CN112508679A - Small and micro enterprise loan risk assessment method and device and storage medium - Google Patents

Abstract

The present application discloses a small and micro enterprise loan risk assessment method, device and storage medium, which relates to the technical field of credit approval and solves the problems of inaccurate assessment of the repayment ability of small and micro enterprises in the prior art and the occupation of a large amount of labor force ; The method includes: obtaining the electricity operation status data and loan data of small and micro enterprises; constructing a loan evaluation index system for small and micro enterprises; assigning weights to each evaluation index in the loan evaluation index system, and determining the comprehensive score of the enterprise. The enterprise comprehensive score is equal to the weighted sum of multiple evaluation indicators; the monitoring threshold is set according to the enterprise comprehensive score to monitor the pre-loan and post-loan risks of small and micro enterprises; this method effectively solves the inaccurate assessment of the repayment ability of the staff, and The problem of occupying a lot of labor is realized, and the loan risk of small and micro enterprises can be determined according to the data of electricity operation status and other data, which improves the accuracy of risk assessment and saves a lot of personnel costs.

Description

A kind of small and micro enterprise loan risk assessment method, device and storage medium

technical field

The present application relates to the technical field of credit approval, and in particular, to a method, device and storage medium for loan risk assessment of small and micro enterprises.

Background technique

Loans are funds provided by financial institutions to consumers. Consumers return funds according to certain interest rates and repayment conditions. Financial institutions release monetary funds through loans to meet the needs of the society to expand production for supplementary funds and promote economic development. Applying for a loan has also become a consumption habit of most consumers today.

At present, with the development of society, there are more and more small and micro enterprises, small and micro enterprises have become the main force in solving the employment rate, and small and micro enterprises have also become the main group with loans. For financial institutions, how to deal with small and micro enterprises? Assessing repayment ability is the primary issue.

The existing evaluation method is based on the investigator's inspection of the qualifications, credit and property status of the small and micro enterprises, that is, the manual review of the loan information of the small and micro enterprises, resulting in inaccurate assessment of the repayment ability of the staff, and occupying a large amount of money. labor force.

SUMMARY OF THE INVENTION

The embodiments of the present application solve the problems of inaccurate assessment of the repayment ability of small and micro enterprises in the prior art and a large amount of labor occupied by providing a small and micro enterprise loan evaluation method, device and storage medium.

In a first aspect, an embodiment of the present invention provides a small and micro enterprise loan evaluation method, characterized in that it includes:

Obtain electricity operation status data and loan data of small and micro enterprises;

constructing a loan evaluation index system for the small and micro enterprises;

A weight is assigned to each evaluation index in the loan evaluation index system, and a comprehensive enterprise score is determined, and the comprehensive enterprise score is equal to the weighted sum of a plurality of the evaluation indicators;

A monitoring threshold is set according to the comprehensive score of the enterprise to monitor the pre-loan and post-loan risks of the small and micro enterprise.

With reference to the first aspect, in a possible implementation manner, the weight assigned to each evaluation index in the loan evaluation index system includes:

The weight of each of the evaluation indicators is determined by using the analytic hierarchy process or the entropy method.

With reference to the first aspect, in a possible implementation manner, the weight assigned to each evaluation index in the loan evaluation index system includes:

A comprehensive weighting method is used to determine the weight of each of the evaluation indicators, and the function of the comprehensive weighting method is:

Among them, W _j is the weight value of the comprehensive weighting, W _1j is the weight value determined by the AHP and TOPSIS method, and W _2j is the weight value determined by the entropy value method and the TOPSIS method.

With reference to the first aspect, in a possible implementation manner, the setting of the monitoring threshold according to the enterprise comprehensive score includes:

The gradient boosting tree method is used to predict the monthly electricity consumption of the small and micro enterprises, and then the monitoring threshold is set.

In combination with the first aspect, in a possible implementation manner, the method further includes:

Associating the power consumption operating status data with the loan data, and filtering fields to form a data wide table;

Data is processed on the basis of the data wide table to resolve sample imbalance, handle outliers and fill in missing values.

With reference to the first aspect, in a possible implementation manner, the data processing on the basis of the data wide table includes:

Adopt undersampling method or adjusting weight sampling method to solve the sample imbalance phenomenon;

Outliers are handled by the box-plot method;

Missing values were filled with linear regression.

With reference to the first aspect, in a possible implementation manner, the scoring system includes: payment credit, electricity consumption stability, industry prosperity, electricity consumption growth, and default electricity consumption information.

In a second aspect, an embodiment of the present invention provides a small and micro enterprise loan evaluation device, the device comprising:

A data acquisition unit, which is used to acquire the power consumption operation status data and loan data of small and micro enterprises;

a system construction unit, used for constructing the loan evaluation index system of the small and micro enterprises;

a comprehensive scoring unit, configured to assign a weight to each evaluation index in the loan evaluation index system, and determine a comprehensive enterprise score, where the comprehensive enterprise score is equal to the weighted sum of a plurality of the evaluation indicators;

A monitoring plan generation unit, configured to set a monitoring threshold according to the comprehensive score of the enterprise, so as to monitor the pre-loan and post-loan risks of the small and micro enterprise.

With reference to the second aspect, in a possible implementation manner, the comprehensive scoring unit is specifically used for:

Using the AHP and/or the entropy method, determine the weight of each of the evaluation indicators, and then determine the comprehensive enterprise score.

With reference to the second aspect, in a possible implementation manner, the comprehensive scoring unit is specifically used for:

The weight of each of the evaluation indicators is determined by the comprehensive weighting method, and then the comprehensive score of the enterprise is determined. The function of the comprehensive weighting method is:

Among them, W _j is the weight value of the comprehensive weighting, W _1j is the weight value determined by the AHP and TOPSIS method, and W _2j is the weight value determined by the entropy value method and the TOPSIS method.

With reference to the second aspect, in a possible implementation manner, the monitoring scheme generating unit is specifically configured to: predict the monthly electricity consumption of the small and micro enterprise by using a gradient boosting tree method, and then set the monitoring threshold.

With reference to the second aspect, in a possible implementation manner, the apparatus further includes a data wide table construction unit and a data processing unit:

The data wide table constructing unit is used for correlating the power consumption operation status data and the loan data, and filtering fields to form a data wide table;

The data processing unit is configured to process the data on the basis of the data wide table, so as to solve the phenomenon of sample imbalance, process outliers and fill in missing values.

With reference to the second aspect, in a possible implementation manner, the data processing unit is specifically configured to:

Adopt undersampling method or adjusting weight sampling method to solve the sample imbalance phenomenon;

Outliers are handled by the box-plot method;

Missing values were filled with linear regression.

With reference to the second aspect, in a possible implementation manner, the loan evaluation index system includes: payment credit, electricity consumption stability, industry prosperity, electricity consumption growth, and default electricity consumption information.

In a third aspect, an embodiment of the present application provides an AHP-based scoring device, the device including a memory and a processor;

the memory for storing computer-executable instructions;

The processor is configured to execute the computer-executable instructions to implement the method of the first aspect and various possible implementations of the first aspect.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores executable instructions, and when a computer executes the executable instructions, the first aspect and the first aspect can be implemented. Aspects various possible implementations of the methods described.

A technical solution provided in the embodiment of the present invention has at least the following technical effects or advantages:

The embodiment of the present invention provides a loan risk assessment method for small and micro enterprises. The method constructs a loan evaluation index system for small and micro enterprises by using power consumption operation status data and loan data, and evaluates the value of each evaluation index in the loan evaluation index system. After the weight is assigned, the comprehensive enterprise score is determined. After setting the monitoring threshold through the enterprise comprehensive score, the pre-loan and post-loan risks of small and micro enterprises can be monitored, thereby effectively solving the problem of inaccurate assessment of the repayment ability of the staff and the occupation of a large amount of labor. It realizes the loan risk assessment of small and micro enterprises based on the data of electricity operation status and other data, which improves the accuracy of risk assessment and saves a lot of personnel costs.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments of the present invention. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without any creative effort.

FIG. 1 is a flowchart of a small and micro enterprise loan evaluation method provided by the embodiment of the present application;

2 is a flowchart of the hierarchical analysis method of the loan evaluation method for small and micro enterprises provided by the embodiment of the present application;

3 is a flowchart of the entropy method of the loan evaluation method for small and micro enterprises provided by the embodiment of the present application;

4 is a flowchart of the TIPSIS method of the small and micro enterprise loan evaluation method provided in the embodiment of the present application;

5 is a flowchart of the comprehensive scoring method of the small and micro enterprise loan evaluation method provided by the embodiment of the present application;

6 is a line chart of monthly electricity consumption forecast made by the XGBoost algorithm of the small and micro enterprise loan evaluation method provided by the embodiment of the present application;

7 is a flow chart of the specific processing data of the small and micro enterprise loan evaluation method provided by the embodiment of the present application;

FIG. 8 is a schematic diagram of the evaluation system of the loan evaluation method for small and micro enterprises provided by the embodiment of the present application;

FIG. 9 is a schematic diagram of a loan evaluation device for small and micro enterprises provided by the embodiment of the present application;

FIG. 10 is a schematic diagram of a small and micro enterprise loan evaluation entity device according to an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

This application provides a small and micro enterprise loan evaluation method. As shown in FIG. 1 , the method includes the following steps S101 to S104, and the specific steps are as follows.

Step S101 , obtaining power consumption operation status data and loan data of the small and micro enterprises.

Step S102, constructing a loan evaluation index system for small and micro enterprises.

In step S103, a weight is assigned to each evaluation index in the loan evaluation index system, and a comprehensive enterprise score is determined, and the comprehensive enterprise score is equal to the weighted sum of multiple evaluation indicators.

In step S104, a monitoring threshold is set according to the comprehensive score of the enterprise to monitor the pre-loan and post-loan risks of the small and micro enterprises.

In the loan risk assessment method for small and micro enterprises provided by the embodiments of the present invention, a loan evaluation index system for small and micro enterprises is constructed by using power consumption operation status data and loan data, and a weight is assigned to each evaluation index in the loan evaluation index system. Enterprise comprehensive score, after setting the monitoring threshold through enterprise comprehensive score, it can monitor the pre-loan and post-loan risks of small and micro enterprises. The loan risk of small and micro enterprises is determined according to the data of electricity operation status and other data, which improves the accuracy of loan risk assessment for small and micro enterprises, and saves a lot of personnel costs.

Specifically, when step S101 is implemented, the data on the operation status of electricity consumption of small and micro enterprises in recent years is obtained, the electricity consumption behavior, bill payment behavior, and breach of contract behavior are counted to obtain relevant indicators, and mutual cooperation with financial institutions is carried out to obtain the Loan data of small and micro enterprises in recent years.

In step S102, the loan evaluation index system of small and micro enterprises is constructed according to the electricity operation status data and loan data obtained in step S101, and specifically according to the electricity consumption growth rate, industry prosperity, electricity consumption in the electricity operation status data Stability, payment credit, defaulted electricity consumption and loan growth to build a loan evaluation index system.

In step S103, specifically, the weights assigned to each evaluation index in the loan evaluation index system include:

The weight of each evaluation index is determined by the analytic hierarchy process or the entropy method.

When the above-mentioned AHP is used to determine the weight of each evaluation index, as shown in Figure 2, the following steps are specifically included.

Step S201, establishing a hierarchical progressive structure model.

Step S202, construct all judgment matrices in each level.

Step S203, ordering and consistency checking of the hierarchical order.

Step S204, total ordering of the hierarchy and consistency check.

When using AHP specifically, an AHP judgment matrix is obtained, five important indicators for credit evaluation of small and micro enterprises are constructed, and the AHP judgment matrix shown in the following table is obtained. The following table is referred to as Table I in the following description.

The scales in Table I are explained as follows: 1 means that the two elements are of equal importance; 3 means that the former is slightly more important than the latter; 5 means that the former is more important than the latter important. The reciprocal means that if the ratio of the importance of element i to element j is a _ij , then the ratio of the importance of element j to element i is

Solve the judgment matrix of each level to get the weight of AHP. After the calculation in the above steps, the weight value of each evaluation index determined by the AHP is obtained.

When specifically using the entropy method, as shown in Figure 3, it is divided into the following steps to calculate the weight of each evaluation index determined by the entropy method.

Step S301, constructing training samples.

Step S302, standardize the training samples.

Step S303: Calculate the proportion value of the index value of the i-th item and the j-th item.

Step S304, calculating the index information entropy.

Step S305, calculating the information entropy redundancy.

Step S306, calculate the index weight.

The training samples constructed in the above step S301 are:

x{{x ₁₁ ,x ₁₂ ,x ₁₃ ...x _1j },{x ₂₁ ,x ₂₂ ,x ₂₃ ...x _2j },{x ₃₁ ,x ₃₂ ,x ₃₃ ...x _3j }. ..{x _i1 ,x _i2 ,x _3i ...x _ij }}.

负向指标：

其中x_i'_j表示标准化后的数据值；min{x_j}表示一组元数中的最小值；max{x_j}表示一组元素的最大值。Specifically, after standardizing the training samples constructed in step S301, it is obtained: the positive index:

Negative indicators:

where x _i ' _j represents the normalized data value; min{x _j } represents the minimum value in a set of arity; max{x _j } represents the maximum value of a set of elements.

其中，y_ij表示第i项指标值与第j项指标值的比重值，m表示元组的个数。然后实施步骤S304：计算指标信息熵。计算公式为：

其中e_j表示指标信息熵，k表示为

Further, step S303 is implemented: the proportion value of the index value of the i-th item and the j-th item is calculated. The calculation formula is:

Among them, y _ij represents the ratio of the i-th index value to the j-th index value, and m represents the number of tuples. Then step S304 is implemented: calculating the index information entropy. The calculation formula is:

where e _j represents the index information entropy, and k represents as

The formula for calculating the information entropy redundancy in step S305 is: d _j =1-e _j . Among them, d _j represents the information entropy redundancy.

其中，W_i表示每个评估指标的权重值，n表示有n个元素。The formula for calculating the index weight in step S306 is:

Among them, Wi represents the weight value of each evaluation index, and _n represents that there are n elements.

After the calculation in the above steps, the index weight value determined by the entropy method can be obtained.

Further, the TOPSIS method (Technique for Order Preference by Similarity to an Ideal Solution) can also be used to determine the weight of each evaluation index, as shown in Figure 4, including the following steps.

Step S401, unify the monotonicity of each evaluation index.

Step S402, normalization processing.

Step S403, performing weighting processing.

Step S404, determine the optimal solution and the worst solution.

Step S405, calculate the distance from each evaluation object to the optimal and worst solutions.

Step S406, calculate the comprehensive evaluation value.

其中，x_ij表示矩阵中i行j列的元素。In step S401, when determining the monotonicity of each same index, the formula is used:

Among them, x _ij represents the element of i row and j column in the matrix.

其中x_ij表示矩阵中i行j列的元素，min_j表示j列的最小元素，max_j表示j列的最大元素。In step S402, the formula for normalization processing is:

where x _ij represents the element in row i and column j in the matrix, min _j represents the smallest element in column j, and max _j represents the largest element in column j.

Further, perform weighting processing on the normalized data. In step S403, the formula for weighting processing is: Z _ij =W _ij *A _ij , where Z _ij represents a weighted score matrix, and W _ij represents an artificial given The weighted value of , A _ij represents the normalized evaluation matrix.

确定最差方案的公式为：

其中，Z⁺表示矩阵Z_ij按行计算的最大解，Z^-表示矩阵Z_ij按行计算的最小解。In step S404, the optimal solution and the worst solution are determined, and the formula for determining the optimal solution is:

The formula for determining the worst case scenario is:

Among them, Z ⁺ represents the maximum solution calculated by the row of the matrix Z _ij , and Z ^- represents the smallest solution of the matrix Z _ij calculated by the row.

与

具体的计算公式为：

其中，X_i表示对象矩阵，

表示对象为最优方案距离，

表示对象为最差方案距离，Z⁺表示最优方案矩阵，Z^-表示最差方案矩阵。In step S405, the distances from each evaluation object to the optimal and worst solutions are calculated, which are respectively

and

The specific calculation formula is:

where X _i represents the object matrix,

Indicates that the object is the optimal solution distance,

Indicates that the object is the worst solution distance, Z ⁺ represents the optimal solution matrix, and Z ^- represents the worst solution matrix.

其中，B_i表示综合评价价值矩阵，

表示对象为最优方案的距离，

表示对象为最方案距离。In step S406, the comprehensive evaluation value is calculated. The formula for calculating the comprehensive evaluation value is:

Among them, B _i represents the comprehensive evaluation value matrix,

represents the distance at which the object is the optimal solution,

Indicates that the object is the optimal distance.

Further, in step S103, the weights assigned to each evaluation index in the loan evaluation index system include:

The weight of each evaluation index is determined by the comprehensive weighting method. The function of the comprehensive weighting method is:

Among them, W _j represents the weight value of the comprehensive weighting, W _1j represents the weight value determined by the AHP and TOPSIS method, and W _2j represents the weight value determined by the entropy value method and the TOPSIS method.

Substitute the weight value obtained by the AHP into the TOPSIS method to obtain the most ideal weight value of the subjective experience method, as shown in the following table, hereinafter referred to as Table II.

serial number index Weights 1 payment credit 0.33 2 power stability 0.06 3 Industry prosperity 0.21 4 Electricity growth rate 0.10 5 Breach of electricity 0.30

Since the optimal weight value calculated according to the subjective experience method cannot be very convincing, the importance of the indicator is more inclined to personal preference, resulting in the deviation of the indicator weight from the actual. Therefore, the entropy value method is introduced to divide the weight distribution of the indicators according to the amount of information contained in the data itself. Use the entropy method and the TOPSIS method to comprehensively calculate the objective optimal weight. The objective weight is shown in the following table, which is hereinafter referred to as Table III.

serial number index Weights 1 payment credit 0.35 2 power stability 0.14 3 Industry prosperity 0.32 4 Electricity growth rate 0.08 5 Breach of electricity 0.11

In the formula of comprehensive weighting, W _1j represents the weight value determined by the AHP and TOPSIS method, and W _2j represents the weight value determined by the entropy value method and the TOPSIS method. W _1j and W _2j are obtained from the above-mentioned Table II and Table III, and the above-mentioned values are substituted into the function formula of the comprehensive weighting method to obtain the final weight value. As shown in the following table, hereinafter referred to as Table IV.

serial number index Weights 1 payment credit 0.34 2 power stability 0.1 3 Industry prosperity 0.265 4 Electricity growth rate 0.09 5 Breach of electricity 0.205

The weights obtained in Table IV are the final weight values of several indicators illustrated.

In step S103, the comprehensive score of the small and micro enterprises is obtained. According to the final weight value in the above table IV, the credit score of the small and micro enterprises is specifically carried out. Since the dimensions of various indicators are inconsistent, it is necessary to further evaluate the obtained data. Perform normalization to eliminate dimensions. The formula for dimension elimination is:

The credit score of the enterprise is obtained from the above formula, and the results shown in the following table are obtained, and the following table is collectively referred to as Table V.

Table V is the pre-loan risk assessment score of the enterprise obtained according to the above steps and algorithms.

In step S104, the monitoring threshold is set according to the comprehensive score of the enterprise, as shown in FIG. 5, including: using the gradient boosting tree method to predict the monthly electricity consumption of the small and micro enterprises, and then setting the monitoring threshold.

Further, using the gradient boosting tree method to predict the monthly electricity consumption of small and micro enterprises after loan can be divided into the following steps.

Step S501, define a regression tree, continuously perform feature splitting to grow a tree, add a tree each time, learn a new function, and fit the residual of the last prediction.

Step S502, define an objective function of XGBoost, which is a gradient boosting algorithm.

Step S503, use the newly generated tree to fit the residual of the previous prediction to obtain the objective function of the prediction score.

Step S504, obtain the first-order derivative and the second-order derivative of the objective function: obtain the optimal prediction score of each leaf node.

Step S505, the optimal prediction score is brought into the objective function, and the minimum loss is obtained.

Step S506, searching for the best branch through a greedy algorithm.

In step S501, after the training is completed, K trees are obtained, and a sample score is predicted:

Among them, K represents the total number of trees, k represents the kth tree, and f _x (x _i ) represents the regression value of a tree at this leaf node.

In step S502, the objective function of XGBoost is:

表示用来衡量预测分数和真实分数的差距，

表示正则化项，T表示叶子节点数，w表示叶子节点的分数，γ表示控制叶子节点的个数，λ表示控制叶子节点的分数不会过大。in,

is used to measure the difference between the predicted score and the true score,

Represents the regularization term, T represents the number of leaf nodes, w represents the score of leaf nodes, γ represents the number of control leaf nodes, and λ represents that the score of control leaf nodes will not be too large.

In step S503, the objective function of the predicted score is:

表示衡量预测分数和真实分数的差值，Ω(f_k)表示正则化项，f_x(x_i)表示一棵树上在这个叶子节点的回归值。in,

represents the difference between the predicted score and the true score, Ω(f _k ) represents the regularization term, and f _x (x _i ) represents the regression value of a tree at this leaf node.

In step S504, the optimal prediction score of each leaf node is:

where _gi and _hi represent removal constants, respectively:

其中，obj表示树的节点分数，T表示叶子节点数，γ表示控制叶子节点的个数，λ表示控制叶子节点的分数不会过大，g_i和h_i分别表示，移除常量：

In step S505, the minimum loss obtained is:

Among them, obj indicates the node score of the tree, T indicates the number of leaf nodes, γ indicates the number of control leaf nodes, λ indicates that the score of the control leaf node will not be too large, _gi and _hi respectively represent, remove the constant:

其中，

表示左子树分数，

表示右子树分数，

表示不可分割我们可以拿到的分数，γ表示加入新叶子节点引入的复杂度代价。In step S506, find the best branch, and the difference of the structure score after the branch is:

in,

represents the left subtree fraction,

represents the right subtree score,

Represents the score we can get for indivisible, and γ represents the complexity cost introduced by adding new leaf nodes.

According to the data of electricity consumption operation status of small and micro enterprises in recent years, the XGBoost algorithm is used to calculate the monitoring thresholds of enterprise indicators before and after lending. For example, Figure 6 uses the XGBoost algorithm to predict the electricity consumption of the small and micro enterprise in July 2020, and the electricity consumption is compared with the chain as the parameter to set the fluctuation threshold of electricity consumption.

When making further forecasts, the year-on-year ratio and ring ratio of monthly electricity consumption of small and micro enterprises are used as reference standards for monitoring indicators. % is defined as a yellow warning, and if the fluctuation in any direction is greater than 10%, it is defined as a yellow warning, and if the fluctuation in any direction is less than 5%, no warning will occur.

For example, the following table shows the monitoring situation of some enterprises in July 2020, hereinafter referred to as Table XI.

Company Name Industry monitor month Evaluation level Enterprise 1 manufacturing July 2020 yellow Enterprise 2 construction industry July 2020 normal Enterprise 3 Agroforestry July 2020 red

The small and micro enterprise loan evaluation method provided by the embodiment of the present invention, as shown in FIG. 7 , further includes step S701 and step S702.

Step S701, correlate the power consumption operation status data and loan data, and filter fields to form a data wide table.

In step S702, the data is processed on the basis of the data wide table to solve the phenomenon of sample imbalance, process outliers and fill in missing values.

In step S701, fields that may be used are screened out from the existing small and micro enterprise electricity consumption operation status data and loan data to form a data wide table, so as to lay a solid foundation for data preprocessing on the basis of the wide table .

Further, in step S702, the data is processed on the basis of the data width table, as shown in FIG. 8, including steps S801 to S803.

Step S801, adopting the under-sampling method or the weight-adjusted sampling method to solve the sample imbalance phenomenon.

Step S802, using the box-line method to process outliers.

Step S803, using a linear regression method to fill in the missing values.

In step S801, the sample data often has an unbalanced phenomenon, and the processing methods for this phenomenon are generally divided into: under-sampling, weight adjustment and synthesis of minority over-sampling technology, referred to as SMOTE for short. SMOTE is a method for dealing with unbalanced data samples. Compared with random oversampling processing technology, the advantage of SMOTE is that it can use a minority of samples for analysis and artificially synthesize these small samples to generate new samples and expand to In the data samples, this processing adopts the synthetic minority oversampling technique to deal with the data imbalance.

In step S802, the abnormal value is processed, and the box-line method adopted in this embodiment is used to process the abnormal value. Specifically, a boxplot is made for the electricity consumption of small and micro enterprises, highlighting the data greater than the 99% quantile, and excluding the data less than 1% quantile.

The linear regression in step S803 is one of the most commonly used methods for dealing with regression problems. In fact, it is a linear modeling method, which can be solved by the convex optimization method. Specifically, the solution can be solved by minimizing the following objective function:

where J(θ) is a function of W ^J X and y. The objective function includes two parts, the regular term is used to control the complexity of the model, the loss term is used to measure the fitting error, and the objective function is a convex function of the same class W. The regularization parameter λ>0 provides a compromise between minimizing error and model complexity to avoid overfitting.

The linear regression method can use the following steps.

Step 1: Given a training data sample set D={(x ₁ ,y ₁ ),(x ₂ ,y ₂ ),...,(x _m ,y _m )}, y _i ∈ R, select the initial value θ ₀ , given the convergence tolerance ε, the maximum number of iterations K, and then solve the following optimization problem.

Step 2: Take the following formula to update θ

When |J(θ(K+1))-J(θ(k))|<ε or when k=K, the output is 0, where k is the number of iterations, otherwise step 2 is repeated until the condition is satisfied.

Construct the regression decision function: f(x) = θ ^T X, and use the regression decision function to supplement the missing values to obtain complete data with no missing data values, so that the small and micro enterprises can be scored later.

The scoring system provided in this application is shown in Figure 8, including: payment credit, electricity consumption stability, industry prosperity, electricity consumption growth, default electricity consumption information, and loan increase.

Payment credit includes: pre-paid payment amount, timely payment rate, arrears amount, and average arrears duration.

The power consumption stability includes: enterprise voltage stability and the number of times of sudden increase in power consumption.

Industry prosperity includes: industry prosperity index, industry total electricity consumption.

The growth rate of electricity consumption includes: the year-on-year monthly electricity consumption of enterprises and the month-on-month ratio of monthly electricity consumption of enterprises.

The breached electricity consumption includes: the number of breached electricity consumption, the amount of breached electricity theft, the number of breached electricity theft, and the recovery of electricity.

The loan increase includes: loan increase index and loan amount.

According to the above data items, a scoring system is constructed to evaluate the loan risk of small and micro enterprises.

Although the present application provides method operation steps as described in the above textual description or flow chart, more or less operation steps may be included based on routine or non-creative work. The sequence of steps enumerated in this embodiment is only one way among the execution sequences of many steps, and does not represent the only execution sequence. When an actual device or client product is executed, the methods shown in this embodiment or the accompanying drawings may be executed sequentially or in parallel (for example, a parallel processor or a multi-threaded processing environment).

The embodiment of the present invention also provides a small and micro enterprise loan evaluation device, the evaluation device includes: a data acquisition unit 901 , a system construction unit 902 , a comprehensive scoring unit 903 , and a monitoring plan generation unit 904 .

Among them, the data acquisition unit 901 is used to acquire the power consumption operation status data and loan data of small and micro enterprises; the system construction unit 902 is used to construct the loan evaluation index system of small and micro enterprises; the comprehensive scoring unit 903 is used to evaluate the loan evaluation index system. The weight assigned to each evaluation index determines the comprehensive enterprise score, which is equal to the weighted sum of multiple evaluation indicators; the monitoring plan generation unit 904 is configured to set a monitoring threshold according to the comprehensive enterprise score, so as to monitor the pre-loan and loan balance of small and micro enterprises. post-loan risk.

The comprehensive scoring unit can be specifically used for: using the analytic hierarchy process and/or the entropy value method, to determine the weight of each evaluation index, and then to determine the comprehensive score of the enterprise.

The comprehensive scoring unit can also be specifically used to: determine the weight of each evaluation index by the comprehensive weighting method, and then determine the comprehensive score of the enterprise. The function of the comprehensive weighting method is:

Among them, W _j is the weight value of the comprehensive weighting, W _1j is the weight value determined by the AHP and TOPSIS method, and W _2j is the weight value determined by the entropy value method and the TOPSIS method.

The monitoring plan generation unit is specifically used to: predict the monthly electricity consumption of small and micro enterprises by using the gradient boosting tree method, and then set the monitoring threshold.

The device also includes a data wide table construction unit and a data processing unit: the data wide table construction unit is used for correlating the data of the electricity consumption running state and the loan data, and filtering fields to form a data wide table; the data processing unit is used for Process data to resolve sample imbalances, handle outliers, and impute missing values.

The data processing unit is specifically used for: adopting the undersampling method or adjusting the weight sampling method to solve the sample imbalance phenomenon; adopting the box-line method to deal with outliers; adopting the linear regression method to fill in the missing values.

The loan evaluation index system includes: payment credit, electricity consumption stability, industry prosperity, electricity consumption growth, and default electricity consumption information.

The devices or modules described in the embodiments of the present invention may be specifically implemented by computer chips or entities, or by products with certain functions. For the convenience of description, when describing the above device, the functions are divided into various modules and described respectively. When implementing the present application, the functions of each module may be implemented in one or more software and/or hardware. Of course, a module that implements a certain function can also be implemented by a combination of multiple sub-modules or sub-units.

As shown in FIG. 10 , an embodiment of the present invention further provides a loan evaluation device for small and micro enterprises, including: a memory 1001 and a processor 1002; the memory 1001 is used for storing computer-executable instructions; the processor 1002 is used for executing computer-executable instructions Instructions are provided to implement the loan evaluation method for small and micro enterprises provided in this embodiment.

Embodiments of the present invention further provide a computer-readable storage medium, where the computer-readable storage medium stores executable instructions, and when the computer executes the executable instructions, the small and micro enterprise loan evaluation method provided in this embodiment is implemented.

The above-mentioned storage medium includes but is not limited to random access memory (English: Random Access Memory; referred to as: RAM), read-only memory (English: Read-Only Memory; referred to as: ROM), cache (English: Cache), hard disk (English: Hard Disk Drive; referred to as: HDD) or memory card (English: Memory Card). The memory may be used to store computer program instructions.

The methods, apparatuses or modules described in this application may be implemented in computer readable program code. The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and the memory may be implemented by the (micro)processing computer-readable medium, logic gates, switches, application-specific integrated circuits (English: Application Specific Integrated Circuit; ASIC for short), programmable logic controllers and embedded microcontrollers Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320, and memory controllers can also be implemented as part of the memory's control logic. Those skilled in the art also know that, in addition to implementing the controller in the form of pure computer-readable program code, the controller can be implemented as logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded devices by logically programming the method steps. The same function can be realized in the form of a microcontroller, etc. Therefore, such a controller can be regarded as a hardware component, and the devices included therein for realizing various functions can also be regarded as a structure within the hardware component. Or even, the means for implementing various functions can be regarded as both a software module implementing a method and a structure within a hardware component.

Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the various embodiments may be referred to each other, and each embodiment focuses on the differences from other embodiments. All or part of this application may be used in numerous general purpose or special purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, including A distributed computing environment for any of the above systems or devices, and the like.

The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still The technical solutions are modified, or some or all of the technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the present application.

Claims (10)

1. A method for evaluating loans for small and micro enterprises, comprising: Obtain electricity operation status data and loan data of small and micro enterprises; constructing a loan evaluation index system for the small and micro enterprises; A weight is assigned to each evaluation index in the loan evaluation index system, and a comprehensive enterprise score is determined, and the comprehensive enterprise score is equal to the weighted sum of a plurality of the evaluation indicators; A monitoring threshold is set according to the comprehensive score of the enterprise to monitor the pre-loan and post-loan risks of the small and micro enterprise. 2. The small and micro enterprise loan evaluation method according to claim 1, wherein the weighting of each evaluation index in the loan evaluation index system comprises: The weight of each of the evaluation indicators is determined by using the analytic hierarchy process or the entropy method. 3. The small and micro enterprise loan evaluation method according to claim 1, wherein the weighting of each evaluation index in the loan evaluation index system comprises: A comprehensive weighting method is used to determine the weight of each of the evaluation indicators, and the function of the comprehensive weighting method is:

Among them, W _j is the weight value of the comprehensive weighting, W _1j is the weight value determined by the AHP and TOPSIS method, and W _2j is the weight value determined by the entropy value method and the TOPSIS method. 4. The small and micro enterprise loan evaluation method according to claim 1, wherein the setting monitoring threshold according to the comprehensive score of the enterprise comprises: The gradient boosting tree method is used to predict the monthly electricity consumption of the small and micro enterprises, and then the monitoring threshold is set. 5. The small and micro enterprise loan evaluation method according to any one of claims 1-4, characterized in that, further comprising: Associating the power consumption operating status data with the loan data, and filtering fields to form a data wide table; Data is processed on the basis of the data wide table to resolve sample imbalance, handle outliers and fill in missing values. 6. The small and micro enterprise loan evaluation method according to claim 5, wherein the data processing on the basis of the data wide table comprises: Adopt undersampling method or adjusting weight sampling method to solve the sample imbalance phenomenon; Outliers are handled by the box-plot method; Missing values were filled with linear regression. 7. The loan evaluation method for small and micro enterprises according to claim 1, wherein the loan evaluation index system comprises: payment credit degree, electricity consumption stability, industry prosperity degree, electricity consumption growth degree, and default electricity consumption information . 8. A small and micro enterprise loan evaluation device, characterized in that, comprising: A data acquisition unit, which is used to acquire the power consumption operation status data and loan data of small and micro enterprises; a system construction unit, used for constructing the loan evaluation index system of the small and micro enterprises; a comprehensive scoring unit, configured to assign a weight to each evaluation index in the loan evaluation index system, and determine a comprehensive enterprise score, where the comprehensive enterprise score is equal to the weighted sum of a plurality of the evaluation indicators; A monitoring plan generation unit, configured to set a monitoring threshold according to the comprehensive score of the enterprise, so as to monitor the pre-loan and post-loan risks of the small and micro enterprise. 9. A loan evaluation device for small and micro enterprises, comprising a memory and a processor; the memory for storing computer-executable instructions; The processor is adapted to execute the computer-executable instructions to implement the method of any of claims 1-7. 10. A computer-readable storage medium, wherein the computer-readable storage medium stores executable instructions, and when a computer executes the executable instructions, the method according to any one of claims 1-7 can be implemented .

Images