the encrypted user ID, and generate a sixth user dataset based on the second platform attribute data, tag data, default risk probability, and credit score of the matched users;

[0136] Here, the fourth user data set [ID', Xb, Yb] can be matched with the fifth user data set [ID', Pa, Sa] to filter out the shared user sets, and associate Xb, Yb, Pa, and Sa of the user sets to generate a sixth user data set. The sixth user data set can be expressed as Uab: [ID', Xb, Yb, Pa, Sa].

[0137] Step 505: Generate a model for evaluating the user's loan amount based on the sixth user data set.

[0138] Here, a model for evaluating the user's loan amount is generated based on the sixth user data set, thereby achieving the purpose of joint modeling based on the user data of the first platform and the second platform.

[0139] In some embodiments, generating a model for evaluating a user's loan amount based on the sixth user dataset includes:

[0140] Counting the proportion of users with financial default and non-financial default corresponding to different credit scores based on the sixth user dataset;

[0141] Based on the mapping relationship between the difference between the proportion of users with financial defaults and the proportion of users with non-financial defaults and the credit score, the credit score threshold corresponding to the loan access threshold is determined.

[0142] In practical applications, based on the sixth user data set Uab:[ID',Xb,Yb,Pa,Sa], statistics and curve drawing of the distribution values ​​of high-quality and risky users are performed. For example, the cumulative proportion Rn of high-quality users (users with Yb=0) and the cumulative proportion Ry of risky users (users with Yb=1) corresponding to each credit score Sa are counted, and Rn and Ry form two curves.

[0143] The entry threshold is selected based on the Rn-Ry hyperbola. Generally, the threshold corresponding to the maximum distance between the two curves is selected as the entry threshold. Financial institutions can also make corresponding adjustments to the threshold based on their own business development needs.

[0144] In some embodiments, generating a model for evaluating a user's loan amount based on the sixth user dataset further includes:

[0145] extracting a seventh user data set having a loan access threshold from the sixth user data set based on the credit score threshold;

[0146] A loan limit assessment model for assessing the user's loan limit is generated for the seventh user data set based on the user's second platform attribute data, default risk probability, and credit score.

[0147] Here, the quota assessment model can be expressed as:

[0148] Db=f(Xb,Pa,Sa)

[0149] Among them, Db is the loan amount, Xb is the attribute data of the second platform, Pa is the default risk probability output by the first platform, Sa is the credit score output by the first platform, and f is the function corresponding to the loan amount assessment model.

[0150] In practice, loan limit assessments typically consider factors such as personal income, consumption, and the price range of consumer loan items. Financial institutions primarily use assessment methods, but specific approaches are not limited. Examples include linear calculations, expert scoring, and machine learning. Financial institutions determine the loan limit assessment model through a joint assessment of user data from both the primary and secondary platforms.

[0151] In this way, the second platform can generate a matching and fused sixth user dataset based on the fourth user dataset on the second platform side and the fifth user dataset of non-sensitive data of the first platform, and jointly model the sixth user dataset to obtain a model for evaluating the loan amount of users. Since the user data of the first platform and the second platform are integrated, the loan amount evaluation is more scientific and accurate.

[0152] The embodiment of the present invention also provides a loan amount evaluation method, which is applied to the second platform, such as Figure 6 As shown, the loan amount assessment method includes:

[0153] Step 601: Obtain the second platform attribute data of the user to be evaluated and the default risk probability and credit score corresponding to the user to be evaluated;

[0154] Step 602: Evaluate the user to be evaluated using the model for evaluating the user's loan amount generated by the loan amount evaluation model training method to obtain a loan amount evaluation result.

[0155] Here, the model used to assess a user's loan limit is generated using the loan limit assessment model training method described in the previous embodiment. The default risk probability and credit score corresponding to the user to be assessed are generated using the credit assessment method described in the previous embodiment, i.e., assessed by the first platform based on the trained credit risk assessment model.

[0156] In some embodiments, evaluating the user to be evaluated to obtain a loan amount evaluation result includes:

[0157] Determining that the credit score of the user to be evaluated is less than a credit score threshold, then determining that the user to be evaluated is not eligible for a loan; or

[0158] Determine that the credit score of the user to be evaluated is greater than or equal to a credit score threshold, and evaluate the loan limit of the user to be evaluated based on the credit limit evaluation model and the second platform attribute data, default risk probability and credit score of the user to be evaluated.

[0159] Since the model for evaluating user loan amounts generated by the loan amount evaluation model training method of the embodiment of the present invention integrates the user data of the first platform and the second platform, and fully taps the value of user data from different platforms, the loan amount evaluation method of the embodiment of the present invention obtains more scientific and accurate evaluation results.

[0160] The present invention will be described in further detail below with reference to application examples.

[0161] like Figure 7 As shown, in this application embodiment, the first platform is a telecom operator platform, and the second platform is a financial institution platform. The training method of the user credit risk and loan amount assessment model is jointly completed by the first platform and the second platform. The training method of the credit risk and loan amount assessment model includes:

[0162] Step 701, user data extraction;

[0163] Step 702: Match and merge users from both parties;

[0164] Step 703: constructing a user's credit risk assessment model;

[0165] Step 704: constructing a user's loan amount assessment model;

[0166] Step 705: Closed-loop feedback and iterative optimization.

[0167] The following is a detailed description of each step.

[0168] Step 701: User data extraction

[0169] The user data set extracted by the first platform is represented by Ua. The user data set Ua contains user attributes collected by telecom operators, such as identity data, communication data, consumption data, location data, Internet behavior data, etc. The above indicator set is represented by Xa. Data reflecting user credit behavior and risks can be extracted from the above telecom operator data: for example, features such as the authenticity of mobile phone numbers and ID card numbers in identity data; features such as loan issuance, overdue payments, and collection contained in communication data; consumption data reflects the user's consumption capacity; features such as user stability reflected by changes in permanent residence in location data; features such as search keywords and financial APP usage behavior contained in Internet behavior data. Ua is represented as Ua:[ID,Xa], such as Figure 8 shown.

[0170] The user data set extracted by the second platform is represented by Ub. The user data set Ub contains user attributes, financial business performance and default labels collected by financial institutions. The above indicator sets are represented by Xb and Yb. Financial institution user attributes include identity data, deposit and loan data, bank flow data, financial management data, etc., and the above indicator set is represented by Xb; financial performance and default labels refer to label data such as whether it is overdue, whether it is in arrears, whether it has been lost, etc., and the above indicator set is represented by Yb. Ub is represented as Ub:[ID,Xb,Yb], such as Figure 9 shown.

[0171] Step 702: Match and merge users from both parties

[0172] like Figure 10 As shown, step 702 includes:

[0173] 1) ID cross encryption processing

[0174] The first and second platforms encrypt the user IDs in the user sets Ua and Ub, respectively, using a cross-encryption algorithm. This implements "cross-encryption" to ensure user data privacy. The encryption algorithms used by the first and second platforms should be unique and irreversible.

[0175] For example, the first platform uses its own encryption algorithm (encryption algorithm A) to perform a first encryption process on the user ID of each user in the user set Ua, and then uses the encryption algorithm of the second platform (encryption algorithm B) to perform a second encryption process on the user ID after the first encryption process to obtain the encrypted user ID of each user, that is, ID', and the user set Ua is converted to Ua:[ID',Xa].

[0176] For example, the second platform uses its own encryption algorithm (encryption algorithm B) to perform a first encryption process on the user ID of each user in the user set Ub, and then uses the encryption algorithm of the first platform (encryption algorithm A) to perform a second encryption process on the user ID after the first encryption process to obtain the encrypted user ID of each user, that is, ID', and the user set Ub is converted to Ub: [ID', Xb, Yb].

[0177] 2) Non-sensitive field transmission

[0178] The second platform transmits the cross-encrypted "non-sensitive fields" in the user set to the first platform. The cross-encrypted user set Ub on the second platform contains [ID', Xb, Yb]. The original ID is encrypted and converted to ID', meeting the transmission requirements. Xb contains user privacy data and is not transmitted. Yb is non-sensitive labeled data. Therefore, the second platform transmits the non-sensitive data [ID', Yb] to the first platform. This cross-encryption and non-sensitive data transmission ensures that both parties meet user data security requirements and that sensitive data never leaves the data center.

[0179] 3) User ID matching and data fusion

[0180] After the first platform obtains the non-sensitive data set transmitted by the second platform, it matches the cross-encrypted ID's of both users to select the common user set as the modeling sample set. For example, the user set Ua of the first platform has 20 million users, and the user set Ub of the second platform has 1 million users. After matching, the common user set Uab is obtained, for example, 600,000 users. The output matching user set Uab includes user ID', telecom operator attribute Xa, and financial institution performance and default label Yb, such as Figure 11 shown.

[0181] Step 703: Construction of user credit risk assessment model

[0182] In step 703, based on the cross-matched user set Uab: [ID', Xa, Yb], the user credit risk identification model is trained, the classifier effectiveness is evaluated and screened, and the identification model integration is constructed using statistical and machine learning methods. The independent variable in the input data is the telecom operator attribute Xa, and the dependent variable is the financial institution's default label data Yb. The risk identification model is based on a multi-classifier integration model. The model training, evaluation, and integration in this step are primarily performed by the telecom operator based on the fused data from both parties. Specifically, it includes:

[0183] 1) Data preprocessing;

[0184] 2) Risk identification classifier training;

[0185] 3) Classification model effect evaluation and screening;

[0186] 4) Classification model integration;

[0187] 5) Probability score converter construction.

[0188] The specific implementation process of step 703 can refer to the description of the aforementioned step 105 and will not be repeated here.

[0189] Step 704: Construction of the user's loan amount assessment model

[0190] Here, step 704 specifically includes:

[0191] 1) Matching and fusion of users from both sides

[0192] The second platform receives the non-sensitive data (ie, the aforementioned fifth user data set) transmitted by the first platform. The non-sensitive data transmitted by the first platform may include: encrypted user ID, default risk probability Pa, and credit score Sa.

[0193] The second platform matches the user dataset [ID', Xb, Yb] with the user dataset [ID', Pa, Sa] to obtain the fused dataset Uab: [ID', Xb, Yb, Pa, Sa].

[0194] 2) Construction of the first-level risk control model

[0195] The goal of the first-level risk control model is to assess the entry threshold for credit loans. Users whose scores are above the threshold are eligible for loans, while users whose scores are below the threshold are not eligible for loans.

[0196] Based on the user set Uab:[ID',Xb,Yb,Pa,Sa], statistics and curve drawing of the distribution values ​​of high-quality and risky users are performed: the cumulative proportion Rn of high-quality users (users with Yb=0) and the cumulative proportion Ry of risky users (users with Yb=1) corresponding to each credit score value Sa are counted, and Rn and Ry form two curves.

[0197] The entry threshold is selected based on the Rn-Ry hyperbola. Generally, the threshold corresponding to the maximum distance between the two curves is selected as the entry threshold, such as Figure 12 As shown in the figure, the maximum difference of 0.64 corresponds to a threshold of 600 points, meaning that users with a score of 600 or above are eligible to apply for loans. Financial institutions can also adjust the threshold based on their business development needs.

[0198] 3) Construction of secondary risk control model

[0199] Here, eligible users undergo a loan limit assessment. This assessment is based on a model constructed from data from both parties, including the financial institution's proprietary data Xb, the telecom operator's risk assessment probability Pa, and the credit score Sa. Factors typically considered during the loan limit assessment process include personal income, consumption, and the price range of consumer loan items. The assessment method is primarily determined by the financial institution, but specific methods are not limited. Examples include linear calculations, expert scoring, and machine learning. After the financial institution evaluates the data from both parties, it outputs the user's loan limit Db, where Db = f(Xb, Pa, Sa).

[0200] Step 705: Closed-loop feedback and iterative optimization.

[0201] Based on the evaluations in steps 701-704 above, a credit risk identification model, credit loan entry threshold, and user loan limit assessment model are obtained. Credit risk and loan limit assessments can be performed for the users to be assessed, allowing for the provision of credit services such as "number loans" to those who meet the requirements. After users apply for and use credit services, new performance and default behavior data is generated. This performance and default behavior data is fed back to telecom operators and financial institutions, respectively, allowing for iteration and optimization of the credit risk assessment model and the loan limit assessment model.

[0202] In order to implement the credit assessment model training method of the embodiment of the present invention, the embodiment of the present invention also provides a credit assessment model training device. The credit assessment model training device corresponds to the above-mentioned credit assessment model training method, and each step in the above-mentioned credit assessment model training method embodiment is also fully applicable to the present credit assessment model training device embodiment.

[0203] like Figure 13 As shown, the credit assessment model training device includes: a first acquisition module 1301, a first encryption module 1302, a second acquisition module 1303, a first matching module 1304 and a first model generation module 1305.

[0204] The first acquisition module 1301 is used to obtain a first user data set corresponding to multiple users of the first platform, where the first user data set includes: a user ID of each user and corresponding first platform attribute data, where the first platform attribute data is user attribute data collected by the mobile communication operator; the first encryption module 1302 is used to encrypt the user ID in the first user data set based on a cross-encryption algorithm of the first platform and the second platform to obtain an encrypted user ID of each user; the second acquisition module 1303 is used to obtain a second user data set corresponding to multiple users of the second platform, where the second user data set includes: non-sensitive data of each user; the non-sensitive data includes: an encrypted user ID generated by encrypting the user ID of the user based on a cross-encryption algorithm of the first platform and the second platform, and marker data for identifying whether the user has financial default; the first matching module 1304 is used to match the first user data set and the second user data set based on the encrypted user ID, and generate a third user data set based on the first platform attribute data and marker data of the matched users; the first model generation module 1305 is used to perform classification training on the third user data set to obtain a model for credit risk assessment of users.

[0205] In some embodiments, the first model generation module 1305 is specifically configured to:

[0206] Performing classification training on the third user data set based on multiple classifiers using the first platform attribute data as an independent variable and the labeled data as a dependent variable to obtain a classification model;

[0207] Select a classification model based on classification evaluation indicators;

[0208] A first model for default risk probability assessment is generated based on the selected classification model.

[0209] In some embodiments, the first model generation module 1305 is specifically configured to:

[0210] A second model is constructed for converting the default risk probability into a credit score.

[0211] In actual application, the first acquisition module 1301, the first encryption module 1302, the second acquisition module 1303, the first matching module 1304, and the first model generation module 1305 can be implemented by a processor in the credit assessment model training device. Of course, the processor needs to run the computer program in the memory to realize its functions.

[0212] It should be noted that the credit assessment model training device provided in the above embodiment, when performing credit assessment model training, uses the aforementioned division of program modules as an example only. In actual applications, the aforementioned processing can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program modules to complete all or part of the aforementioned processing. Furthermore, the credit assessment model training device provided in the above embodiment and the credit assessment model training method embodiment are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0213] In order to implement the credit assessment method of the embodiment of the present invention, the embodiment of the present invention also provides a credit assessment device. The credit assessment device corresponds to the above-mentioned credit assessment method, and each step in the above-mentioned credit assessment method embodiment is also fully applicable to the present credit assessment device embodiment.

[0214] like Figure 14 As shown, the credit assessment device includes: a third acquisition module 1401 and a first assessment module 1402, wherein the third acquisition module 1401 is used to obtain the first platform attribute data of the user to be assessed; the first assessment module 1402 is used to generate a model for credit risk assessment of the user by the credit assessment model training device described in an embodiment of the present invention, and assess the credit risk result of the user to be assessed.

[0215] In some embodiments, the first evaluation module 1402 is specifically configured to:

[0216] Assessing the default risk probability of the user to be assessed;

[0217] Evaluate the credit score of the user to be evaluated.

[0218] In actual application, the third acquisition module 1401 and the first evaluation module 1402 can be implemented by a processor in the credit evaluation device. Of course, the processor needs to run the computer program in the memory to implement its functions.

[0219] In order to implement the loan amount assessment model training method of the embodiment of the present invention, the embodiment of the present invention also provides a loan amount assessment model training device. The loan amount assessment model training device corresponds to the above-mentioned loan amount assessment model training method, and each step in the above-mentioned loan amount assessment model training method embodiment is also fully applicable to the present loan amount assessment model training device embodiment.

[0220] like Figure 15 As shown, the loan amount assessment model training device includes: a fourth acquisition module 1501, a second encryption module 1502, a fifth acquisition module 1503, a second matching module 1504 and a second model generation module 1505.

[0221] A fourth acquisition module 1501 is used to acquire a fourth user dataset corresponding to multiple users of the second platform. The fourth user dataset includes: a user ID of each user, corresponding second platform attribute data, and tag data used to identify whether the user has a financial default, where the second platform attribute data is user attribute data collected by a financial institution. A second encryption module 1502 is used to encrypt the user IDs in the fourth user dataset based on a cross-encryption algorithm of the first and second platforms to obtain an encrypted user ID of each user. A fifth acquisition module 1503 is used to acquire a fifth user dataset corresponding to multiple users of the first platform. The fifth user dataset includes: non-sensitive data of each user. The non-sensitive data includes: an encrypted user ID generated by encrypting the user ID of the user based on the cross-encryption algorithm of the first and second platforms, and a default risk probability and credit score obtained by evaluating the user's first platform attribute data. A second matching module 1504 is used to match the fourth user dataset and the fifth user dataset based on the encrypted user ID, and generate a sixth user dataset based on the second platform attribute data, tag data, default risk probability, and credit score of the matched users. A second model generation module 1505 is used to generate a model for evaluating a user's loan amount based on the sixth user dataset.

[0222] In some embodiments, the second model generation module 1505 is specifically configured to:

[0223] Counting the proportion of users with financial default and non-financial default corresponding to different credit scores based on the sixth user dataset;

[0224] Based on the mapping relationship between the difference between the proportion of users with financial defaults and the proportion of users with non-financial defaults and the credit score, the credit score threshold corresponding to the loan access threshold is determined.

[0225] In some embodiments, the second model generation module 1505 is specifically configured to:

[0226] extracting a seventh user data set having a loan access threshold from the sixth user data set based on the credit score threshold;

[0227] A loan limit assessment model for assessing the user's loan limit is generated for the seventh user data set based on the user's second platform attribute data, default risk probability, and credit score.

[0228] In actual application, the fourth acquisition module 1501, the second encryption module 1502, the fifth acquisition module 1503, the second matching module 1504, and the second model generation module 1505 can be implemented by a processor in the loan amount assessment model training device. Of course, the processor needs to run the computer program in the memory to implement its functions.

[0229] It should be noted that the loan limit assessment model training device provided in the above embodiment only uses the division of the above-mentioned program modules as an example to illustrate the training of the loan limit assessment model. In actual applications, the above-mentioned processing can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program modules to complete all or part of the above-described processing. In addition, the loan limit assessment model training device provided in the above embodiment and the loan limit assessment model training method embodiment are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0230] In order to implement the loan amount assessment method of the embodiment of the present invention, the embodiment of the present invention also provides a loan amount assessment device, which corresponds to the above-mentioned loan amount assessment method. The various steps in the above-mentioned loan amount assessment method embodiment are also fully applicable to the loan amount assessment embodiment of the present invention.

[0231] like Figure 16 As shown, the loan amount assessment device includes: a sixth acquisition module 1601 and a second assessment module 1602, wherein the sixth acquisition module 1601 is used to obtain the second platform attribute data of the user to be assessed and the default risk probability and credit score corresponding to the user to be assessed; the second assessment module 1602 is used to evaluate the user to be assessed using the model for assessing the user's loan amount generated by the loan amount assessment model training device described in an embodiment of the present invention to obtain a loan amount assessment result.

[0232] In some embodiments, the second evaluation module 1602 is specifically configured to:

[0233] Determining that the credit score of the user to be evaluated is less than a credit score threshold, then determining that the user to be evaluated is not eligible for a loan; or

[0234] Determine that the credit score of the user to be evaluated is greater than or equal to a credit score threshold, and evaluate the loan limit of the user to be evaluated based on the credit limit evaluation model and the second platform attribute data, default risk probability and credit score of the user to be evaluated.

[0235] In actual application, the sixth acquisition module 1601 and the second evaluation module 1602 can be implemented by a processor in the loan amount evaluation device. Of course, the processor needs to run the computer program in the memory to implement its functions.

[0236] Based on the hardware implementation of the above program modules, and in order to implement the credit assessment model training method of an embodiment of the present invention, an embodiment of the present invention also provides a credit assessment model training device. Figure 17 Only the exemplary structure of the credit assessment model training device is shown, not the entire structure. Figure 17 Partial or complete structure shown.

[0237] like Figure 17 As shown, the credit assessment model training device 1700 provided by the embodiment of the present invention includes: at least one processor 1701, a memory 1702, a user interface 1703 and at least one network interface 1704. The various components in the credit assessment model training device 1700 are coupled together through a bus system 1705. It can be understood that the bus system 1705 is used to achieve connection and communication between these components. In addition to including a data bus, the bus system 1705 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Figure 17 Various buses are labeled as bus system 1705.

[0238] The user interface 1703 may include a display, a keyboard, a mouse, a trackball, a click wheel, keys, buttons, a touch pad or a touch screen.

[0239] The memory 1702 in this embodiment of the present invention is used to store various types of data to support the operation of the credit assessment model training device. Examples of such data include any computer program used to operate on the credit assessment model training device.

[0240] The credit assessment model training method disclosed in the embodiments of the present invention can be applied to or implemented by processor 1701. Processor 1701 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the credit assessment model training method can be completed by hardware integrated logic circuits or software instructions in processor 1701. Processor 1701 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. Processor 1701 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium located in memory 1702. Processor 1701 reads information from memory 1702 and, in conjunction with its hardware, completes the steps of the credit assessment model training method provided in the embodiments of the present invention.

[0241] In an exemplary embodiment, the credit assessment model training device 1700 can be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), FPGAs, general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned method.

[0242] Based on the hardware implementation of the above program modules, and in order to implement the credit evaluation method of the embodiment of the present invention, the embodiment of the present invention further provides a credit evaluation device. Figure 18 Only an exemplary structure of the credit assessment device is shown, not all structures, and it can be implemented as needed. Figure 18 Partial or complete structure shown.

[0243] like Figure 18 As shown, the credit assessment device 1800 provided by the embodiment of the present invention includes: at least one processor 1801, a memory 1802, a user interface 1803, and at least one network interface 1804. The various components in the credit assessment device 1800 are coupled together via a bus system 1805. It can be understood that the bus system 1805 is used to achieve connection and communication between these components. In addition to including a data bus, the bus system 1805 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, the bus system 1805 is not described in detail. Figure 18 Various buses are labeled as bus system 1805.

[0244] The user interface 1803 may include a display, a keyboard, a mouse, a trackball, a click wheel, keys, buttons, a touch pad or a touch screen.

[0245] The memory 1802 in this embodiment of the present invention is used to store various types of data to support the operation of the credit assessment device. Examples of such data include any computer program used to operate on the credit assessment device.

[0246] The credit assessment method disclosed in the embodiments of the present invention can be applied to or implemented by processor 1801. Processor 1801 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the credit assessment method can be completed by hardware integrated logic circuits or software instructions within processor 1801. Processor 1801 may be a general-purpose processor, a digital signal processor, or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. Processor 1801 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules within the decoding processor. The software modules may be located in a storage medium located in memory 1802. Processor 1801 reads information from memory 1802 and, in conjunction with its hardware, completes the steps of the credit assessment method provided in the embodiments of the present invention.

[0247] In an exemplary embodiment, the credit evaluation device 1800 may be implemented by one or more ASICs, DSPs, PLDs, CPLDs, FPGAs, general-purpose processors, controllers, MCUs, Microprocessors, or other electronic components to perform the aforementioned method.

[0248] Based on the hardware implementation of the above program modules, and in order to implement the loan amount assessment model training method of the embodiment of the present invention, the embodiment of the present invention also provides a loan amount assessment model training device. Figure 19 Only the exemplary structure of the loan amount assessment model training device is shown, not the entire structure. Figure 19 Partial or complete structure shown.

[0249] like Figure 19 As shown, the loan amount assessment model training device 1900 provided by the embodiment of the present invention includes: at least one processor 1901, a memory 1902, a user interface 1903 and at least one network interface 1904. The various components in the loan amount assessment model training device 1900 are coupled together through a bus system 1905. It can be understood that the bus system 1905 is used to achieve connection and communication between these components. In addition to including a data bus, the bus system 1905 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Figure 19 Various buses are labeled as bus system 1905.

[0250] The user interface 1903 may include a display, a keyboard, a mouse, a trackball, a click wheel, keys, buttons, a touch pad or a touch screen.

[0251] The memory 1902 in this embodiment of the present invention is used to store various types of data to support the operation of the loan amount assessment model training device. Examples of such data include: any computer program used to operate on the loan amount assessment model training device.

[0252] The loan amount assessment model training method disclosed in the embodiments of the present invention can be applied to or implemented by processor 1901. Processor 1901 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the loan amount assessment model training method can be completed by hardware integrated logic circuits or software instructions in processor 1901. Processor 1901 can be a general-purpose processor, a digital signal processor, or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. Processor 1901 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software modules can be located in a storage medium located in memory 1902. Processor 1901 reads information from memory 1902 and, in conjunction with its hardware, completes the steps of the loan amount assessment model training method provided in the embodiments of the present invention.

[0253] In an exemplary embodiment, the loan amount assessment model training device 1900 can be implemented by one or more ASICs, DSPs, PLDs, CPLDs, FPGAs, general-purpose processors, controllers, MCUs, Microprocessors, or other electronic components to perform the aforementioned method.

[0254] Based on the hardware implementation of the above program modules, and in order to implement the loan amount evaluation method of the embodiment of the present invention, the embodiment of the present invention also provides a loan amount evaluation device. Figure 20 Only the exemplary structure of the loan amount assessment device is shown, not the entire structure, and can be implemented as needed. Figure 20 Partial or complete structure shown.

[0255] like Figure 20As shown, the loan amount assessment device 2000 provided by the embodiment of the present invention includes: at least one processor 2001, a memory 2002, a user interface 2003 and at least one network interface 2004. The various components in the loan amount assessment device 2000 are coupled together via a bus system 2005. It can be understood that the bus system 2005 is used to achieve connection and communication between these components. In addition to including a data bus, the bus system 2005 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the bus system 2005 is not shown in FIG. Figure 20 Various buses are labeled as bus system 2005.

[0256] The user interface 2003 may include a display, a keyboard, a mouse, a trackball, a click wheel, keys, buttons, a touch pad or a touch screen.

[0257] The memory 2002 in the embodiment of the present invention is used to store various types of data to support the operation of the loan amount assessment device. Examples of such data include: any computer program used to operate on the loan amount assessment device.

[0258] The loan amount assessment method disclosed in the embodiments of the present invention can be applied to or implemented by processor 2001. Processor 2001 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the loan amount assessment method can be completed by hardware integrated logic circuits or software instructions in processor 2001. Processor 2001 may be a general-purpose processor, a digital signal processor, or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. Processor 2001 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium located in memory 2002. Processor 2001 reads information from memory 2002 and, in conjunction with its hardware, completes the steps of the loan amount assessment method provided in the embodiments of the present invention.

[0259] In an exemplary embodiment, the loan amount evaluation device 2000 may be implemented by one or more ASICs, DSPs, PLDs, CPLDs, FPGAs, general-purpose processors, controllers, MCUs, Microprocessors, or other electronic components to perform the aforementioned method.

[0260] It is understood that memories 1702, 1802, 1902, and 2002 may be volatile or non-volatile memories, or may include both volatile and non-volatile memories. Non-volatile memories may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disk, or compact disc read-only memory (CD-ROM); magnetic surface memory may include magnetic disk storage or magnetic tape storage. Volatile memories may include random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM).The memories described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memories.

[0261] The embodiment of the present invention also provides a credit loan system, such as Figure 21 As shown, the credit loan system includes: the credit assessment device 1800 of the aforementioned embodiment and the loan amount assessment device 2000 of the aforementioned embodiment, wherein the credit assessment device 1800 and the loan amount assessment device 2000 are communicatively connected. The credit assessment method performed by the credit assessment device 1800 and the loan amount assessment method performed by the loan amount assessment device 2000 refer to the aforementioned method embodiments and are not described in detail here.

[0262] In an exemplary embodiment, the present invention further provides a storage medium, namely a computer storage medium, which may be a computer-readable storage medium, for example, a memory 1702 storing a computer program, which may be executed by the processor 1701 of the credit assessment model training device 1700 to complete the steps of the credit assessment model training method according to the embodiment of the present invention; another example, a memory 1802 storing a computer program, which may be executed by the processor 1801 of the credit assessment device 1800 to complete the steps of the credit assessment method according to the embodiment of the present invention; another example, a memory 1902 storing a computer program, which may be executed by the processor 1901 of the loan amount assessment model training device 1900 to complete the steps of the loan amount assessment model training method according to the embodiment of the present invention; another example, a memory 2002 storing a computer program, which may be executed by the processor 2001 of the loan amount assessment device 2000 to complete the steps of the loan amount assessment method according to the embodiment of the present invention; the computer-readable storage medium may be a memory such as a ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface storage, optical disk, or CD-ROM.

[0263] It should be noted that: "first", "second", etc. are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0264] In addition, the technical solutions described in the embodiments of the present invention can be arbitrarily combined without conflict.

[0265] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A credit assessment model training method, It is characterized in that include: Acquire a first user data set corresponding to a plurality of users of a first platform, wherein the first user data set includes: a user identity ID of each user and corresponding first platform attribute data, wherein the first platform attribute data is user attribute data collected by a mobile communication operator; Encrypting the user IDs in the first user data set based on the cross encryption algorithm of the first platform and the second platform to obtain encrypted user IDs of each user; Acquire a second user data set corresponding to a plurality of users of the second platform, the second user data set comprising: non-sensitive data of each user; the non-sensitive data comprising: an encrypted user ID generated by encrypting the user ID of the user based on a cross encryption algorithm of the first platform and the second platform, and marking data for identifying whether the user has financial default; matching the first user data set and the second user data set based on the encrypted user ID, and generating a third user data set based on the first platform attribute data and the tag data of the matching users; Performing classification training on the third user data set to obtain a model for credit risk assessment of users; The step of encrypting the user IDs in the first user data set based on the cross encryption algorithm of the first platform and the second platform to obtain the encrypted user IDs of the users includes: The first platform uses its own encryption algorithm to perform a first encryption process on the user ID of each user in the first user data set; The first platform uses the encryption algorithm of the second platform to perform a second encryption process on the user ID after the first encryption process to obtain an encrypted user ID of each user.

2. The method according to claim 1, It is characterized in that The performing classification training on the third user data set to obtain a model for credit risk assessment of the user includes: For the third user data set, the first platform attribute data is used as an independent variable and the labeled data is used as a dependent variable, and classification training is performed based on multiple classifiers to obtain a classification model; Select a classification model based on classification evaluation indicators; A first model for default risk probability assessment is generated based on the selected classification model.

3. The method according to claim 2, It is characterized in that The method further comprises: A second model is constructed for converting the probability of default risk into a credit score.

4. A credit assessment method, It is characterized in that include: Obtaining the first platform attribute data of the user to be evaluated; Based on the first platform attribute data of the user to be evaluated and the model for credit risk evaluation of the user generated by the method according to any one of claims 1 to 3, the credit risk result of the user to be evaluated is evaluated.

5. The method according to claim 4, It is characterized in that The credit risk result of the user to be evaluated includes at least one of the following: Assessing the default risk probability of the user to be assessed; Evaluate the credit score of the user to be evaluated.

6. A loan amount assessment model training method, It is characterized in that include: Acquire a fourth user data set corresponding to a plurality of users of the second platform, the fourth user data set comprising: a user ID of each user, corresponding second platform attribute data, and marking data for identifying whether the user is in financial default, the second platform attribute data being user attribute data collected by a financial institution; Encrypting the user IDs in the fourth user data set based on the cross encryption algorithm of the first platform and the second platform to obtain encrypted user IDs of each user; Acquire a fifth user data set corresponding to multiple users of the first platform, the fifth user data set including: non-sensitive data of each user; the non-sensitive data including: an encrypted user ID generated by encrypting the user ID of the user based on a cross encryption algorithm of the first platform and the second platform, and a default risk probability and a credit score obtained by evaluating the attribute data of the user on the first platform; Matching the fourth user data set and the fifth user data set based on the encrypted user ID, and generating a sixth user data set based on the second platform attribute data, the marking data, the default risk probability and the credit score of the matching user; generating a model for evaluating a user's loan amount based on a sixth user data set; The step of encrypting the user IDs in the fourth user data set based on the cross encryption algorithm of the first platform and the second platform to obtain the encrypted user IDs of the users includes: The second platform uses its own encryption algorithm to perform a first encryption process on the user ID of each user in the fourth user data set; The second platform uses the encryption algorithm of the first platform to perform a second encryption process on the user ID after the first encryption process to obtain an encrypted user ID of each user.

7. The method according to claim 6, It is characterized in that The generating a model for evaluating the loan amount of the user based on the sixth user data set includes: Based on the sixth user data set, statistics are given on the proportions of users who have financial defaults and those who have non-financial defaults corresponding to different credit scores; Based on the mapping relationship between the difference between the proportion of users with financial defaults and the proportion of users with non-financial defaults and the credit score, the credit score threshold corresponding to the loan access threshold is determined.

8. The method according to claim 7, It is characterized in that The generating a model for evaluating the loan amount of the user based on the sixth user data set further includes: extracting a seventh user data set having a loan access threshold from the sixth user data set based on the credit score threshold; A loan amount assessment model for assessing the user's loan amount is generated for the seventh user data set based on the user's second platform attribute data, default risk probability and credit score.

9. A loan amount assessment method, It is characterized in that include: Obtaining the second platform attribute data of the user to be evaluated and the default risk probability and credit score corresponding to the user to be evaluated; The user to be evaluated is evaluated using the model for evaluating the user's loan amount generated by the method according to any one of claims 6 to 8 to obtain a loan amount evaluation result.

10. The method according to claim 9, It is characterized in that The step of evaluating the user to be evaluated and obtaining a loan amount evaluation result includes: Determining that the credit score of the user to be evaluated is less than a credit score threshold, then determining that the user to be evaluated is not eligible for a loan; or, Determine that the credit score of the user to be evaluated is greater than or equal to a credit score threshold, and evaluate the loan amount of the user to be evaluated based on the credit limit evaluation model and the second platform attribute data, default risk probability and credit score of the user to be evaluated.

11. A credit assessment model training device, It is characterized in that include: A first acquisition module, configured to acquire a first user data set corresponding to a plurality of users of a first platform, wherein the first user data set includes: a user ID of each user and corresponding first platform attribute data, wherein the first platform attribute data is user attribute data collected by a mobile communication operator; A first encryption module, used to encrypt the user IDs in the first user data set based on a cross encryption algorithm of the first platform and the second platform to obtain an encrypted user ID of each user; A second acquisition module is used to acquire a second user data set corresponding to a plurality of users of the second platform, wherein the second user data set includes: non-sensitive data of each user; the non-sensitive data includes: an encrypted user ID generated by encrypting the user ID of the user based on a cross encryption algorithm of the first platform and the second platform, and marking data for identifying whether the user has financial default; A first matching module, configured to match the first user data set and the second user data set based on the encrypted user ID, and generate a third user data set based on the first platform attribute data and the tag data of the matching user; A first model generation module, used for performing classification training on the third user data set to obtain a model for credit risk assessment of the user; The step of encrypting the user IDs in the first user data set based on the cross encryption algorithm of the first platform and the second platform to obtain the encrypted user IDs of the users includes: The first platform uses its own encryption algorithm to perform a first encryption process on the user ID of each user in the first user data set; The first platform uses the encryption algorithm of the second platform to perform a second encryption process on the user ID after the first encryption process to obtain an encrypted user ID of each user.

12. A credit assessment device, It is characterized in that include: A third acquisition module is used to acquire the first platform attribute data of the user to be evaluated; The first evaluation module is used to evaluate the credit risk result of the user to be evaluated based on the model for performing credit risk evaluation on the user generated by the credit evaluation model training device as claimed in claim 11.

13. A loan amount assessment model training device, It is characterized in that include: a fourth acquisition module, configured to acquire a fourth user data set corresponding to a plurality of users of the second platform, wherein the fourth user data set includes: a user ID of each user, corresponding second platform attribute data, and marking data for identifying whether the user is in financial default, wherein the second platform attribute data is user attribute data collected by a financial institution; A second encryption module, used for encrypting the user IDs in the fourth user data set based on the cross encryption algorithm of the first platform and the second platform to obtain the encrypted user IDs of each user; a fifth acquisition module, configured to acquire a fifth user data set corresponding to a plurality of users of the first platform, wherein the fifth user data set includes: non-sensitive data of each user; the non-sensitive data includes: an encrypted user ID generated by encrypting the user ID of the user based on a cross encryption algorithm of the first platform and the second platform, and a default risk probability and a credit score obtained by evaluating the attribute data of the user on the first platform; A second matching module, configured to match the fourth user data set and the fifth user data set based on the encrypted user ID, and generate a sixth user data set based on the second platform attribute data, the marking data, the default risk probability and the credit score of the matching user; A second model generating module, used to generate a model for evaluating a user's loan amount based on a sixth user data set; The step of encrypting the user IDs in the fourth user data set based on the cross encryption algorithm of the first platform and the second platform to obtain the encrypted user IDs of the users includes: The second platform uses its own encryption algorithm to perform a first encryption process on the user ID of each user in the fourth user data set; The second platform uses the encryption algorithm of the first platform to perform a second encryption process on the user ID after the first encryption process to obtain an encrypted user ID of each user.

14. A loan amount assessment device, It is characterized in that include: A sixth acquisition module, used to acquire the second platform attribute data of the user to be evaluated and the default risk probability and credit score corresponding to the user to be evaluated; The second evaluation module is used to evaluate the user to be evaluated using the model for evaluating the user's loan amount generated by the loan amount evaluation model training device as claimed in claim 13 to obtain a loan amount evaluation result.

15. A credit assessment model training device, It is characterized in that include: A processor and a memory for storing a computer program that can be executed on the processor, wherein: The processor is used to execute the steps of the method according to any one of claims 1 to 3 when running a computer program.

16. A credit assessment device, It is characterized in that include: A processor and a memory for storing a computer program that can be executed on the processor, wherein: The processor is used to execute the steps of the method according to any one of claims 4 to 5 when running a computer program.

17. A loan amount assessment model training device, It is characterized in that include: A processor and a memory for storing a computer program that can be executed on the processor, wherein: The processor is used to execute the steps of the method according to any one of claims 6 to 8 when running a computer program.

18. A loan amount assessment device, It is characterized in that include: A processor and a memory for storing a computer program that can be executed on the processor, wherein: The processor is used to execute the steps of the method according to any one of claims 9 to 10 when running a computer program.

19. A credit loan system, It is characterized in that include: The credit assessment device according to claim 16 and the loan amount assessment device according to claim 18, wherein the credit assessment device and the loan amount assessment device are communicatively connected.

20. A storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.