Risk assessment method and device, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides a risk assessment method and device, electronic equipment, a medium and a product, and relates to the field of artificial intelligence. The method comprises the following steps: acquiring to-be-processed data; inputting the to-be-processed data into a pre-trained fuzzy logic model to obtain a risk assessment result output by the fuzzy logic model; wherein the fuzzy logic model is obtained by training based on a data sample, a simulation sample and an adversarial sample, the data sample comprises multiple groups of credit data and corresponding risk results, the simulation sample is generated based on the data sample, and the adversarial sample is generated based on the polluted data sample. According to the method, credit risk assessment can be accurately carried out by using fuzzy data.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a risk assessment method, electronic device, storage medium, and program product. Background Art

[0002] With the rapid development of big data and artificial intelligence technologies, the financial sector has undergone profound changes. Financial risk management based on AI has become a new trend. Credit risk assessment is a crucial component of financial risk management. Accurately assessing and predicting a customer's default risk can help banks effectively manage risk and reduce bad debt losses.

[0003] Existing credit risk assessment methods typically rely on scoring card models, logistic regression models, or decision tree models in machine learning. Although these models can assess borrowers' risk levels to a certain extent, they mostly rely on clear rules and clear historical data. The models have high requirements for historical data and have difficulty processing data information that is ambiguous, uncertain, or subjective, such as borrowers' behavioral patterns, the predictability of future income, and behavioral performance in social networks.

[0004] Therefore, how to use fuzzy data to accurately assess credit risk has become an urgent problem to be solved. Summary of the Invention

[0005] The present application provides a risk assessment method, device, electronic device, medium and product for accurately performing credit risk assessment using fuzzy data.

[0006] In a first aspect, the present application provides a risk assessment method, comprising:

[0007] Get the data to be processed;

[0008] Inputting the data to be processed into a pre-trained fuzzy logic model to obtain a risk assessment result output by the fuzzy logic model;

[0009] The fuzzy logic model is trained based on data samples, simulated samples and adversarial samples. The data samples include multiple sets of credit data and corresponding risk results. The simulated samples are generated based on the data samples, and the adversarial samples are generated based on the contaminated data samples.

[0010] In a second aspect, the present application provides a risk assessment device, comprising:

[0011] An acquisition module is used to obtain data to be processed;

[0012] A prediction module is used to input the data to be processed into a pre-trained fuzzy logic model to obtain a risk assessment result output by the fuzzy logic model; wherein the fuzzy logic model is trained based on data samples, simulated samples and adversarial samples, the data samples include multiple sets of credit data and corresponding risk results, the simulated samples are generated based on the data samples, and the adversarial samples are generated based on the contaminated data samples.

[0013] In a third aspect, the present application provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0014] The memory stores computer-executable instructions;

[0015] The processor executes the computer-executable instructions stored in the memory to implement the above method.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to implement the method as described above when executed by a processor.

[0017] In a fifth aspect, the present application provides a computer program product, characterized in that it includes a computer program, which implements the method described above when executed by a processor.

[0018] The risk assessment method, device, electronic device, medium and product provided in this application significantly enhance the system's ability to process fuzzy data and uncertain information by introducing a fuzzy logic model; furthermore, the fuzzy logic model is trained based on data samples, simulation samples and adversarial samples, which effectively improves the robustness of the fuzzy logic model, not only enhancing the model's ability to cope with different types of borrower groups, but also improving the model's stability and accuracy when facing abnormal data or extreme situations, thereby effectively reducing the model's assessment errors caused by changes in the external environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0020] Figure 1 A schematic diagram of a risk assessment method according to an embodiment of the present invention;

[0021] Figure 2 A schematic diagram of the structure of a risk assessment device provided in an embodiment of the present application;

[0022] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0023] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0024] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0025] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0026] In addition, this application involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.), and the use of artificial intelligence technology for automated decision-making, and a technical solution for making decisions that have a significant impact on personal rights and interests based on the results of automated decision-making. The application provides users with corresponding operation entrances for users to choose to agree or reject the results of automated decision-making; if the user chooses to reject, the expert decision-making process will be entered.

[0027] It should be noted that the risk assessment methods, devices, electronic devices, media and products provided in this application can be used in the field of artificial intelligence, and can also be used in any field other than artificial intelligence. The application fields of the risk assessment methods, devices, electronic devices, media and products in this application are not limited.

[0028] With the rapid development of big data and artificial intelligence technologies, the financial sector has undergone profound changes. Financial risk management based on AI has become a new trend. Credit risk assessment is a crucial component of financial risk management. Accurately assessing and predicting a customer's default risk can help banks effectively manage risk and reduce bad debt losses.

[0029] Existing credit risk assessment methods typically rely on scoring card models, logistic regression models, or decision tree models in machine learning. Although these models can assess borrowers' risk levels to a certain extent, they mostly rely on clear rules and clear historical data. The models have high requirements for historical data and have difficulty processing data information that is ambiguous, uncertain, or subjective, such as borrowers' behavioral patterns, the predictability of future income, and behavioral performance in social networks.

[0030] Therefore, how to use fuzzy data to accurately assess credit risk has become an urgent problem to be solved.

[0031] The risk assessment method, device, electronic device, medium and product provided in this application are intended to solve the above technical problems of the prior art.

[0032] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0033] Figure 1 This is a flow chart of the risk assessment method provided in the embodiment of the present application. Figure 1 As shown, the method includes:

[0034] S101, obtaining data to be processed;

[0035] S102. Input the data to be processed into a pre-trained fuzzy logic model to obtain the risk assessment results output by the fuzzy logic model; wherein the fuzzy logic model is trained based on data samples, simulated samples and adversarial samples, the data samples include multiple groups of credit data and corresponding risk results, the simulated samples are generated based on the data samples, and the adversarial samples are generated based on the contaminated data samples.

[0036] In a specific implementation, after obtaining the data to be processed, a fuzzy logic model can be used to process it. The fuzzy logic model can effectively analyze and process the fuzzy information in the data, incorporating difficult-to-quantify subjective factors into the assessment system and performing reasoning based on fuzzy rules. This enables the fuzzy logic model to more accurately reflect the borrower's risk characteristics and provide more reliable risk assessment results when faced with incomplete or ambiguous data. The fuzzy logic model is trained based on data samples, simulated samples, and adversarial samples. Simulated samples are used to improve the fuzzy logic model's ability to identify data samples, while adversarial samples are used to challenge the fuzzy logic model's robustness, resulting in the trained fuzzy logic model being able to accurately output risk assessment results.

[0037] The risk assessment method provided in this embodiment significantly enhances the system's ability to process fuzzy data and uncertain information by introducing a fuzzy logic model. Furthermore, the fuzzy logic model is trained based on data samples, simulated samples, and adversarial samples, effectively improving the robustness of the fuzzy logic model. This not only enhances the model's ability to cope with different types of borrower groups, but also improves the model's stability and accuracy when facing abnormal data or extreme situations, thereby effectively reducing the model's assessment errors caused by changes in the external environment.

[0038] In some embodiments, the method further comprises:

[0039] Get data samples;

[0040] Generate a simulated sample using a first generative adversarial network based on the data sample; and perform contamination on the data sample, and generate an adversarial sample using a second generative adversarial network based on the contaminated data sample;

[0041] The fuzzy logic model is trained based on the data samples, the simulation samples and the adversarial samples.

[0042] In a specific implementation, in order to expand the amount of training data, after obtaining real data samples, a generative adversarial network can be used to generate samples, and the data samples and the generated samples are used together as training data to train the fuzzy logic model. Among them, a first generative adversarial network can be used to generate simulated samples based on the data samples. The simulated samples have a high degree of similarity with the data samples. By generating simulated samples, the fuzzy logic model's ability to recognize data samples can be enhanced when the number of real data samples is limited. In addition, a second generative adversarial network can be used to generate adversarial samples based on the contaminated data samples. Adversarial samples are used to cause the fuzzy logic model to output inaccurate risk assessment results, which can identify the weaknesses of the fuzzy logic model and effectively challenge the robustness of the fuzzy logic model.

[0043] By combining the fuzzy logic model with a two-layer generative adversarial network, the adaptability and accuracy of credit risk assessment are significantly improved.

[0044] Optionally, the first generative adversarial network includes a first generator and a first discriminator; and generating a simulated sample using the first generative adversarial network according to the data sample includes:

[0045] generating a first sample using a first generator according to random noise of the data sample;

[0046] According to the data sample, using the first discriminator to determine the probability that the first sample is a data sample, and obtain an output result;

[0047] Using the loss function, the parameters of the first generator and the first discriminator are optimized until the output result of the first discriminator is close to 50%, and the current first sample is used as a simulation sample.

[0048] In the specific implementation, the first-layer generative adversarial network model GAN1 is pre-designed and initialized. The first-layer generative adversarial network model GAN1 includes a first generator G1 and a first discriminator D1. The first generator G1 is based on the preprocessed data sample set D′ s Generate a simulated first sample, the first discriminator D1 is used to distinguish the generated first sample from the actual data sample, and the initial parameter set of the first generator G1 is defined as The initial parameter set of the first discriminator D1 is defined as

[0049] Define the generating function of the first generator G1 as Among them, z is the data sample set D′ s The noise distribution p z (z) is a randomly drawn vector, the first generator G1 is based on the generating function Generate the first sample The generated set of simulated samples is defined as

[0050] The first discriminator D1 is used to determine whether the input sample x is a real data sample, where x is an actual data sample or the first sample generated by the first generator G1. Judgment results The probability that the input sample x is the actual data sample.

[0051] The first generator G1 and the first discriminator D1 are trained through the objective function to optimize the first generative adversarial network model. The objective function is as follows:

[0052]

[0053] in, Represents the data distribution p of the actual data sample data expectations on (x), Denotes the noise distribution p z (z), optimizing the first generator parameters of the first generator G1 and the first discriminator D1 through multiple iterations and the first discriminator parameters Until the first sample When the first discriminator D1 is input, the output result is close to 50%. At this time, it can be determined that the first discriminator D1 cannot distinguish the first sample Is it an actual data sample? The first sample generated by the current first generator G1 is As a simulation sample.

[0054] Through the first generative adversarial network, simulated samples similar to data samples can be accurately generated, effectively expanding the sample size of training data.

[0055] In one possible implementation, the second generative adversarial network includes a second generator and a second discriminator; and generating adversarial samples using the second generative adversarial network includes:

[0056] generating a second sample using a second generator according to the random noise of the contaminated data sample;

[0057] According to the contaminated data sample, using the second discriminator to judge the probability that the second sample is the contaminated data sample, and obtaining an output result;

[0058] Use the loss function to optimize the parameters of the second generator and the second discriminator until the output result of the second discriminator is close to 50%, and use the current second sample as an adversarial sample.

[0059] In the specific implementation, the second-layer generative adversarial network model GAN2 is pre-designed and initialized. The second-layer generative adversarial network model GAN2 includes a second generator G2 and a second discriminator D2. The second generator G2 is used to generate a second sample to challenge the robustness of the fuzzy logic model FLS, and the second discriminator D2 is used to distinguish the generated second sample from the actual contaminated data sample. The initial parameter set of the second generator G2 is defined as The initial parameter set of the second discriminator D2 is defined as

[0060] Define the generating function of the second generator G2 Where z' is the noise distribution p of the contaminated data sample z′ A randomly drawn vector in (z′), generating function Generate the second sample The second sample is used to cause the fuzzy logic model FLS to output an inaccurate risk score or risk level.

[0061] The second discriminator D2 is used to determine whether the input sample x is an actual contaminated data sample or a second sample generated by the second generator G2. The discrimination result of the second discriminator D2 The probability that the input sample x is the actual contaminated data sample.

[0062] The second generator G2 and the second discriminator D2 are trained through the objective function to optimize the second generative adversarial network model. The objective function is as follows:

[0063]

[0064] in, Denotes the noise distribution p z′ The expectation on (z′) is to optimize the first generator parameters of the second generator G2 and the second discriminator D2 through multiple iterations and the first discriminator parameters Until the second sample When the second discriminator D2 is input, the output result is close to 50%. At this time, it can be determined that the second discriminator D2 cannot distinguish the second sample Is it an actual contaminated data sample? The second sample generated by the current second generator G2 as adversarial examples.

[0065] Through the second generative adversarial network, adversarial samples similar to the contaminated data samples can be accurately generated, effectively expanding the sample size of training data and improving the robustness of the trained fuzzy logic model.

[0066] Optionally, a fuzzy logic model is trained based on the data samples, simulated samples, and adversarial samples, including:

[0067] Build an initial model;

[0068] Determine a first training set based on the data sample and the simulation sample, use the first training set to train the initial model until the accuracy of the risk assessment result corresponding to the first training set output by the initial model is greater than a first threshold, and use the current initial model as the first model;

[0069] According to the data sample, the second training set is determined, and according to the adversarial sample, the third training set is determined; the first model is trained using the second training set and the third training set until the accuracy of the risk assessment result corresponding to the second training set output by the first model is greater than the second threshold, and the accuracy of the risk assessment result corresponding to the third training set output by the first model is less than the third threshold, and the current first model is used as a fuzzy logic model.

[0070] In the specific implementation, the actual data sample set D can be used s And the set of simulated samples generated by the first generator G1 are used to construct the fuzzy logic model FLS. The fuzzy logic model FLS is used to deal with the fuzziness and uncertainty in the borrower behavior data, and represents the impact of different input variables on the output results in the fuzzy reasoning mechanism.

[0071]

[0072] Where n is the number of fuzzy rules in the fuzzy rule base, x iis the input variable corresponding to the i-th fuzzy rule in the fuzzy rule base, is x i For fuzzy set A i The membership function, w i is x i The corresponding weight factor, y i is x i The corresponding output variable; FLS(x) is the total evaluation result of the input sample x. The fuzzy logic model combines multiple input variables x of the input sample x. i The membership degree of each of the two variables is used to determine the total evaluation result of the output in a weighted average manner.

[0073] Membership function μ Ai (x i ) by input variable x i Nonlinear mapping to generate membership in high-dimensional space:

[0074]

[0075] Among them, b i is the fuzzy set A i The center point, σ i is the scale parameter that controls the expansion of the fuzzy set, γ i Control parameters for the shape.

[0076] Preliminary construction of fuzzy rule base R:

[0077]

[0078] Among them, ∧ represents the intersection operation of fuzzy logic, Represents a generalized synthesis operation of fuzzy logic, used to synthesize the effects of multiple input variables on the output. The operator is defined as:

[0079]

[0080] Set up the fuzzy inference mechanism and calculate the membership function μ of the output variable y to the fuzzy set B B (y):

[0081]

[0082] Where φ(y) is the membership distribution function of the output variable y. The fuzzy inference mechanism transfers the membership of the input variable to the output terminal through a maximum-minimum compound operation to calculate the final risk score or risk level.

[0083] Use the actual data sample set D s and simulated sample sets As the first training set, the initial model of the fuzzy logic model is carried out, and when the accuracy of the initial model on the first training set is greater than a first threshold, the current initial model is used as the first model.

[0084] Exemplarily, the first training set can be further divided into a training set and a validation set, which is not limited here.

[0085] Using adversarial example sets and the actual data sample set D s , retrain the first model, update the fuzzy rule base R and membership function μ Ai (x i ), μ Bi (y i ) parameters.

[0086]

[0087] in, is the predicted output of the adversarial sample data, ω k is the fuzzy rule weight, is the distance between the adversarial sample data and the actual dataset, δ m is the parameter adjustment factor updated during the training process, M is the number of membership functions, is the input variable of the adversarial sample data.

[0088] The first model is trained by minimizing the loss function to optimize the fuzzy rule base and membership function:

[0089]

[0090] Among them, y i is the actual risk score or risk level, FLS(x i ; θ)) is the risk score or risk level output by the model, θ is the parameter set of the first model, λ is the regularization coefficient used to control the smoothness of the membership function, and N is the number of data samples.

[0091] Through multiple iterative training, the parameter set of the first model is tuned to make the first model more effective in processing the adversarial sample set. The accuracy is low when processing the actual data sample set D s The accuracy is high, so it can remain effective in different credit risk scenarios. According to the parameter set optimized during the training process, the fuzzy rule weight ω in the fuzzy rule base R is updated. k and membership function The shape parameter σ i , center point b i The updated first model is used as the final fuzzy logic model.

[0092] In one possible implementation, the data to be processed is input into a pre-trained fuzzy logic model to obtain a risk assessment result output by the fuzzy logic model, including:

[0093] Use fuzzy logic model to calculate the risk score of the data to be processed under various fuzzy rules;

[0094] According to the risk scores of the data to be processed under various fuzzy rules, the total risk score of the data to be processed is obtained as the risk assessment result.

[0095] In the specific implementation, the fuzzy logic model can calculate the risk score of the data to be processed under various fuzzy rules based on the updated FLS(x) calculation formula, and accurately calculate the total risk score of the data to be processed according to algorithms such as the weighted average method, the maximum membership method, and the center method.

[0096] Exemplarily, based on the risk scores of the data to be processed under various fuzzy rules, the total risk score of the data to be processed is obtained, including:

[0097] Based on the weighted average method, the total risk score of the data to be processed is obtained according to the risk score of the data to be processed under each fuzzy rule and the membership function.

[0098] In the specific implementation, the membership function μ of the output variable y can be calculated through the fuzzy reasoning mechanism B′ (y), the fuzzified output is defuzzified using the weighted average method to obtain the borrower's total risk score

[0099]

[0100] Among them, y i is the risk score corresponding to the i-th fuzzy rule, μ B′ (y i ) is the value of the corresponding membership function, N is the number of fuzzy rules, and the membership function μ B′ (y i ) is determined based on the final trained fuzzy logic model.

[0101] Optionally, the method further includes:

[0102] Based on the preset correspondence between the total risk score and the risk level, the risk level of the data to be processed is determined according to the total risk score of the data to be processed as the risk assessment result.

[0103] In specific implementations, the risk levels corresponding to different total risk scores can be predefined as follows:

[0104]

[0105] in For the total risk score, The higher the value, the greater the risk. is the final risk level. Get the total risk score of the data to be processed Afterwards, you can The scoring range to which the risk belongs determines the corresponding risk level.

[0106] In one possible implementation, the data to be processed includes at least one of the borrower's income, total debt, disposable income, number of loans, repayment record, number of defaults, social network credit score, job stability, personal credit score, and length of credit history.

[0107] In some embodiments, the method further comprises:

[0108] Provide loan decision recommendations to financial institutions based on the overall risk score or risk level, including whether to grant loans, loan amounts and loan terms.

[0109] In practice, the system can provide financial institutions with decision-making recommendations on whether to grant loans based on the borrower's overall risk score or risk level. For example, if a borrower is assessed as low risk, a loan may be granted, with relatively loose loan amounts and terms. For borrowers with medium or low risk, a moderate loan may be granted, with the amount capped based on the borrower's income level and debt situation. For borrowers with medium risk, a loan may be granted based on a rigorous review, with the loan amount and terms carefully calculated.

[0110] When setting the loan amount, the borrower's risk score, income level, historical lending record, social behavior, and personal credit score can be comprehensively considered. The loan amount must be set in accordance with the borrower's actual repayment ability and adjusted according to the risk score. The loan amount for high-risk borrowers is lower than that for low-risk borrowers.

[0111] The setting of interest rates is related to the borrower's risk score. For low-risk borrowers, the interest rate is close to the benchmark interest rate. For medium-risk and high-risk borrowers, the interest rate should be increased accordingly.

[0112] Other loan conditions are also set based on the risk category of the borrower, such as loan term, repayment method and guarantee requirements. Low-risk borrowers enjoy long loan terms and repayment methods, while high-risk borrowers have their loan terms fixed at a short term and require guarantee measures.

[0113] Finally, a loan decision recommendation report can be generated, which lists the borrower's risk score, recommended loan amount, interest rate, loan term and other relevant terms, effectively improving the efficiency and reliability of loan assessment.

[0114] Figure 2 This is a schematic diagram of the structure of the risk assessment device provided in the embodiment of the present application. Figure 2 As shown, the device includes:

[0115] An acquisition module 21 is used to acquire data to be processed;

[0116] The prediction module 22 is used to input the data to be processed into a pre-trained fuzzy logic model to obtain a risk assessment result output by the fuzzy logic model; wherein the fuzzy logic model is trained based on data samples, simulated samples and adversarial samples, the data samples include multiple sets of credit data and corresponding risk results, the simulated samples are generated based on the data samples, and the adversarial samples are generated based on the contaminated data samples.

[0117] It should be noted that the risk assessment device is used to execute the risk assessment method as described above. Its specific implementation method can be found in the method embodiment provided in the embodiments of the present application, and will not be repeated here.

[0118] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 3 As shown, the electronic device includes:

[0119] The electronic device includes a processor 291 and a memory 292; a communication interface 293, and a bus 294. The processor 291, memory 292, and communication interface 293 can communicate with each other via bus 294. Communication interface 293 can be used for information transmission. The processor 291 can invoke logic instructions in memory 292 to execute the methods of the above embodiments.

[0120] In addition, the logic instructions in the memory 292 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product.

[0121] Memory 292, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of the present application. Processor 291 executes the software programs, instructions, and modules stored in memory 292 to perform functional applications and data processing, thereby implementing the methods in the above-mentioned method embodiments.

[0122] Memory 292 may include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function; the data storage area may store data generated based on the use of the terminal device. Memory 292 may also include high-speed random access memory and non-volatile memory.

[0123] An embodiment of the present application provides a non-transitory computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are executed by a processor, they are used to implement the method described in the above embodiment.

[0124] An embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, the method provided in any of the above embodiments of the present application is implemented.

[0125] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all optional embodiments, and the actions and modules involved are not necessarily required by this application.

[0126] It should be further noted that, although the various steps in the flowchart are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps may be performed in other orders. Moreover, at least a portion of the steps in the flowchart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily performed at the same time, but may be performed at different times. The execution order of these sub-steps or stages is not necessarily to be performed in sequence, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0127] It should be understood that the above-described device embodiments are merely illustrative, and the device of the present application may also be implemented in other ways. For example, the division of units / modules in the above-described embodiments is merely a logical functional division, and actual implementations may employ other division methods. For example, multiple units, modules, or components may be combined or integrated into another system, or some features may be omitted or not implemented.

[0128] In addition, unless otherwise specified, the functional units / modules in the various embodiments of the present application may be integrated into a single unit / module, each unit / module may exist physically separately, or two or more units / modules may be integrated together. The aforementioned integrated units / modules may be implemented in the form of hardware or software program modules.

[0129] If the integrated unit / module is implemented in hardware, the hardware may be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor may be any appropriate hardware processor, such as a CPU, GPU, FPGA, DSP, and ASIC. Unless otherwise specified, the storage unit may be any appropriate magnetic storage medium or magneto-optical storage medium, such as resistive random access memory (RRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (EDRAM), high-bandwidth memory (HBM), hybrid memory cube (HMC), etc.

[0130] If the integrated unit / module is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0131] In the above embodiments, the description of each embodiment has its own emphasis. For parts not described in detail in a particular embodiment, please refer to the relevant description of other embodiments. The technical features of the above embodiments can be combined in any way. To keep the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0132] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0133] It should be understood that the present application is not limited to the exact structure described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A risk assessment method, characterized in that: include: Get the data to be processed; Inputting the data to be processed into a pre-trained fuzzy logic model to obtain a risk assessment result output by the fuzzy logic model; The fuzzy logic model is trained based on data samples, simulated samples and adversarial samples. The data samples include multiple sets of credit data and corresponding risk results. The simulated samples are generated based on the data samples, and the adversarial samples are generated based on the contaminated data samples.

2. The method according to claim 1, characterized in that The method further comprises: Get data samples; Generate a simulated sample using a first generative adversarial network based on the data sample; and perform contamination on the data sample, and generate an adversarial sample using a second generative adversarial network based on the contaminated data sample; The fuzzy logic model is trained based on the data sample, the simulation sample and the adversarial sample.

3. The method according to claim 2, characterized in that The first generative adversarial network includes a first generator and a first discriminator; and generating a simulated sample using the first generative adversarial network according to the data sample includes: generating a first sample using the first generator according to the random noise of the data sample; According to the data sample, using the first discriminator to determine the probability that the first sample is a data sample, and obtain an output result; The loss function is used to optimize the parameters of the first generator and the first discriminator until the output result of the first discriminator is close to 50%, and the current first sample is used as the simulation sample.

4. The method according to claim 2, characterized in that The second generative adversarial network includes a second generator and a second discriminator; and generating adversarial samples using the second generative adversarial network includes: generating a second sample using the second generator according to the random noise of the contaminated data sample; According to the contaminated data sample, using the second discriminator to determine the probability that the second sample is the contaminated data sample, and obtain an output result; Using the loss function, the parameters of the second generator and the second discriminator are optimized until the output result of the second discriminator is close to 50%, and the current second sample is used as the adversarial sample.

5. The method according to claim 2, characterized in that The step of training a fuzzy logic model based on the data sample, the simulation sample, and the adversarial sample includes: Build an initial model; Determining a first training set based on the data sample and the simulation sample, training the initial model using the first training set until the accuracy of the risk assessment result corresponding to the first training set output by the initial model is greater than a first threshold, and using the current initial model as the first model; According to the data sample, a second training set is determined, and according to the adversarial sample, a third training set is determined; the first model is trained using the second training set and the third training set until the accuracy of the risk assessment result corresponding to the second training set output by the first model is greater than a second threshold, and the accuracy of the risk assessment result corresponding to the third training set output by the first model is less than a third threshold, and the current first model is used as the fuzzy logic model.

6. The method according to claim 1, wherein Inputting the data to be processed into a pre-trained fuzzy logic model to obtain a risk assessment result output by the fuzzy logic model includes: Using the fuzzy logic model, calculating the risk score of the data to be processed under various fuzzy rules; According to the risk scores of the data to be processed under various fuzzy rules, the total risk score of the data to be processed is obtained as the risk assessment result.

7. The method according to claim 6, characterized in that The total risk score of the data to be processed is obtained according to the risk scores of the data to be processed under various fuzzy rules, including: Based on the weighted average method, the total risk score of the data to be processed is obtained according to the risk scores of the data to be processed under various fuzzy rules and the membership function.

8. The method according to claim 6, characterized in that The method further comprises: Based on a preset correspondence between the total risk score and the risk level, the risk level of the data to be processed is determined according to the total risk score of the data to be processed as the risk assessment result.

9. The method according to any one of claims 1 to 8, characterized in that The data to be processed includes at least one of the borrower's income, total debt, disposable income, number of loans, repayment record, number of defaults, social network credit score, job stability, personal credit score and length of credit history.

10. A risk assessment device comprising: An acquisition module is used to obtain data to be processed; A prediction module is used to input the data to be processed into a pre-trained fuzzy logic model to obtain a risk assessment result output by the fuzzy logic model; The fuzzy logic model is trained based on data samples, simulated samples and adversarial samples. The data samples include multiple sets of credit data and corresponding risk results. The simulated samples are generated based on the data samples, and the adversarial samples are generated based on the contaminated data samples.

11. An electronic device, characterized in that: include: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 9 when executed by a processor.

13. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 9 when being executed by a processor.

Images