Digitalized unhealthy asset valuation system-based divisional model generation method

Abstract

The invention relates to the technical field of divisional models, and discloses a divisional model generation method based on a digital unhealthy asset valuation system, and the method comprises the following steps: processing personal loan business data with sequential input and output; distinguishing the output data samples from real data samples, and identifying whether the generated samples are from real business data distribution; learning potential distribution of business data through a set model, and generating new business introduction data; training, learning and evaluating the reputation ability of the user on a wide personal loan business database, and performing fine adjustment in downstream tasks to meet different task requirements; and the service data training result and the comprehensive ability of the debtor are monitored in real time. According to the method, the historical repayment data of the debtor is analyzed, so that the average delay rate, the delayed repayment amount and the ratio of the total loan amount to the income of the personal loan are determined, and the repayment ability of each debtor can be accurately and efficiently evaluated.

Description

Technical Field

[0001] The present invention relates to the technical field of case division models, and more specifically to a method for generating case division models based on a digital non-performing asset valuation system. Background Art

[0002] The asset valuation model system can help investors, analysts and companies evaluate the market value of different types of assets. It provides reliable asset value assessment by comprehensively analyzing various data and factors. In particular, the case division of non-performing assets refers to the allocation of cases based on the current caseload of the collection company or collection personnel when allocating loan cases, but this lacks a comprehensive evaluation of the debtor's credit.

[0003] The existing patent discloses a loan case assignment management method (CN 112950095A), including obtaining multiple case information of loan cases to be assigned; preprocessing the multiple case information; obtaining the weight value corresponding to each case information based on the current case cycle, and extracting multiple target assignment features from the multiple case information after preprocessing according to the weight value; inputting the multiple target assignment features into the loan case assignment model, so as to output the target company matching the loan case to be assigned through the loan case assignment model. The technology disclosed in the patent cannot make a comprehensive and accurate evaluation of the actual credit status and repayment ability of each debtor, and cannot judge whether to increase the frequency of debt collection and the allocation of debt collection personnel according to the credit status and repayment ability of the debtor. Summary of the invention

[0004] The main technical problem solved by the present invention is to provide a case-by-case model generation method based on a digital non-performing asset valuation system, which solves the problems in the above-mentioned background technology.

[0005] To solve the above technical problems, according to one aspect of the present invention, more specifically, a method for generating a case-by-case model based on a digital non-performing asset valuation system comprises the following steps:

[0006] S 1. Processing personal loan business data whose input and output are both sequences;

[0007] S2. Distinguish the output data samples from the real data samples, and identify whether the generated samples come from the real business data distribution;

[0008] S3, learn the potential distribution of business data through the set model and generate new business introduction data;

[0009] S4. Train on a wide range of personal loan business databases to learn and evaluate the user's creditworthiness, and fine-tune in downstream tasks to meet different task requirements;

[0010] S5. Monitor the business data training results and the comprehensive capabilities of the debtor in real time, and dynamically adjust the case division strategy based on the progress and effect of the collection results;

[0011] S6. Match based on case characteristics and professional expertise of post-loan managers.

[0012] Furthermore, in step S1, the personal loan business data of the input sequence is converted into a vector of fixed length, and then the financial business data of the output sequence is generated from the context vector.

[0013] Furthermore, the step S2 includes varying the output business data, and generating data samples that are more consistent with the actual state based on the varied business data samples.

[0014] Furthermore, in step S2, a case classification model for non-performing assets is generated according to the differentiated data samples, and:

[0015] S201. Decompose the relevant business data in the complex sample into smaller and more manageable businesses;

[0016] S202, automatically assigning newly generated overdue business to post-loan management personnel with stronger business capabilities according to preset rules and algorithms;

[0017] S203. Analyze the historical collection data of debtors and evaluate the comprehensive capabilities of each debtor;

[0018] S204: Feedback the relevant data of the debtor's comprehensive ability to step S5.

[0019] Furthermore, the repayment capacity of each debtor is assessed based on the analysis of the debtor's historical repayment data, including the average delay rate of personal loans, the delayed repayment amount, and the total loan amount to income ratio:

[0020]

[0021] In the formula, C represents the debtor's repayment ability coefficient, a represents the average delay rate of the debtor's personal loan business, b represents the delayed repayment amount of the debtor's personal loan business, and s represents the ratio of the debtor's income to the total loan amount.

[0022] Furthermore, based on the training results on the personal loan business database, the comprehensive credit of the debtor is evaluated as follows:

[0023]

[0024] In the formula, H represents the comprehensive credit coefficient of the debtor, k represents the total overdue records and default times of the debtor, u represents the total number of contract defaults in history, and s represents the ratio of the debtor's income to the total loan amount.

[0025] Furthermore, the loan recovery strategy for the debtor can be adjusted based on the debtor's repayment ability and the debtor's overall creditworthiness:

[0026]

[0027] In the formula, m represents the probability coefficient of collecting debts from debtors and reducing the loan amount;

[0028] When m is less than 60%, it means that the frequency of debt collection from the debtor needs to be increased and the total loan amount needs to be reduced.

[0029] Furthermore, in step S4, training the personal loan business data specifically includes:

[0030] 1) Text classification training: based on the collected and discriminated variational sample data, personal loan business data and its text content are divided;

[0031] 2) Entity recognition training, used to identify and classify physical objects from text data and digital information;

[0032] 3) Machine translation training, converting digital information into user-readable text and graphic information.

[0033] The beneficial effects of the case-by-case model generation method based on the digital non-performing asset valuation system of the present invention are:

[0034] 1. The present invention analyzes the historical repayment data of debtors to determine the average delay rate of personal loans, the delayed repayment amount, and the total loan amount to income ratio, so as to accurately and efficiently evaluate the repayment ability of each debtor.

[0035] 2. The present invention evaluates the comprehensive credit of the debtor by analyzing all overdue records, number of defaults, total number of contract defaults in history, and the ratio of the debtor's income to total loan amount. The comprehensive credit can evaluate the status of non-performing assets efficiently and intuitively.

[0036] 3. The present invention determines whether it is necessary to increase the frequency of debt collection and reduce the total loan amount by analyzing the debtor and the comprehensive credit. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0038] Figure 1The flowchart of the case division model generation method in the present invention. DETAILED DESCRIPTION

[0039] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0040] like Figure 1 According to one aspect of the present invention, a method for generating a case-by-case model based on a digital non-performing asset valuation system is provided, comprising the following steps:

[0041] Step 1: Processing personal loan business data whose input and output are both sequences; converting the input sequence personal loan business data into a vector of fixed length, and then generating output sequence financial business data from the context vector.

[0042] Step 2: Distinguish the output data samples from the real data samples, and identify whether the generated samples come from the real business data distribution; this step includes varying the output business data, and generating data samples that are more in line with the real state based on the varied business data samples.

[0043] Generate a case classification model for non-performing assets based on differentiated data samples, and:

[0044] 1) Decompose the relevant business data in complex samples into smaller and more manageable businesses;

[0045] 2) Automatically assign newly generated overdue business to post-loan management personnel with stronger business capabilities based on preset rules and algorithms;

[0046] 3) Analyze the historical collection data of debtors and evaluate the comprehensive ability of each debtor; Analyze the historical collection data of debtors to determine the average delay rate of personal loans, delayed repayment amount, and total loan amount to income ratio, so as to accurately and efficiently evaluate the repayment ability of each debtor.

[0047] 4) Feed the relevant data of the debtor’s comprehensive capabilities back to step five.

[0048] Step 3: Use the set model to learn the potential distribution of business data and generate new business introduction data;

[0049] Step 4: Train on a wide range of personal loan business databases to learn and evaluate the user's creditworthiness, and fine-tune in downstream tasks to meet different task requirements. Analyze the debtor's entire overdue record, number of defaults, total number of contract defaults in history, and the ratio of the debtor's income to total loan amount to evaluate the debtor's comprehensive creditworthiness. This comprehensive creditworthiness can efficiently and intuitively evaluate the status of non-performing assets.

[0050] In addition, the training of personal loan business data specifically includes:

[0051] 1) Text classification training: based on the collected and discriminated variational sample data, personal loan business data and its text content are divided;

[0052] 2) Entity recognition training, used to identify and classify physical objects from text data and digital information;

[0053] 3) Machine translation training, converting digital information into user-readable text and graphic information.

[0054] Step 5: Monitor the business data training results and the comprehensive capabilities of the debtor in real time, and dynamically adjust the case division strategy according to the progress and effect of the collection results, analyze the debtor and the overall credit, and then determine whether it is necessary to increase the frequency of collection and reduce the total loan amount for the debtor;

[0055] Step six: Match based on case characteristics and professional expertise of post-loan managers.

[0056] In this embodiment, the repayment ability of each debtor is evaluated based on the analysis of the debtor's historical repayment data, including the average delay rate of personal loans, the delayed repayment amount, and the total loan amount to income ratio, as follows:

[0057]

[0058] In the formula, C represents the debtor's repayment ability coefficient, a represents the average delay rate of the debtor's personal loan business, b represents the delayed repayment amount of the debtor's personal loan business, and s represents the ratio of the debtor's income to the total loan amount.

[0059] Among them, the repayment ability of debtor X is determined, the average delay rate of debtor X's personal loan business is a=13%, the delayed repayment amount of debtor X's personal loan business is b=5300 (yuan), and the ratio of debtor X's income to the total loan amount is s=2.3, then:

[0060]

[0061] From the above calculation, we can know that the repayment capacity coefficient of debtor X is C = 51%. When C < 60%, it means that the debtor's repayment capacity is insufficient.

[0062] In this embodiment, the comprehensive credit of the debtor is evaluated based on the training results on the personal loan business database:

[0063]

[0064] In the formula, H represents the comprehensive credit coefficient of the debtor, k represents the total overdue records and default times of the debtor, u represents the total number of contract defaults in history, and s represents the ratio of the debtor's income to the total loan amount.

[0065] Among them, for all overdue records and default times of debtor X, k=7, the total number of contract defaults in history is u=13, and the ratio of debtor X's income to total loan amount is s=2.3, then:

[0066]

[0067] From the above calculation, we can know that the comprehensive credit coefficient of debtor X is H = 48%. Among them, the debtor's comprehensive credit coefficient H> 60% means that the credit is qualified.

[0068] In this embodiment, the loan recovery strategy for the debtor is adjusted based on the debtor's repayment ability and the debtor's comprehensive creditworthiness, as follows:

[0069]

[0070] In the formula, m represents the probability coefficient of collecting debts from debtors and reducing the loan amount;

[0071] When m is less than 60%, it means that the frequency of debt collection from the debtor needs to be increased and the total loan amount needs to be reduced.

[0072] Based on the above analysis of the repayment ability coefficient and comprehensive credit coefficient of debtor X, the probability coefficients of collecting debts from the debtor and reducing the loan amount are:

[0073]

[0074] From the above calculation, it can be known that the probability coefficient of collecting debts from debtor X and reducing the loan amount is m=40%.

[0075] That is, at this time, it is necessary to increase the frequency of collection for debtor X and reduce the total loan amount.

[0076] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention also fall within the protection scope of the present invention.

Claims

1. A method for generating case-by-case models based on a digital non-performing asset valuation system, characterized in that: The following steps are involved: S1, processing personal loan business data where both input and output are sequences; S2. Distinguish the output data samples from the real data samples, and identify whether the generated samples come from the real business data distribution; S3, learn the potential distribution of business data through the set model and generate new business introduction data; S4. Train on a wide range of personal loan business databases to learn and evaluate the user's creditworthiness, and fine-tune in downstream tasks to meet different task requirements; S5. Monitor the business data training results and the comprehensive capabilities of the debtor in real time, and dynamically adjust the case division strategy based on the progress and effect of the collection results; S6. Match based on case characteristics and professional expertise of post-loan managers.

2. The method for generating case-by-case models based on a digital non-performing asset valuation system according to claim 1 is characterized in that: In the step S1, the personal loan business data of the input sequence is converted into a vector of fixed length, and then the financial business data of the output sequence is generated from the context vector.

3. The method for generating case-by-case models based on a digital non-performing asset valuation system according to claim 1 is characterized in that: The step S2 includes performing variation on the output business data, and generating data samples that are more consistent with the actual state based on the varied business data samples.

4. The method for generating case-by-case models based on a digital non-performing asset valuation system according to claim 1 is characterized in that: In the step S2, a case classification model for non-performing assets is generated according to the differentiated data samples, and: S201. Decompose the relevant business data in the complex sample into smaller and more manageable businesses; S202, automatically assigning newly generated overdue business to post-loan management personnel with stronger business capabilities according to preset rules and algorithms; S203. Analyze the historical collection data of debtors and evaluate the comprehensive capabilities of each debtor; S204: Feedback the relevant data of the debtor's comprehensive ability to step S5.

5. The method for generating case-by-case models based on a digital non-performing asset valuation system according to claim 4 is characterized in that: Based on the analysis of the debtors’ historical repayment data, including the average delay rate of personal loans, the delayed repayment amount, and the total loan amount to income ratio, the repayment ability of each debtor is assessed: In the formula, C represents the debtor's repayment ability coefficient, a represents the average delay rate of the debtor's personal loan business, b represents the delayed repayment amount of the debtor's personal loan business, and s represents the ratio of the debtor's income to the total loan amount.

6. The method for generating case-by-case models based on a digital non-performing asset valuation system according to claim 5 is characterized in that: According to the training results on the personal loan business database, the comprehensive credit of the debtor is evaluated as follows: In the formula, H represents the comprehensive credit coefficient of the debtor, k represents the total overdue records and default times of the debtor, u represents the total number of contract defaults in history, and s represents the ratio of the debtor's income to the total loan amount.

7. The method for generating case-by-case models based on a digital non-performing asset valuation system according to claim 6 is characterized in that: In combination with the debtor's repayment ability and the debtor's overall credit, the loan recovery strategy for the debtor is adjusted as follows: In the formula, m represents the probability coefficient of collecting debts from debtors and reducing the loan amount; When m is less than 60%, it means that the frequency of debt collection from the debtor needs to be increased and the total loan amount needs to be reduced.

8. The method for generating case-by-case models based on a digital non-performing asset valuation system according to claim 1 is characterized in that: In step S4, training the personal loan business data specifically includes: 1) Text classification training: based on the collected and discriminated variational sample data, personal loan business data and its text content are divided; 2) Entity recognition training, used to identify and classify physical objects from text data and digital information; 3) Machine translation training, converting digital information into user-readable text and graphic information.

Images