Risk monitoring method and system after loan guarantee

Abstract

The invention discloses a risk monitoring method and system after loan guarantee, and relates to the technical field of big data processing, target object data is acquired, the target object data is preliminarily screened by using a k-means clustering algorithm, and a preliminary screening variable is acquired; using a Pearson's correlation coefficient to check the linear relationship of the preliminary screening variables, and using a distance correlation coefficient to check the nonlinear relationship of the preliminary screening variables to obtain a final variable; performing level classification on the final variable to divide the final variable into variable sets of all levels, and dividing the variable sets of all levels again according to risk attributes to obtain variable sets of different types; and obtaining risk signals corresponding to different levels and types of final variables, and calculating a guarantee risk value of the target object by using a hierarchical risk assessment mechanism. According to the invention, risk monitoring after intelligent loan guarantee can be carried out, and risk analysis can be carried out on the target object and the corresponding group.

Description

Technical Field

[0001] The present invention relates to the technical field of big data processing, and more particularly to a risk monitoring method and system after loan guarantee. Background Art

[0002] With the development of the economy, loan efficiency and loan amount are gradually increasing. In order to ensure effective repayment after the loan, loan guarantee has also become a common means of loan. Among them, loan guarantee refers to the legal act of requiring the borrower to provide guarantee when issuing a loan to ensure the realization of the loan debt.

[0003] However, there are still certain risks after the loan guarantee. In order to timely discover early warning signals and actively take corresponding remedial measures, the guarantee company still needs to continuously monitor and analyze the guarantors and related factors that affect the security of the guarantee business. However, monitoring and analysis in existing technologies requires a lot of manpower, and the efficiency is particularly low.

[0004] Therefore, how to conduct intelligent risk monitoring and analysis is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides a method and system for risk monitoring after loan guarantee, which can perform risk monitoring after intelligent loan guarantee, and perform risk analysis on target objects and corresponding groups.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A risk monitoring method after loan guarantee, comprising:

[0008] Obtain target object data, and preliminarily screen the target object data using a k-means clustering algorithm to obtain preliminary screening variables;

[0009] The linear relationship of the preliminary screening variables was tested using the Pearson correlation coefficient, and the nonlinear relationship of the preliminary screening variables was tested using the distance correlation coefficient to obtain the final variables;

[0010] Classifying the final variables into variable sets at different levels, and further classifying the variable sets at different levels according to risk attributes to obtain variable sets of different types;

[0011] The risk signals corresponding to the final variables of different levels and types are obtained, and the guarantee risk value of the target object is calculated using a hierarchical risk assessment mechanism.

[0012] Preferably, the obtaining of the initial screening variables specifically includes:

[0013] S1.1, use the target object data as the training set of the k-means algorithm and initialize k cluster centers;

[0014] S1.2, calculate the Euclidean distance from the variable to the centroid of each cluster, assign it to the cluster with the closest distance, and recalculate the centroid of the cluster;

[0015] S1.3, repeating steps S1.1 and S1.2 until the centroid of each cluster no longer changes;

[0016] S1.4, draw a curve graph of the sum of squared errors (SSE) and k value, use the elbow rule to determine the optimal number of clusters, and select the centroid of each cluster as the initial screening variable.

[0017] Preferably, the obtaining of the final variable specifically includes: calculating the Pearson correlation coefficient between each variable and other variables to test linear correlation, calculating the distance correlation coefficient between each variable and other variables to test nonlinear correlation, obtaining the Pearson and distance correlation matrices of the variables based on the linear and nonlinear correlations, calculating the mean of the correlation coefficients of each variable and other variables, eliminating variables whose means are greater than a threshold, and obtaining the final variable.

[0018] Preferably, the calculation of the guaranteed risk value of the target object specifically includes:

[0019] Obtain risk signals corresponding to different types of variable sets in variable sets at each level, monitor the risk monitoring data of the final variable in the target object data in real time, set a risk assessment cycle, and regularly summarize the risk monitoring data;

[0020] A hierarchical risk assessment mechanism is established, based on the variable risk value of the target object, and the risk values ​​corresponding to the variable set and the variable collection are calculated in sequence.

[0021] Preferably, based on the variable risk value of the target object, the risk values ​​corresponding to the variable set and the variable collection are calculated in sequence, specifically including: obtaining the risk monitoring data of each target object, and preferentially calculating the variable risk value of each target object; according to the hierarchical classification of the target object, calculating the set risk value of the variable collection to which the target object belongs by cumulative calculation and taking the average value.

[0022] Preferably, the level classification of the final variables into variable sets of each level specifically includes: standardizing the final variables and constructing a variable information matrix; setting the number of clustering clusters, performing clustering operations on the final variables using an ant colony clustering algorithm improved by a genetic algorithm, and dividing the final variables into variable sets of different levels.

[0023] A risk monitoring system after loan guarantee includes:

[0024] A primary variable screening module obtains target object data, performs preliminary screening on the target object data using a k-means clustering algorithm, and obtains preliminary screening variables;

[0025] The secondary variable screening module uses the Pearson correlation coefficient to test the linear relationship of the primary screening variables and uses the distance correlation coefficient to test the nonlinear relationship of the primary screening variables to obtain the final variables;

[0026] A variable division module is used to classify the final variables into variable sets of each level, and further divide the variable sets of each level according to risk attributes to obtain variable sets of different types;

[0027] The guarantee risk value calculation module obtains the risk signals corresponding to the final variables of different levels and types, and calculates the guarantee risk value of the target object using a hierarchical risk assessment mechanism.

[0028] Through the above technical solutions, it can be seen that compared with the prior art, the present invention discloses a risk monitoring method and system after loan guarantee, which fully considers the inherent structure and distribution characteristics of the data by obtaining the target object data and using the k-means clustering algorithm for preliminary screening. The algorithm can cluster data points with similar characteristics together, using the cluster center as the initial screening variable, and effectively summarizes the main information and potential risk patterns in the data. This step not only reduces the data dimension and avoids processing a large amount of redundant information in subsequent analysis, but also focuses on key risk factors, laying the foundation for more accurate risk monitoring; when the final variables are classified into variable sets of each level, the ant colony clustering algorithm improved by genetic algorithm has significant advantages. This algorithm combines the global search ability of genetic algorithm and the local optimization ability of ant colony algorithm, and can better handle complex data distribution and multimodal problems; by setting the number of clusters, the algorithm can divide the final variables into variable sets of different levels according to the similarity and difference between variables. Each variable set represents a combination of variables with similar risk characteristics. This classification method makes the risk level division more scientific and reasonable, and can accurately reflect the difference in risk levels of different variable groups; according to the risk attributes, the variable sets of each level are divided again to obtain different types of variable collections, which further deepens the risk classification management. According to different risk attributes (such as financial risk, market risk, credit risk, guarantee risk, etc.), the variable set is subdivided, and more accurate and targeted monitoring strategies can be formulated for different types of risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0030] Figure 1 A diagram of the method steps provided by the present invention;

[0031] Figure 2 A schematic diagram of the structure provided by the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] The embodiment of the present invention discloses a risk monitoring method after loan guarantee, such as Figure 1 As shown, including:

[0034] Obtain target object data, use k-means clustering algorithm to preliminarily screen the target object data, and obtain preliminary screening variables;

[0035] The Pearson correlation coefficient was used to test the linear relationship of the initial screening variables, and the distance correlation coefficient was used to test the nonlinear relationship of the initial screening variables to obtain the final variables;

[0036] The final variables are classified into variable sets at each level, and the variable sets at each level are further divided according to risk attributes to obtain variable sets of different types;

[0037] Obtain risk signals corresponding to final variables of different levels and types, and use the hierarchical risk assessment mechanism to calculate the guarantee risk value of the target object.

[0038] In a specific embodiment, obtaining the initial screening variables specifically includes:

[0039] S1.1, use the target object data as the training set of the k-means algorithm and initialize k cluster centers;

[0040] S1.2, calculate the Euclidean distance from the variable to the centroid of each cluster, assign it to the cluster with the closest distance, and recalculate the centroid of the cluster;

[0041] S1.3, repeat steps S1.1 and S1.2 until the centroid of each cluster no longer changes;

[0042] S1.4, draw a curve graph of the sum of squared errors (SSE) and k value, use the elbow rule to determine the optimal number of clusters, and select the centroid of each cluster as the initial screening variable.

[0043] In a specific embodiment, obtaining the final variable specifically includes: calculating the Pearson correlation coefficient between each variable and other variables to test linear correlation, calculating the distance correlation coefficient between each variable and other variables to test nonlinear correlation, obtaining the Pearson and distance correlation matrices of the variables based on linear and nonlinear correlations, calculating the mean of the correlation coefficients of each variable and other variables, eliminating variables with means greater than a threshold, and obtaining the final variable.

[0044] Among them, the Pearson correlation coefficient of each variable and other variables is calculated to test the linear correlation:

[0045]

[0046] Calculate the distance correlation coefficient between each variable and other variables to test nonlinear correlation:

[0047]

[0048] Among them, the distance covariance and They are:

[0049]

[0050]

[0051]

[0052] Similarly, calculate the distance covariance and A and B are two variables for correlation analysis, n represents the number of samples, and d represents the distance.

[0053] In a specific embodiment, calculating the guarantee risk value of the target object specifically includes:

[0054] Obtain risk signals corresponding to different types of variable sets in variable sets at each level, monitor the risk monitoring data of the final variables in the target object data in real time, set the risk assessment cycle, and regularly summarize the risk monitoring data;

[0055] A hierarchical risk assessment mechanism is established, based on the variable risk value of the target object, and the risk values ​​corresponding to the variable set and the variable collection are calculated in sequence.

[0056] In a specific embodiment, based on the variable risk value of the target object, the risk values ​​corresponding to the variable set and the variable collection are calculated in sequence, specifically including: obtaining the risk monitoring data of each target object, and preferentially calculating the variable risk value of each target object; according to the hierarchical classification of the target object, calculating the set risk value of the variable collection to which the target object belongs by cumulative calculation and taking the average value.

[0057] In a specific embodiment, classifying the final variables into variable sets of each level specifically includes: standardizing the final variables and constructing a variable information matrix; setting the number of clusters, performing clustering operations on the final variables using an ant colony clustering algorithm improved by a genetic algorithm, and dividing the final variables into variable sets of different levels.

[0058] Among them, the ant colony clustering algorithm improved by genetic algorithm is used to perform clustering operation on the final variables, and the final variables are divided into variable sets of different levels, including:

[0059] The number of clusters is pre-set, and a certain number of virtual ants are randomly placed in the variable information matrix; the walking path of the virtual ants is determined according to the pheromone value and state transition probability in the variable information matrix, and the path is marked. The state transition probability operation expression is:

[0060]

[0061] In the formula, p ij (t) indicates that the virtual ant is clustered from the data X in the tth clustering process. i To data X j The state transition probability of ij (t) indicates that the virtual ant is clustered from the data X in the tth clustering process. i To data X j pheromone; τ is (t) indicates that the virtual ant is clustered from the data X in the tth clustering process. i To data X s pheromone; s represents data X i neighbor nodes; q represents the direct transfer threshold.

[0062] The cluster center in the current iteration process is determined according to the path identification, and the deviation of all measured final variables to the corresponding cluster center is calculated. The deviation operation expression is:

[0063]

[0064] In the formula, J represents the deviation; K represents the number of types of final variables; S(j) represents the set of the jth final variables; C j represents the cluster center of the jth final variable; ωij Represents data X i and the decision variable between the jth final variable.

[0065] The path corresponding to the minimum deviation is mutated, and the new deviation of the cluster center corresponding to all the measured final variables under the path is calculated using the deviation operation expression. If the new deviation is less than the minimum deviation, the mutated path is used as the path selected by the virtual ant in the next iteration. If the new deviation is greater than or equal to the minimum deviation, the original path is returned.

[0066] After each clustering iteration is completed, the pheromone that generates the new path is updated; the iteration is repeated until the deviation is stable or the maximum number of iterations is reached, and the clustering iteration operation is stopped;

[0067] All measured final variables in the variable information matrix are divided into corresponding clusters, so that the measured final variables corresponding to all target objects are divided into variable sets of different levels to achieve level classification.

[0068] After each clustering iteration is completed, the pheromone that generates the new path is updated, including the following steps:

[0069] After each iteration of clustering is completed, all paths generated by virtual ants are sorted in ascending order according to the value of the deviation, and the pheromone concentration on each path is updated;

[0070] The pheromone evaporates according to the preset ratio, and the path where the R virtual ants with the smallest deviation are located is incremented. The information is updated through the pheromone change calculation. The pheromone change calculation expression is:

[0071]

[0072] In the formula, τ ij (t+1) represents the pheromone concentration of the next iteration; 1-u represents the residual degree of pheromone after volatilization; Indicates that the kth virtual ant stays in the data X during the tth clustering iteration i The information on the path to cluster j.

[0073] The Ant Colony Clustering Algorithm is a heuristic algorithm inspired by the behavior of ants when looking for food. It is an optimization algorithm based on simulating biological behavior and is often used to solve combinatorial optimization problems, such as graph segmentation, clustering, and path planning. The basic idea of ​​the Ant Colony Clustering Algorithm is to simulate the behavior of ants releasing pheromones in the process of looking for food. Ants release pheromones on the path to guide other ants to choose the same path, and this pheromone will gradually evaporate over time. When the ant chooses a path, the volatilization rate of the pheromone and the length of the path will affect the path chosen by the ant next time. In this way, over time, the pheromone concentration on shorter paths will increase, causing more ants to choose these paths, and eventually forming a convergence of paths. In the Ant Colony Clustering Algorithm, there are usually the following key concepts:

[0074] Ant: represents a possible solution in the search space and can be regarded as a local solution.

[0075] Path: represents the sequence of nodes that the ant passes through, usually corresponding to a solution in the problem.

[0076] Pheromone: indicates the quality of the path and is the basis for ants to choose the path. Pheromones accumulate and evaporate along the path.

[0077] Pheromone Update: After the ant selects a path, it updates the pheromone concentration based on the path quality. The better the path, the higher the pheromone concentration.

[0078] The basic process of the ant colony clustering algorithm includes initializing pheromones, ant path selection, pheromone update and other steps. When applied to clustering problems, ants will select appropriate clusters for clustering based on a certain distance metric (such as Euclidean distance). Through multiple iterations, the algorithm will gradually converge and eventually obtain a set of clustering results.

[0079] In a specific embodiment, the final variables include, for example:

[0080] Debt-paying ability indicators: debt-to-asset ratio (ratio of total liabilities to total assets), reflecting the long-term debt-paying ability of an enterprise; current ratio (ratio of current assets to current liabilities) and quick ratio (ratio of quick assets to current liabilities, quick assets are current assets after deducting inventory), measuring the short-term debt-paying ability of an enterprise; interest coverage ratio (ratio of earnings before interest and taxes to interest expenses), reflecting the ability of an enterprise to pay interest. Changes in the values ​​of these indicators directly affect the ability of an enterprise to repay its loans and are an important basis for assessing the risk of loan guarantees. For example, an excessively high debt-to-asset ratio may mean that the enterprise has a greater financial risk, and when faced with market fluctuations or operational difficulties, it may not be able to repay the principal and interest of the loan in full and on time, thereby increasing the risk of loan guarantees.

[0081] Profitability indicators: including gross profit margin (ratio of gross profit to operating income), net profit margin (ratio of net profit to operating income), return on net assets (ratio of net profit to average net assets), etc. Gross profit margin reflects the initial profitability of the company's products or services; net profit margin reflects the company's final profitability after deducting all costs and expenses; return on net assets measures the investment income level of the company's own funds. The strength of profitability is directly related to the company's source of repayment funds. If the company's profitability continues to decline, it may lead to tight cash flow, making it difficult to fulfill loan repayment obligations, and posing a threat to loan guarantees.

[0082] Operating capacity indicators: Inventory turnover rate (the ratio of operating costs to average inventory balance) reflects the efficiency of inventory management and capital turnover speed of the enterprise; accounts receivable turnover rate (the ratio of net credit sales income to the average balance of accounts receivable) reflects the recovery speed and management level of accounts receivable of the enterprise; total asset turnover rate (the ratio of operating income to average total assets) measures the operating efficiency of the overall assets of the enterprise. Poor operating capacity may lead to increased capital occupation and slow capital recovery, affecting the normal operation and debt repayment ability of the enterprise, and further affecting the security of loan guarantees. For example, a too low inventory turnover rate may indicate that the enterprise has an inventory backlog problem, which occupies a large amount of funds and increases financial costs and operating risks.

[0083] Credit score: A score that is comprehensively assessed by a professional credit rating agency or a financial institution's internal credit assessment system based on a company or individual's credit history, repayment record, debt situation, etc., which directly reflects their credit status. Borrowers with high credit scores usually have low credit risks and are more likely to repay their loans in full and on time; low credit scores indicate high credit risks and a possible tendency to default, posing a greater threat to loan guarantees.

[0084] Overdue records: Detailed records of the number of overdue payments, duration of overdue payments, overdue amounts, and other information in the borrower's past credit activities such as loans, credit card overdrafts, and utility bill payments. Multiple or long-term overdue payments indicate that the borrower has a weak sense of credit and may have problems with repayment ability, increasing the possibility of loan defaults and being an important negative factor in loan guarantee risk assessment. For example, if an enterprise has repeatedly overdue loan repayments in the past year, financial institutions will be highly cautious when evaluating its new loan guarantee application and believe that its loan guarantee risk is high.

[0085] Number of credit inquiries and reasons: the number of inquiries on the borrower's credit report and the reasons for the inquiries within a certain period. Too many credit inquiries may indicate that the borrower has urgent capital needs in the near future, frequently applies for credit or other credit services, or may have hidden credit risks, resulting in multiple institutions conducting credit reviews on it. In particular, frequent inquiries due to loan approval, credit card application, etc. may increase the credit risk of the borrower and affect the loan guarantee risk assessment. For example, if an individual's credit report is frequently inquired by multiple financial institutions in a short period of time, and all of them are loan application inquiries, it may indicate that his financial situation is unstable or his debt repayment ability is questioned, and the loan guarantee risk increases accordingly.

[0086] A risk monitoring system after loan guarantee, such as Figure 2 As shown, including:

[0087] A variable screening module obtains the target object data, uses the k-means clustering algorithm to preliminarily screen the target object data, and obtains the initial screening variables;

[0088] The secondary variable screening module uses the Pearson correlation coefficient to test the linear relationship of the initial screening variables, and uses the distance correlation coefficient to test the nonlinear relationship of the initial screening variables to obtain the final variables;

[0089] The variable division module classifies the final variables into variable sets of each level, and further divides the variable sets of each level according to risk attributes to obtain variable sets of different types;

[0090] The guarantee risk value calculation module obtains the risk signals corresponding to the final variables of different levels and types, and calculates the guarantee risk value of the target object using a hierarchical risk assessment mechanism.

[0091] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0092] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A risk monitoring method after loan guarantee, characterized in that: include: Obtain target object data, and preliminarily screen the target object data using a k-means clustering algorithm to obtain preliminary screening variables; The linear relationship of the preliminary screening variables was tested using the Pearson correlation coefficient, and the nonlinear relationship of the preliminary screening variables was tested using the distance correlation coefficient to obtain the final variables; Classifying the final variables into variable sets of each level, and further dividing the variable sets of each level according to risk attributes to obtain variable sets of different types; The risk signals corresponding to the final variables of different levels and types are obtained, and the guarantee risk value of the target object is calculated using a hierarchical risk assessment mechanism.

2. A risk monitoring method after loan guarantee according to claim 1, characterized in that: The obtaining of the initial screening variables specifically includes: S1.1, use the target object data as the training set of the k-means algorithm and initialize k cluster centers; S1.2, calculate the Euclidean distance from the variable to the centroid of each cluster, assign it to the cluster with the closest distance, and recalculate the centroid of the cluster; S1.3, repeating steps S1.1 and S1.2 until the centroid of each cluster no longer changes; S1.4, draw a curve graph of the sum of squared errors (SSE) and k value, use the elbow rule to determine the optimal number of clusters, and select the centroid of each cluster as the initial screening variable.

3. The risk monitoring method after loan guarantee according to claim 1 is characterized in that: The obtaining of the final variable specifically includes: calculating the Pearson correlation coefficient of each variable with other variables to test linear correlation, calculating the distance correlation coefficient of each variable with other variables to test nonlinear correlation, obtaining the Pearson and distance correlation matrices of the variables based on the linear and nonlinear correlations, calculating the mean of the correlation coefficients of each variable with other variables, eliminating variables with a mean greater than a threshold, and obtaining the final variable.

4. The risk monitoring method after loan guarantee according to claim 1 is characterized in that: The calculation of the guaranteed risk value of the target object specifically includes: Obtain risk signals corresponding to different types of variable sets in variable sets at each level, monitor the risk monitoring data of the final variable in the target object data in real time, set a risk assessment cycle, and regularly summarize the risk monitoring data; A hierarchical risk assessment mechanism is established, based on the variable risk value of the target object, and the risk values ​​corresponding to the variable set and the variable collection are calculated in sequence.

5. A method for monitoring risk after loan guarantee according to claim 4, characterized in that: Based on the variable risk value of the target object, the risk values ​​corresponding to the variable set and the variable collection are calculated in sequence, specifically including: obtaining the risk monitoring data of each target object, and preferentially calculating the variable risk value of each target object; according to the hierarchical classification of the target object, calculating the set risk value of the variable collection to which the target object belongs by cumulative calculation and taking the average value.

6. The risk monitoring method after loan guarantee according to claim 1 is characterized in that: The step of classifying the final variables into sets of variables at different levels specifically includes: performing standardization preprocessing on the final variables and constructing a variable information matrix; setting the number of clusters, performing clustering operations on the final variables using an ant colony clustering algorithm improved by a genetic algorithm, and dividing the final variables into sets of variables at different levels.

7. A risk monitoring system after loan guarantee, using a risk monitoring method after loan guarantee according to any one of claims 1 to 6, characterized in that: include: A primary variable screening module obtains target object data, performs preliminary screening on the target object data using a k-means clustering algorithm, and obtains preliminary screening variables; The secondary variable screening module uses the Pearson correlation coefficient to test the linear relationship of the primary screening variables and uses the distance correlation coefficient to test the nonlinear relationship of the primary screening variables to obtain the final variables; A variable division module is used to classify the final variables into variable sets of each level, and further divide the variable sets of each level according to risk attributes to obtain variable sets of different types; The guarantee risk value calculation module obtains the risk signals corresponding to the final variables of different levels and types, and calculates the guarantee risk value of the target object using a hierarchical risk assessment mechanism.

Images