Method for calculating credit score of loan person according to loan person information

Abstract

The invention discloses a method for calculating a credit score of a loan person according to information of the loan person. The method comprises the following steps: processing credit information classification of the loan person; quantifying credit scoring indexes; calculating credit classification and credit scoring index weights by an analytic hierarchy process; and carrying out weighted calculation on a loan classification credit score and a credit total score, and generating a radar user portrait. According to the score of each credit category, the category credit score of the user is made into a radar map, the credit portrait of the loan person is formed, the credit conditions of the loan person in all aspects are visually displayed, and the loan person is assisted in classification and loan approval decision making.

Description

Technical Field

[0001] The present invention belongs to the technical field of credit scoring, and in particular is a method for calculating a lender's credit score based on lender information. Background Art

[0002] Due to information asymmetry, some information of lenders is difficult to quantify and evaluate, and other reasons, bank loan overdue, fraud and other cases occur frequently. The traditional loan review process mainly relies on platform data statistical analysis, and more on expert experience to manually judge the risk level of the lender to determine the loan approval result. This leads to low accuracy and poor timeliness of loan approval. In addition, as the number of lenders continues to increase, the amount of data that needs to be processed is getting larger and larger. A method that can classify and quantify loan information is needed to help loan reviewers gain an intuitive and three-dimensional insight into the credit status of lenders.

[0003] Currently, the commonly used loan review methods on the market include manual review combined with rule engines, credit assessment based on lender information, etc. There are relatively few studies on the application of algorithms for risk analysis, data classification, quantitative calculation, and intuitive display.

[0004] The common loan review methods have the following shortcomings:

[0005] 1. Some credit scoring indicators for lenders are enumerated types with limited options. Such indicators can only be used as a reference during manual review and are difficult to quantify.

[0006] 2. During the loan review process, there is a lack of classification and grading of the borrower's information, and the borrower's credit assessment indicators are not differentiated in terms of their relevance to the credit score, i.e., the calculation of weights.

[0007] 3. In the process of credit assessment of lenders, there is a lack of classification and separate quantitative scoring according to the type of credit information, and there is a lack of cases that separately quantify and intuitively display all aspects of the lender's credit information. Summary of the invention

[0008] In view of the above problems, the present invention is proposed to provide a method for calculating a lender's credit score based on lender information, which overcomes the above problems or at least partially solves the above problems.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] A method for calculating a lender's credit score based on lender information, the method comprising:

[0011] Processing credit classification of lenders;

[0012] Quantification of credit scoring indicators;

[0013] Calculate the credit classification and credit scoring indicator weights using the analytic hierarchy process;

[0014] Calculate the lender's classified credit score and total credit score by weight, and generate a radar user profile.

[0015] Optionally, the credit score indicator quantification includes:

[0016] Quantify the enumeration type credit score indicators. Select unique hot encoding to quantify the enumeration type credit score indicators. Through quantization, the enumeration type credit score indicators are converted into numerical types.

[0017] The credit scoring indicators are standardized, the original data is scaled, and data normalization is performed to map the data to the [0, 1] space.

[0018] Optionally, the AHP method for calculating the credit classification and credit scoring indicator weights includes:

[0019] Modeling: Establish a hierarchical model of credit scoring indicators based on the analysis purpose, determine the credit scoring indicators that affect the evaluation objectives based on the analysis statistics, and establish a hierarchical model consisting of objectives and credit scoring indicators;

[0020] Establish a credit scoring indicator matrix, and establish a weight matrix by comparing the importance of the upper and lower credit scoring indicators;

[0021] Calculate the weight value and obtain the data sequence value of the lower credit score indicator to the upper credit score indicator through the square root method;

[0022] Check consistency. The consistency value calculation formula is:

[0023] CI=λ max -n / n-1

[0024] Where CI is the consistency value; λ max is the largest characteristic root; n is the order of A (characteristic matrix), when A is completely consistent, CI = 0; λ max -The larger the n, the worse the consistency;

[0025] The calculation formula to determine whether the consistency is satisfactory is:

[0026] CR=CI / RI

[0027] Where CI is the consistency value; RI is the index value of average random consistency, which is a set of constants (first order: 0.00, second order: 0.00, third order: 0.52, fourth order: 0.89...); CR<0.1 means that the consistency meets the requirements; if the consistency does not meet the requirements, it is necessary to adjust the weight matrix and repeat the calculation until the consistency meets the requirements.

[0028] Optionally, establish a credit scoring indicator matrix including:

[0029] A weight matrix is ​​established for the indicator category layer, and the weight value of each indicator category is calculated;

[0030] A weight matrix is ​​established for each indicator category to obtain the weight of each indicator.

[0031] Optionally, a weighted calculation is performed on the lender's classified credit score and total credit score, and a radar user profile is generated, including:

[0032] First, obtain the risk assessment value of each indicator class;

[0033] Then, according to the weight of each indicator class and the risk assessment value of the indicator class, the user's comprehensive risk assessment value is calculated through a weighted algorithm;

[0034] Finally, the risk assessment value L for each classification is obtained 1， L 2... L n And the comprehensive risk assessment value L generates the credit score profile of the radar user.

[0035] Optionally, the risk assessment value of each indicator class is obtained by a weighted algorithm according to the following calculation formula:

[0036] L i =(l i1 *W i1 +l i2 *W i2 +...+l im *W im )*100

[0037] Where: L i represents the evaluation value of the ith indicator class; if there are m lender credit indicators in the ith indicator class, l im Indicates the value of the mth credit indicator in this indicator class, W im Indicates the value weight of the mth credit indicator in this indicator class.

[0038] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0039] 1. The present invention makes a radar chart of the user's category credit score based on the score of each credit category, forms a credit profile of the lender, intuitively displays the lender's credit status in all aspects, and assists in lender classification and loan approval decision-making.

[0040] 2. Through the present invention, when launching different types of loan products, the credit indicators of different categories of users are paid different attention, taking mortgage loans and credit loans as examples. When the loan product is a mortgage loan, in the process of evaluating the credit score of the lender, we can increase the proportion of the lender's personal deposit asset indicators to make the lender's personal deposit asset indicators occupy a more significant decisive role in the credit scoring process, so as to screen out users who are more suitable for such products after the score is sorted. Similarly, in the evaluation process of credit loan products, we will make personal credit and basic information indicators occupy a heavier proportion to screen out users suitable for such products. The calculation complexity of the weight matrix is ​​greatly reduced by the analytic hierarchy process to calculate the indicator weight by classifying the indicators. For example, there are 30 indicators, which are divided into 4 indicator classes after the indicator classification, and the number of indicators in each indicator class is 7, 9, 6, and 8 respectively. Then the calculation complexity is changed from the original 30-order operation to the highest 9-order operation, which greatly improves the efficiency of the calculation and reduces the calculation cost and time. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of a method for calculating a lender's credit score based on lender information provided in an embodiment of the present application;

[0042] Figure 2 This is a hierarchical model diagram of an embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] See also Figure 1 This embodiment provides a method for calculating a lender's credit score based on lender information, the method comprising the following steps:

[0045] S1. Process the credit information classification of lenders.

[0046] S2. Quantification of credit scoring indicators.

[0047] Credit scoring indicators include:

[0048] Enumeration type credit score indicators are quantified. Unique hot encoding is selected to quantify the enumeration type credit score indicators. Through quantization, the enumeration type credit score indicators are converted into numerical types.

[0049] One-hot encoding, also known as one-bit effective encoding, uses an N-bit state register to encode N states, and at any time, only one of them is valid. Such credit scoring indicators include: education, unit nature, age group, etc. The quantization results are shown in Table 1 Educational coding table and Table 2 Unit nature coding table. Through quantization, these credit scoring indicators are converted into numerical types.

[0050] Table 1 Educational qualification coding table

[0051]

[0052] In dataset 1, the education code 0001 represents high school or below, in dataset 2, the education code 0010 represents undergraduate, and in dataset 3, the education code 0100 represents graduate.

[0053] Table 2 Unit nature coding table

[0054]

[0055] The credit scoring indicators are standardized, the original data is scaled, and data normalization is performed to map the data to the [0, 1] space.

[0056] When processing data, there will also be operations to standardize the data, that is, to scale the original data. The units and value ranges of different attribute values ​​are different (for example, a deposit amount of 100,000 yuan and 3 overdue times). Attributes with larger value ranges may affect the calculation results during calculations, weakening the influence of other factors and causing larger deviations in the calculation results. Therefore, the standardization step is indispensable. Using data normalization to map the data to the [0, 1] space can solve this problem well. Note that when the indicator value is used for calculations, if the indicator value is negatively correlated with the evaluation target, such as the overdue number indicator, the indicator value should be normalized with a negative sign. Calculation formula, where lmin represents the minimum value of the indicator and lmax represents the maximum value of the indicator:

[0057]

[0058] S3. Calculate the credit classification and credit scoring index weights using the analytic hierarchy process.

[0059] The AHP method for calculating the credit classification and credit scoring index weights includes:

[0060] Modeling: Establish a hierarchical model of credit scoring indicators based on the analysis purpose, determine the credit scoring indicators that affect the assessment objectives based on analytical statistics, and establish a hierarchical model consisting of objectives and credit scoring indicators.

[0061] like Figure 2The hierarchical model diagram shown in the figure, because there are many actual credit scoring indicators, multiple indicators are classified, and the indicators of the same category are more comparable. For example: personal basic information includes age, education, nature of unit, salary, etc.; asset information includes the number of real estates, the number of vehicles, real estate valuation, vehicle valuation, etc.; personal credit information includes the number of bank cards, the number of overdue payments, historical loan amounts, etc. The evaluation target is personal credit.

[0062] A credit scoring indicator matrix is ​​established. By comparing the importance of the upper and lower credit scoring indicators, a weight matrix A is established, where aij represents the importance of ai to aj, and the value range is 1 to 9 and its reciprocal. 1 means that the former is equally important to the latter, and 9 means that the former has the greatest difference in importance with the latter and the former is important.

[0063] In the actual operation process, the weight matrix is ​​first established for the indicator category layer, the weight value of each indicator category is calculated, and then the weight matrix is ​​established for the indicators under each indicator category to obtain the weight of each indicator. Figure 2 In this example, four weight matrices need to be established.

[0064]

[0065] Calculate the weight value and obtain the data sequence value of the lower credit score indicator to the upper credit score indicator through the square root method.

[0066] The steps include: 1. Get the product of the elements in each row. 2. Calculate the nth square root. 3. Normalize the vector and find the approximate weight value. 4. Calculate the maximum eigenvalue λmax.

[0067] Check consistency. The consistency value calculation formula is:

[0068] CI=λmax-n / n-1

[0069] Where CI is the consistency value; λ max is the largest characteristic root; n is the order of A (characteristic matrix), when A is completely consistent, CI = 0; λ max The larger the -n, the worse the consistency.

[0070] The calculation formula to determine whether the consistency is satisfactory is:

[0071] CR=CI / RI

[0072] Where CI is the consistency value; RI is the index value of average random consistency, which is a set of constants (first order: 0.00, second order: 0.00, third order: 0.52, fourth order: 0.89...); CR<0.1 means that the consistency meets the requirements; if the consistency does not meet the requirements, it is necessary to adjust the weight matrix and repeat the calculation until the consistency meets the requirements.

[0073] Using the same method, we first construct a weight matrix for all indicator types, and then construct a weight judgment matrix for each indicator type to obtain the weight q value of each characteristic indicator.

[0074] like Figure 2 A weight matrix is ​​established for the indicator category layer, and the normalized weight values ​​of each indicator category are calculated as W1, W2, and W3; then a weight matrix is ​​established for each of the three indicator categories to obtain the weight values ​​of the indicator to the indicator category, for example, the weight of indicator four to indicator category one is W 41 .

[0075] The credit scoring indicator matrix includes:

[0076] A weight matrix is ​​established for the indicator category layer, and the weight value of each indicator category is calculated.

[0077] A weight matrix is ​​established for each indicator category to obtain the weight of each indicator.

[0078] S4. Calculate the lender’s classified credit score and total credit score by weight, and generate a radar user profile.

[0079] The lender's classified credit score and total credit score are calculated by weight, and a radar user profile is generated, including:

[0080] First, the risk assessment value of each indicator class is obtained.

[0081] Then, according to the weight of each indicator class and the risk assessment value of the indicator class, the user's comprehensive risk assessment value is calculated through a weighted algorithm:

[0082] L=L1*W1+L2*W2+...+L n *W n

[0083] Finally, the risk assessment value L for each classification is obtained 1， L 2... L n And the comprehensive risk assessment value L generates the credit score profile of the radar user.

[0084] The risk assessment value of each indicator class is obtained by a weighted algorithm according to the following calculation formula:

[0085] L i =(l i1 *W i1 +l i2 *W i2 +...+l im *W im )*100

[0086] Where: L irepresents the evaluation value of the ith indicator class; if there are m lender credit indicators in the ith indicator class, l im Indicates the value of the mth credit indicator in this indicator class, W im Indicates the value weight of the mth credit indicator in this indicator class.

[0087] After the indicator set is segmented, we get N indicator classes. During the data processing, we get the weight of each indicator class and the weight of each indicator. Then we can use the weighted algorithm to get the indicator class lender risk assessment score for each indicator class, and then use the weighted algorithm to calculate the lender comprehensive assessment score for all indicator classes. After getting the assessment score of the indicator class and the richness of the comprehensive assessment, we can use these scoring results to generate radar user portraits.

[0088] This embodiment makes a radar chart of the user's category credit score based on the score of each credit category, forms a credit profile of the lender, intuitively displays the lender's credit status in all aspects, and assists in lender classification and loan approval decision-making.

[0089] When launching different types of loan products, different types of credit indicators of users are given different attention. Take mortgage loans and credit loans as examples. When the loan product is a mortgage loan, in the process of evaluating the credit score of the borrower, we can increase the proportion of the borrower's personal deposit asset indicators to make the borrower's personal deposit asset indicators play a more significant role in the credit scoring process, so as to screen out users who are more suitable for such products after the score is sorted. Similarly, in the evaluation process of credit loan products, we will make personal credit and basic information indicators occupy a heavier proportion to screen out users who are suitable for such products. The calculation of indicator weights by the analytic hierarchy process greatly reduces the computational complexity of the weight matrix by classifying the indicators. For example, there are 30 indicators, which are divided into 4 indicator classes after indicator classification. The number of indicators in each indicator class is 7, 9, 6, and 8 respectively. Then the computational complexity is changed from the original 30-order operation to the highest 9-order operation, which greatly improves the efficiency of the calculation and reduces the computational overhead and time.

[0090] In order to better understand the solution of this embodiment, an application case is provided as follows:

[0091] The information of customer Zhang San includes basic information: education level, income, nature of unit; deposit information: annual inflow, annual outflow, balance; litigation information: number of litigation processes, number of litigation judgments, number of executed cases, amount of executed cases; credit information: bad debt amount, total credit amount, remaining balance.

[0092] Indicator weights: W1 (basic information): 0.13, W2 (deposit information): 0.25, W3 (litigation information): 0.24, W4 (credit information): 0.38.

[0093] Indicator weight, W 11 (Education): 0.17, W 12 (income): 0.68, W 13 (Unit nature): 0.15...

[0094] Index value, l 11 (Education): Postgraduate, 12 (Income): 25,000 yuan, l 13 (Nature of unit): State-owned enterprise...

[0095] The quantified and standardized indicator value, l 11 (Education): 0.75 (3-0 / 4-0), l 12 (income): 0.5(25000-0 / 50000-0), l 13 (Unit nature): 1(5-0 / 5-0)...

[0096] Calculate the score: L1 (basic information) = (0.75*0.17+0.5*0.68+1*0.15)*100 = 61.75 Calculate the comprehensive score: L = (60.31*0.13+L2*0.25+L3*0.24+L4*0.38)*100

[0097] Practical applications: 1. Comprehensive score ranking is used for comprehensive credit assessment of lenders and provides decision-making suggestions from all aspects. 2. Comprehensive score ranking and credit information score ranking are used to assist loan review.

[0098] Application effect: In the process of comprehensive customer evaluation, the scores of various indicators of customers can be displayed intuitively. The radar chart can help business personnel better understand the basic situation of customers and help business personnel make decisions. For example, when the score of customer credit information indicators is low, the loan can be rejected during loan review, saving a lot of time.

[0099] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A method for calculating a lender's credit score based on lender information, characterized in that: The method comprises: Processing credit classification of lenders; Quantification of credit scoring indicators; Calculate the credit classification and credit scoring indicator weights using the analytic hierarchy process; Calculate the lender's classified credit score and total credit score by weight, and generate a radar user profile.

2. A method for calculating a lender's credit score based on lender information as claimed in claim 1, characterized in that: Credit scoring indicators include: Quantify the enumeration type credit score indicators. Select unique hot encoding to quantify the enumeration type credit score indicators. Through quantization, the enumeration type credit score indicators are converted into numerical types. The credit scoring indicators are standardized, the original data is scaled, and data normalization is performed to map the data to the [0, 1] space.

3. A method for calculating a lender's credit score based on lender information as claimed in claim 1, characterized in that: The AHP method for calculating the credit classification and credit scoring index weights includes: Modeling: Establish a hierarchical model of credit scoring indicators based on the analysis purpose, determine the credit scoring indicators that affect the evaluation objectives based on the analysis statistics, and establish a hierarchical model consisting of objectives and credit scoring indicators; Establish a credit scoring indicator matrix, and establish a weight matrix by comparing the importance of the upper and lower credit scoring indicators; Calculate the weight value and obtain the data sequence value of the lower credit score indicator to the upper credit score indicator through the square root method; Check consistency. The consistency value calculation formula is: CI=λ max -n / n-1 Where CI is the consistency value; λ max is the largest characteristic root; n is the order of A (characteristic matrix), when A is completely consistent, CI = 0; λ max -The larger the n, the worse the consistency; The calculation formula to determine whether the consistency is satisfactory is: CR=CI / RI Where CI is the consistency value; RI is the index value of average random consistency, which is a set of constants (first order: 0.00, second order: 0.00, third order: 0.52, fourth order: 0.89...); CR<0.1 means that the consistency meets the requirements; if the consistency does not meet the requirements, it is necessary to adjust the weight matrix and repeat the calculation until the consistency meets the requirements.

4. A method for calculating a lender's credit score based on lender information as claimed in claim 3, characterized in that: The credit scoring indicator matrix includes: A weight matrix is ​​established for the indicator category layer, and the weight value of each indicator category is calculated; A weight matrix is ​​established for each indicator category to obtain the weight of each indicator.

5. A method for calculating a lender's credit score based on lender information as claimed in claim 1, characterized in that: The lender's classified credit score and total credit score are calculated by weight, and a radar user profile is generated, including: First, obtain the risk assessment value of each indicator class; Then, according to the weight of each indicator class and the risk assessment value of the indicator class, the user's comprehensive risk assessment value is calculated through a weighted algorithm; Finally, the risk assessment value L for each classification is obtained 1， L 2... L n And the comprehensive risk assessment value L generates the credit score profile of the radar user.

6. A method for calculating a lender's credit score based on lender information as claimed in claim 5, characterized in that: The risk assessment value of each indicator class is obtained by a weighted algorithm according to the following calculation formula: L i =(l i1 *W i1 +l i2 *W i2 +...+l im *W im )*100 Where: L i represents the evaluation value of the ith indicator class; if there are m lender credit indicators in the ith indicator class, l im Indicates the value of the mth credit indicator in this indicator class, W im Indicates the value weight of the mth credit indicator in this indicator class.

Images