A business risk prediction method and system

Abstract

The present invention relates to the field of data processing technology, and in particular to a business risk prediction method and system. The method of the present invention comprises the following steps: obtaining credit application information of pending business users and credit data of historical users; establishing a first credit data set based on the credit data, performing data clustering and expansion on the first credit data set to obtain a second credit data set; extracting user credit features from the second credit data set, and obtaining multiple feature similarity parameters and risk similarities between historical users based on the user credit features; constructing a business risk assessment model using the feature similarity parameters and risk similarities; obtaining a business risk index of the pending business user through the business risk assessment model based on the credit application information of the pending business user and the credit data of historical users, and then completing the business risk prediction of the pending business user. The present invention can efficiently and accurately predict the risk of credit business.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a business risk prediction method and system. Background Art

[0002] Consumer loans, also known as "consumer loans", refer to loans lent to individual consumers for the purchase of durable consumer goods or payment of various expenses. With the rise of artificial intelligence and the Internet, Internet finance, as an emerging industry, is in a stage of rapid development in China. Different from traditional financial institutions, especially in terms of business, the loan threshold of Internet finance may be lower than that of traditional financial institutions, so the risk of overdue payments by users will also be higher than that of traditional financial institutions. At the same time, the dimensions that Internet financial institutions can use to comprehensively assess user risks are, to a certain extent, less than those of traditional institutions, which means that Internet finance is destined to face more new challenges in all aspects.

[0003] With the rapid development of Internet finance in recent years, new social consumption concepts have led to a continuous increase in the number of people choosing Internet lending and other services. However, due to information asymmetry, the information of borrowers collected by Internet financial platforms is incomplete. Due to the needs of laws and regulations and data privacy protection, some customers' personal financial information and past credit record information cannot be fully disclosed. Most of the information about customers on Internet financial platforms comes from credit reports and platform data. Therefore, in the face of a variety of loan applicants, in the absence of complete credit records, how to efficiently and accurately predict credit business risks, retain high-quality customers to the greatest extent, avoid non-performing loans and thus improve the quality of credit issuance is an issue that needs to be urgently addressed. Summary of the invention

[0004] In view of the shortcomings of existing methods and the needs of practical applications, in order to retain high-quality customers to the greatest extent, avoid bad loans and thus improve the quality of credit issuance, the problem of efficiently and accurately predicting credit business risks is solved. On the one hand, the present invention provides a business risk prediction method, comprising the following steps: obtaining the credit application information of the user of pending business and the credit data of historical users, the credit data including the credit application information and repayment records; establishing a first credit data set based on the credit data, performing data clustering and expansion on the first credit data set to obtain a second credit data set; extracting user credit features from the second credit data set, and based on the user credit features, obtaining a variety of feature similarity parameters and risk similarities between historical users; using the feature similarity parameters and the risk similarity to construct a business risk assessment model; based on the credit application information of the user of pending business and the credit data of historical users, obtaining the business risk index of the user of pending business through the business risk assessment model, and then completing the business risk prediction of the user of pending business. The present invention reasonably expands the data set for those without complete credit records, and then obtains the business risk relationship of historical users from the expanded data set, and predicts the business risk of users with pending business based on the business risk relationship, thereby reasonably restricting customer credit, reducing bad debts, and reducing corporate losses. It solves the problem of efficiently and accurately predicting credit business risks, retains high-quality customers to the greatest extent, avoids bad loans, and thus improves the quality of credit issuance, which is further conducive to the sustainable development of the Internet consumer credit industry.

[0005] Optionally, performing data clustering and expansion on the first credit data set to obtain a second credit data set comprises the following steps:

[0006] Based on the first credit data set, an initial membership is set; according to the initial membership, an initial cluster center is set; through the initial cluster center, a cluster loss function value is obtained; according to the cluster loss function value, the membership is adjusted to obtain the best cluster center, and then the first credit data set is fuzzy classified; according to the fuzzy classification result, the second credit data set is generated. The present invention effectively increases the number of samples by fuzzy classifying the first credit data set and then generating a second credit data set with more data samples according to the classification result, thereby further improving the prediction accuracy of the present invention.

[0007] Optionally, the initial cluster center is set according to the initial membership degree, satisfying the following formula:

[0008] ,in, represents the probability of taking sample C as the initial cluster center, represents the number of samples, Indicates samples, Indicates samples, Indicates samples and similar The area composed of samples is represents the Euclidean distance, Indicates The samples belong to The membership degree of the cluster centers is represents the fuzzy index, specifically The constant, represents the cluster center, represents a value function that satisfies The present invention calculates the probability of a sample becoming a cluster point center, and then determines the initial cluster center, which is conducive to obtaining the final clustering result and further improves the prediction efficiency of the present invention.

[0009] Optionally, extracting user credit features from the second credit data set comprises the following steps:

[0010] Extracting index parameters from the second credit data set; constructing an index value model, and using the index value model to obtain the index value of the index parameter; screening valuable index parameters according to the index value, evaluating the valuable index parameters, and then extracting the user credit characteristics. The present invention screens valuable index parameters by constructing an index value model, thereby reducing the computational complexity, improving the utilization efficiency of the index parameters, and further ensuring the effectiveness of the extracted user credit characteristics.

[0011] Optionally, the constructed indicator value model satisfies the following formula:

[0012] ,in, Indicates The indicator value of the indicator parameter, Indicates the number of groups. Indicates Parameters of the indicators Group The number of users with different risk levels, Indicates The weight coefficient of the risk level, Indicates Parameters of the indicators The number of risky users in the group, Indicates The total number of users who do not have risk for this indicator parameter, Indicates The total number of users with risky indicators. Indicates The number of users in the group who are not at risk, Indicates The number of risky users in the group, Indicates The number of risky users in the group The proportion of the total number of risky users under the indicator parameters, Indicates The number of users who are not at risk in the group is The present invention determines the index value by means of a model, reduces the human interference factors in the credit risk prediction, and improves the objectivity of the present invention.

[0013] Optionally, obtaining multiple feature similarity parameters and risk similarity between historical users based on the user credit feature comprises the following steps:

[0014] The user credit characteristics are encoded to obtain a credit characteristic matrix; and multiple characteristic similarity parameters and risk similarity between historical users are obtained using multiple similarity calculation models and the credit characteristic matrix.

[0015] Optionally, the similarity calculation model includes a Euclidean distance calculation model, a Manhattan distance calculation model, a Minkowski distance calculation model, a Mahalanobis distance calculation model and a cosine similarity calculation model. The present invention evaluates the similarity between historical users through a variety of similarity calculation models, improves the comprehensiveness and accuracy of the evaluation results, and further helps to improve the prediction accuracy of the present invention.

[0016] Optionally, the constructing a business risk assessment model by using the feature similarity parameter and the risk similarity comprises the following steps:

[0017] Each similarity calculation model is sorted according to the feature similarity parameter, and the indicator weight of the similarity calculation model is obtained according to the sorting result and the evaluation result of the valuable indicator parameter; based on the indicator weight and the risk similarity, the business risk assessment model is constructed. The present invention is conducive to quantitatively evaluating the business relevance between users by obtaining the indicator weight of the similarity calculation model, and further improves the prediction accuracy of the present invention.

[0018] Optionally, the index weight of the similarity calculation model is obtained based on the ranking result and the valuable index parameter evaluation result, satisfying the following formula:

[0019] ,

[0020] in, Indicates The indicator weights of the similarity calculation model, represents the number of similarity calculation models, Indicates the number of valuable indicator parameters, Indicates The standard deviation of the weight distribution of valuable indicator parameters in the similarity calculation model, Indicates Similarity calculation model and Similarity calculation model about Similarity parameters of valuable indicator parameters, Indicates The similarity calculation model is assigned to the The weight coefficients of the valuable indicator parameters, Indicates that the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The weight coefficients of the valuable indicator parameters, Indicates Similarity calculation model about The importance scale value of the valuable indicator parameter, Indicates Similarity calculation model about The present invention obtains the index weight of the similarity calculation model through a mathematical formula, thereby avoiding the situation that the index parameter weight allocation scheme of an individual similarity calculation model is seriously different from that of other similarity calculation models and completely deviates from the actual situation.

[0021] In the second aspect, in order to efficiently execute a business risk prediction method provided by the present invention, the present invention also provides a business risk prediction system, including a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute a business risk prediction method as described in the first aspect of the present invention. A business risk prediction system of the present invention has a compact structure and stable performance, and can stably execute a business risk prediction method provided by the present invention, further improving the overall applicability and practical application capabilities of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0023] Figure 1 A flow chart of a business risk prediction method provided by an embodiment of the present invention;

[0024] Figure 2 A business risk prediction system framework diagram provided by an embodiment of the present invention;

[0025] Figure 3 A schematic diagram of the structure of a business risk prediction device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0027] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.

[0028] See also Figure 1 In order to solve the problem of efficiently and accurately predicting credit business risks, retain high-quality customers to the greatest extent, avoid bad loans and thus improve the quality of credit issuance, the present invention provides a business risk prediction method, such as Figure 1 As shown, the method comprises the following steps:

[0029] S1. Obtain credit application information of pending business users and credit data of historical users.

[0030] In an embodiment, the credit application information of the user of pending business and the credit data of other historical users are obtained from the corresponding Internet financial platform database, and the credit data includes the credit application information and repayment records. In other embodiments, in order to further improve the credit data of historical users, the corresponding credit data can also be obtained from a public database.

[0031] Furthermore, the credit application information includes information filled in by the user, smart device usage log data and historical credit information. The information filled in by the user includes basic user information, including the user's name, telephone number, address, age, occupation, whether WeChat is bound, whether QQ is bound, income, gender, education and birthday information; the smart device usage log data includes online shopping information such as the number of user orders, the maximum amount of user orders, the minimum amount of user orders, the user's payment method and the user's order time, and also includes smart device positioning information; historical credit information includes credit time, credit amount, repayment method, the number and amount of loans shown in the credit report within the past 30 days, and the number of lending institutions shown in the credit report within the past 30 days. In the embodiment, the credit application information can also include more credit-related or unrelated variables, which are selected according to actual conditions.

[0032] S2. Establish a first credit data set based on the credit data, perform data clustering and expansion on the first credit data set, and obtain a second credit data set.

[0033] Specifically, user risk level assessment rules may be pre-set, and the credit risk level of the historical user may be assessed based on the historical user's repayment record and the user risk level assessment rules, and then the risk level label of the historical user may be determined to establish the first credit data set.

[0034] Further, performing data clustering and expansion on the first credit data set to obtain a second credit data set includes the following steps:

[0035] S21. Based on the first credit data set, set an initial membership degree.

[0036] Randomly initialize the membership matrix , Representation sample Belong to the sample is the membership degree of the cluster center, and , , Indicates the number of cluster centers.

[0037] S22. Setting an initial cluster center according to the initialized membership degree.

[0038] Specifically, according to the initial membership, the initial cluster center is set to satisfy the following formula:

[0039] ,

[0040] in, represents the probability of taking sample C as the initial cluster center, represents the number of samples, Indicates samples, Indicates samples, Indicates Samples and similar The area composed of samples, represents the Euclidean distance, Indicates The samples belong to The membership degree of the cluster centers is represents the fuzzy index, specifically In this embodiment, , represents the cluster center, represents a value function that satisfies .

[0041] Sample C according to To sort, The larger the value, the greater the probability that sample C is the initial cluster center. The corresponding samples are used as the initial cluster centers.

[0042] S23. Obtain a clustering loss function value through the initial cluster center.

[0043] According to the initial cluster center, the clustering loss function value is calculated to satisfy the following formula:

[0044] ,in, Indicates The clustering loss function value of the cluster center, represents the clustering hesitation, which can be adjusted according to the clustering effect and actual requirements. In this embodiment, .

[0045] S24. Adjust the degree of membership according to the clustering loss function value to obtain the best cluster center, and then perform fuzzy classification on the first credit data set.

[0046] In an embodiment, if the clustering loss function value is less than a preset value or reaches a convergence condition, the iteration is stopped, and the current clustering result is used as the final clustering result to obtain the best cluster center, and then the first credit data set is fuzzy classified.

[0047] If the clustering loss function value is greater than the preset value and the convergence condition is not reached, the initialized membership matrix is ​​updated, and the update formula satisfies:

[0048] ,in, represents the fuzzy index, Indicates The samples belong to The membership degree of the cluster centers is represents the Euclidean distance.

[0049] S25. Generate the second credit data set according to the fuzzy classification result.

[0050] Specifically, in Python, based on the obtained fuzzy classification results, categories and membership functions are defined, and sample features are generated. For each sample, its membership to each category is calculated according to the defined membership function, and the features and corresponding membership of each sample are combined into a data set to obtain a second credit data set.

[0051] S3. Extracting user credit features from the second credit data set, and obtaining multiple feature similarity parameters and risk similarities between historical users based on the user credit features.

[0052] Specifically, extracting user credit features from the second credit data set includes the following steps:

[0053] S31. Extracting indicator parameters from the second credit data set.

[0054] The indicator parameters are extracted using a text feature extraction model or a lightweight neural network model. The indicator parameters are variable parameters in the second credit data set, such as whether there is an overdue record, whether the person is married, whether the person has a vehicle, whether the person has a mortgage, whether the person has a car loan, annual income, credit card limit, limit utilization rate, income limit ratio, number of overdue payments within 6 months, number of overdue payments within 24 months, number of credit inquiries within 1 month, average number of transactions per month, and average monthly transaction amount.

[0055] S32: construct an indicator value model, and use the indicator value model to obtain the indicator value of the indicator parameter.

[0056] In the embodiment, the constructed indicator value model satisfies the following formula:

[0057] ,

[0058] in, Indicates The indicator value of the indicator parameter, Indicates the number of groups. Indicates Parameters of the indicators Group The number of users with different risk levels, Indicates The weight coefficient of the risk level, Indicates Parameters of the indicators The number of risky users in the group, Indicates The total number of users who do not have risk for this indicator parameter, Indicates The total number of users with risky indicators. Indicates The number of users in the group who are not at risk, Indicates The number of risky users in the group, Indicates The number of risky users in the group The proportion of the total number of risky users under the indicator parameters, Indicates The number of users who are not at risk in the group is It should be understood that the indicator parameters can be grouped randomly, preferably in combination with the Best-KS binning technology.

[0059] S33. According to the indicator value, valuable indicator parameters are screened, the valuable indicator parameters are evaluated, and then the user credit characteristics are extracted.

[0060] Specifically, the indicator parameters whose indicator values ​​are less than the threshold are screened out, and the remaining ones are used as valuable indicator parameters; secondly, a prediction model is established based on the XGboost algorithm; then the valuable indicator parameters are sorted from high to low according to the indicator value, and then the data set is randomly set as a training set and a validation set in a ratio of 8:3; then, the valuable indicator parameters are added as feature variables one by one in the order of the indicator parameters, and a round of training is performed to generate an AUC value as the evaluation result of the valuable indicator parameters. According to the evaluation result, the valuable indicator parameters with low value can be further screened out to obtain more representative valuable indicator parameters.

[0061] Further, based on the user credit characteristics, a plurality of characteristic similarity parameters and risk similarity between historical users are obtained, including the following steps:

[0062] S34. Encode the user's credit characteristics to obtain a credit characteristics matrix.

[0063] The valuable indicator parameters in the user's credit characteristics are encoded using the One-hot encoding or bag-of-words model method, and then a credit feature matrix is ​​generated based on the encoding results.

[0064] S35. Utilize a plurality of similarity calculation models and the credit feature matrix to obtain a plurality of feature similarity parameters and risk similarity between historical users.

[0065] In the embodiment, the similarity calculation model includes a Euclidean distance calculation model, a Manhattan distance calculation model, a Minkowski distance calculation model, a Mahalanobis distance calculation model and a cosine similarity calculation model. In some other embodiments, the similarity calculation model may also use other models.

[0066] S4. Construct a business risk assessment model using the feature similarity parameters and the risk similarity.

[0067] In an embodiment, constructing a business risk assessment model using the feature similarity parameter and the risk similarity includes the following steps:

[0068] S41. Sort each similarity calculation model according to the feature similarity parameter, and obtain the indicator weight of the similarity calculation model based on the sorting result and the valuable indicator parameter evaluation result.

[0069] For each similarity calculation model, obtain the similarity of each valuable indicator parameter and sort them from high to low according to similarity;

[0070] Furthermore, according to the ranking results and the evaluation results of the valuable indicator parameters, the indicator weights of the similarity calculation model are obtained and satisfy the following formula:

[0071] ,

[0072] in, Indicates The indicator weights of the similarity calculation model, represents the number of similarity calculation models, Indicates the number of valuable indicator parameters, Indicates The standard deviation of the weight distribution of valuable indicator parameters in the similarity calculation model, Indicates Similarity calculation model and Similarity calculation model about Similarity parameters of valuable indicator parameters, Indicates The similarity calculation model is assigned to the The weight coefficients of the valuable indicator parameters, Indicates that the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The weight coefficients of the valuable indicator parameters, Indicates Similarity calculation model about The importance scale value of the valuable indicator parameter, Indicates Similarity calculation model about The valuable indicator parameter evaluation result coefficient of the valuable indicator parameter.

[0073] S42: construct the business risk assessment model based on the indicator weights and the risk similarity.

[0074] Specifically, the constructed business risk assessment model satisfies the following formula:

[0075] ,in, represents the business risk assessment value, represents the number of similarity calculation models, Indicates The indicator weights of the similarity calculation model, Indicates The risk similarity of the similarity calculation model is Indicates The risk coefficient of the sample with the highest similarity among the similarity calculation models.

[0076] S5. Based on the credit application information of the pending business user and the credit data of historical users, the business risk index of the pending business user is obtained through the business risk assessment model, thereby completing the business risk prediction of the pending business user.

[0077] In an embodiment, the business risk assessment model is used to traverse the credit data of historical users based on the credit application information of the users of pending business, and the business risk of the users of pending business is predicted based on the risk coefficient of the most similar users.

[0078] See also Figure 2 In an embodiment, in order to efficiently execute a business risk prediction method provided by the present invention, the present invention also provides a business risk prediction system, including: an input device, an output device, a processor, and a memory, wherein the input device, the output device, the processor, and the memory are interconnected, and the memory contains program instructions, and the program instructions are used for the steps of the business risk prediction method. The business risk prediction system of the present invention has a compact structure and stable performance, and can stably execute a business risk prediction method of the present invention, further improving the overall applicability and practical application ability of the present invention.

[0079] In an embodiment, the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. An input device may be used to obtain data information. An output device may be used to output the results obtained by storing program instructions contained in a computer program in a memory provided by the present invention. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory.

[0080] In yet another alternative embodiment, see Figure 3 In order to efficiently execute a business risk prediction method provided by the present invention, this embodiment also provides a business risk prediction device, such as Figure 3 As shown, including:

[0081] The memory 10 is used to store computer programs; the processor 20 is used to execute the computer programs to implement the above-mentioned business risk prediction method. The memory 10, the processor 20, the communication interface 31 and the communication bus 32. The memory 10, the processor 20, and the communication interface 31 all communicate with each other through the communication bus 32.

[0082] In an embodiment, the memory 10 is used to store one or more program instructions. The memory 10 may store program instructions for implementing the following functions:

[0083] Acquire the credit application information of the user of pending business and the credit data of historical users, wherein the credit data includes the credit application information and repayment records; establish a first credit data set based on the credit data, expand the first credit data set to obtain a second credit data set; extract user credit features from the second credit data set, and based on the user credit features, obtain a variety of feature similarity parameters and risk similarities between historical users; construct a business risk assessment model using the feature similarity parameters and the risk similarities; based on the credit application information of the user of pending business and the credit data of historical users, obtain the business risk index of the user of pending business through the business risk assessment model, thereby completing the business risk prediction of the user of pending business.

[0084] In one possible implementation, the memory 10 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, and at least one application required for a function, etc.; the data storage area may store data created during use. In addition, the memory 10 may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include NVRAM. The memory stores an operating system and operating instructions, executable modules or data structures, or a subset thereof, or an extended set thereof, wherein the operating instructions may include various operating instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and processing hardware-based tasks.

[0085] The processor 20 may be a central processing unit (CPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array or other programmable logic device, and the processor 20 may be a microprocessor or any conventional processor, etc. The processor 20 may call a program stored in the memory 10. The communication interface 31 may be an interface of a communication module, and is used to connect with other devices or systems.

[0086] Of course, it should be noted that Figure 3 The structure shown does not constitute a limitation on the business risk prediction device in this embodiment. In actual applications, the business risk prediction device may include Figure 3 More or fewer components than shown, or combinations of certain components.

[0087] An embodiment also provides a storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned business risk prediction method are implemented.

[0088] The storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0089] In summary, the present invention reasonably expands the data set without complete credit records, and then obtains the business risk relationship of historical users from the expanded data set, predicts the business risk of pending business users based on this business risk relationship, reasonably limits customer credit, reduces bad debts, reduces corporate losses, solves the problem of efficient and accurate prediction of credit business risks, retains high-quality customers to the greatest extent, avoids bad loans, thereby improving the quality of credit issuance, and further facilitates the sustainable development of the Internet consumer credit industry. Therefore, the present invention effectively overcomes the various shortcomings of the prior art and has a high industrial utilization value.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope recorded in the present invention.

Claims

1. A business risk prediction method, characterized in that: The business risk prediction method comprises the following steps: Obtaining credit application information of pending business users and credit data of historical users, wherein the credit data includes the credit application information and repayment records; Establishing a first credit data set according to the credit data, performing data clustering and expansion on the first credit data set to obtain a second credit data set; extracting user credit features from the second credit data set, and obtaining multiple feature similarity parameters and risk similarities between historical users based on the user credit features; Using the feature similarity parameters and the risk similarity, construct a business risk assessment model; Based on the credit application information of the pending business user and the credit data of the historical users, the business risk index of the pending business user is obtained through the business risk assessment model, thereby completing the business risk prediction of the pending business user; The step of performing data clustering and expansion on the first credit data set to obtain a second credit data set comprises the following steps: Based on the first credit data set, setting an initial membership degree; According to the initial membership, setting the initial cluster center; Obtaining a clustering loss function value through the initial cluster center; adjusting the membership degree according to the clustering loss function value to obtain the best cluster center, and then performing fuzzy classification on the first credit data set; According to the fuzzy classification result, generating the second credit data set; The method of obtaining multiple feature similarity parameters and risk similarity between historical users based on the user credit features includes the following steps: Encoding the user's credit characteristics to obtain a credit characteristics matrix; Using multiple similarity calculation models and the credit feature matrix, multiple feature similarity parameters and risk similarity between historical users are obtained; The similarity calculation model includes a Euclidean distance calculation model, a Manhattan distance calculation model, a Minkowski distance calculation model, a Mahalanobis distance calculation model and a cosine similarity calculation model; The method of constructing a business risk assessment model by using the feature similarity parameter and the risk similarity comprises the following steps: Sorting each similarity calculation model according to the feature similarity parameter, and obtaining the indicator weight of the similarity calculation model according to the sorting result and the evaluation result of the valuable indicator parameter; Based on the indicator weights and the risk similarity, construct the business risk assessment model; According to the ranking results and the evaluation results of the valuable indicator parameters, the indicator weights of the similarity calculation model are obtained, which satisfy the following formula: ,in, Indicates The indicator weights of the similarity calculation model, represents the number of similarity calculation models, Indicates the number of valuable indicator parameters, Indicates The standard deviation of the weight distribution of valuable indicator parameters in the similarity calculation model, Indicates Similarity calculation model and Similarity calculation model about Similarity parameters of valuable indicator parameters, Indicates The similarity calculation model is assigned to the The weight coefficients of the valuable indicator parameters, Indicates that the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The average value of the weight coefficients of the valuable indicator parameters, Indicates The similarity calculation model is assigned to the The weight coefficients of the valuable indicator parameters, Indicates Similarity calculation model about The importance scale value of the valuable indicator parameter, Indicates Similarity calculation model about The valuable indicator parameter evaluation result coefficient of the valuable indicator parameter; The constructed business risk assessment model satisfies the following formula: ,in, represents the business risk assessment value, represents the number of similarity calculation models, Indicates The indicator weights of the similarity calculation model, Indicates The risk similarity of the similarity calculation model is Indicates The risk coefficient of the sample with the highest similarity among the similarity calculation models; The step of extracting user credit features from the second credit data set comprises the following steps: extracting an indicator parameter from the second credit data set, wherein the indicator parameter is a variable parameter in the second credit data set; Constructing an indicator value model, and using the indicator value model to obtain the indicator value of the indicator parameter; According to the indicator value, valuable indicator parameters are screened, the valuable indicator parameters are evaluated, and then the user credit characteristics are extracted; The constructed indicator value model satisfies the following formula: ， in, Indicates The indicator value of the indicator parameter, Indicates the number of groups. Indicates Parameters of the indicators Group The number of users with different risk levels, Indicates The weight coefficient of the risk level, Indicates Parameters of the indicators The number of risky users in the group, Indicates The total number of users who do not have risk for this indicator parameter, Indicates The total number of users with risky indicators. Indicates The number of users in the group who are not at risk, Indicates The number of risky users in the group, Indicates The number of risky users in the group The proportion of the total number of risky users under the indicator parameters, Indicates The number of users who are not at risk in the group is The proportion of the total number of risky users under the indicator parameters; Evaluating the valuable indicator parameters and then extracting the user credit characteristics includes the following steps: Establish a prediction model based on the XGboost algorithm; Sort the valuable indicator parameters from high to low according to the indicator value; Add the valuable indicator parameters as feature variables one by one in the order of the indicator parameters, perform a round of training, and generate the AUC value as the evaluation result of the valuable indicator parameters; According to the evaluation results, valuable indicator parameters with low values ​​are screened out to obtain more representative valuable indicator parameters, and the credit characteristics corresponding to the more representative valuable indicator parameters are used as the user credit characteristics.

2. A business risk prediction method according to claim 1, characterized in that: The initial cluster center is set according to the initial membership degree, satisfying the following formula: ， in, represents the probability of taking sample C as the initial cluster center, represents the number of samples, Indicates samples, Indicates samples, Indicates samples and similar The area composed of samples, represents the Euclidean distance, Indicates The samples belong to The membership degree of the cluster centers is represents the fuzzy index, specifically The constant, represents the cluster center, represents a value function that satisfies .

3. A business risk prediction system, characterized in that: The business risk prediction system includes: an input device, an output device, a processor, and a memory. The input device, output device, processor, and memory are interconnected. The memory includes program instructions, and the program instructions are used to execute the business risk prediction method described in any one of claims 1-2.

Images