CN114676936A - Method for predicting default time and related device - Google Patents

Abstract

The present application provides a method for predicting default time and a related device, which relate to the field of artificial intelligence. The method includes: acquiring data of a first customer, where the data of the first customer includes parameters used to reflect the credit risk of the first customer; inputting the above data into a first model to obtain the probability of default of the first customer; When the default probability of the customer is greater than the preset value, the time interval to which the default time of the first customer belongs is predicted by the second model; The model predicts the time to default for the first customer. By predicting the default time of the first customer, financial institutions can formulate more reasonable risk control measures for the default time, which is beneficial to improve the efficiency of credit risk control of financial institutions.

Description

A method for predicting default time and related device

technical field

The present application relates to the field of artificial intelligence, and in particular, to a method for predicting default time and a related device.

Background technique

With the rapid development of my country's economy and finance, consumer credit is gradually accepted by people. In recent years, various credit businesses such as auto loans, education loans, petty cash loans, and beauty loans have flourished. For credit business, financial institutions' credit risk assessment of customers is crucial. In the era of big data, although the data used for credit risk assessment is becoming more and more abundant, it also brings many challenges to credit risk assessment.

The current mainstream credit risk assessment method is to use statistical models to predict whether a customer will default or calculate the customer's default probability, but simply predicting the customer's default probability is not comprehensive enough for financial institutions' risk control, and the efficiency of risk control is not sufficient. high.

SUMMARY OF THE INVENTION

The present application provides a method for predicting default time and a related device. By predicting the default time of a customer, a financial institution can formulate more reasonable risk control measures for the default time, thereby improving the efficiency of credit risk control of the financial institution.

In the first aspect, the present application provides a method for predicting default time, the method can be executed by a server, or can also be executed by a component (such as a chip, a chip system, etc.) configured in the server, or can also be executed by a server capable of A logic module or software implementation that implements all or part of the server functions, which is not limited in this application.

Wherein, the above server is configured with a first model, a second model and at least one third model, the first model is used to predict the default probability of the customer, and the second model is used to predict the time interval to which the customer's default time belongs, The at least one third model corresponds to at least one time interval, and each third model is used to predict the number of defaults and the default time of the customer within the corresponding time interval.

Exemplarily, the method includes: acquiring data of a first customer, the data of the first customer including parameters for reflecting the credit risk of the first customer; inputting the data into the first model, Obtain the default probability of the first customer; when the default probability of the first customer is greater than a preset value, predict the time interval to which the default time of the first customer belongs by using the second model; based on the The time interval to which the default time of the first customer predicted by the second model belongs, and the default time of the first customer is predicted by the corresponding third model.

Based on the above technical solution, the acquired data of the first customer is input into the first model to obtain the default probability of the customer, and when the default probability is greater than the preset value, the second model is used to predict the default of the customer The time interval to which the time belongs, that is, how long the customer may default in the future. Further, based on the time interval to which the predicted default time of the customer belongs, the specific default time of the customer is predicted through the corresponding third model. , not only can predict the default probability of customers, but also further predict the default time for customers who may default, so more comprehensive information can be obtained, which is convenient for financial institutions to formulate more reasonable risk control measures based on this, which is conducive to improving financial institutions. the efficiency of credit risk control.

With reference to the first aspect, in a certain possible implementation manner of the first aspect, the first model is a multivariate logistic model, the second model is a Gaussian mixture model (GMM), and the first model is a Gaussian mixture model (GMM). The three models are autoregressive integrated moving average (ARIMA) models.

In combination with the first aspect, in a possible implementation manner of the first aspect, the data includes one or more of the following: industry category, execution interest rate, loan amount, number of loan periods, gender, age, education, and annual family income , employment status, unit type, residence status, job title, social security mark, and customer rating.

With reference to the first aspect, in a certain possible implementation manner of the first aspect, the first model includes multiple first sub-models, the multiple first sub-models correspond to different customer types, and the customer types are based on the customer The age group or region to which it belongs is determined; the inputting the data into the first model to obtain the probability of default of the first customer includes: determining from the plurality of first sub-models the relationship with the first sub-model A first sub-model corresponding to the customer type of a customer; inputting the data of the first customer into the first sub-model to obtain the default probability of the first customer.

With reference to the first aspect, in a certain possible implementation manner of the first aspect, the second model includes multiple second sub-models, the multiple second sub-models correspond to different customer types, and the customer types are based on the customer The age group or region to which the first customer belongs is determined; the predicting the time interval to which the default time of the first customer belongs by using the second model includes: determining from the plurality of second sub-models the relationship between the first customer and the first customer The second sub-model corresponding to the customer type; the time interval to which the default time of the first customer belongs is predicted by the second sub-model.

With reference to the first aspect, in a certain possible implementation manner of the first aspect, each third model includes multiple third sub-models, and customers corresponding to any two third sub-models in the multiple third sub-models different types, the customer type is determined according to the age group or region to which the customer belongs; the time interval to which the default time of the first customer predicted based on the second model belongs, the corresponding third model is used to predict the first customer The default time of a customer includes: in a third model corresponding to the time interval to which the default time of the first customer predicted by the second model belongs, determining a third model corresponding to the customer type of the first customer sub-model; predicting the default time of the first customer through the third sub-model.

With reference to the first aspect, in a certain possible implementation manner of the first aspect, the method further includes: acquiring a training set, where the training set includes historical data of multiple customers; A model, the second model and the at least one third model are trained.

With reference to the first aspect, in a possible implementation manner of the first aspect, the method further includes: grouping the training set based on the customer types corresponding to the multiple customers respectively, to obtain multiple sets of training sets, where the The customer types corresponding to the multiple sets of training sets are different, and the customer types are determined according to the age group or region to which the customer belongs; and, based on the training set, the first model, the second model and the Training the at least one third model includes: training the first model, the second model and the at least one third model based on each group of training sets, respectively, to obtain a trained first sub-model , a second sub-model, and multiple trained third sub-models.

With reference to the first aspect, in a certain possible implementation manner of the first aspect, the method further includes: updating the training set according to a preset period to obtain an updated training set; based on the updated training set, The first model, the second model and the at least one third model are trained separately.

In a second aspect, the present application provides a model training method, which can be executed by a server. The server is configured with a first model, a second model and at least one third model, the first model is used to predict the default probability of the customer, the second model is used to predict the time interval to which the customer's default time belongs, the At least one third model corresponds to at least one time interval, and each third model is used to predict the number of defaults and the default time of the customer within the corresponding time interval.

Exemplarily, the method includes: acquiring data of a plurality of customers, and the data of each customer includes a parameter used to reflect the credit risk of the customer; inputting the data of the plurality of customers into the first model, and obtaining the obtained data. the default probability of the multiple customers; when the default probability of the customer is greater than the preset value, the second model is used to predict the time interval to which the default time of the customer belongs; the default time of the customer predicted based on the second model belongs to , and predict the default time of customers through the corresponding third model.

Based on the above technical solution, the obtained data of multiple customers is input into the first model to obtain the default probability of multiple customers. For customers whose default probability is greater than the preset value, the second model is used to predict the default time of the customer. The time interval, that is, how long the customer may default in the future, and further based on the time interval to which the predicted default time of the customer belongs, the specific default time of the customer is predicted through the corresponding third model. The first model, the second model and the at least one third model are trained on the data of each customer, which is beneficial to improve the accuracy of the model.

In a third aspect, the present application provides a server, the server is configured with a first model, a second model and at least one third model, the first model is used to predict the default probability of a customer, and the second model uses The at least one third model corresponds to the at least one time interval in the time interval to which the default time of the customer is predicted, and each third model is used to predict the number of default times and the default time of the customer in the corresponding time interval.

Exemplarily, the server includes an acquisition unit, an input unit and a processing unit. Wherein, the acquisition unit is used for acquiring the data of the first customer, the data of the first customer includes parameters used to reflect the credit risk of the first customer; the input unit is used for inputting the data into the first model , obtain the default probability of the first customer; the processing unit is configured to predict the default time of the first customer through the second model when the default probability of the first customer is greater than a preset value. time interval; the processing unit is further configured to predict the default time of the first customer by using the corresponding third model based on the time interval to which the default time of the first customer predicted by the second model belongs.

In a fourth aspect, the present application provides a server, the apparatus including a processor. The processor is coupled to the memory and is operable to execute a computer program in the memory to implement the methods described in the first to second aspects and any possible implementations of the first to second aspects.

Optionally, the server in the fourth aspect further includes a memory.

Optionally, the server in the fourth aspect further includes a communication interface, and the processor is coupled to the communication interface.

In a fifth aspect, the present application provides a chip system, where the chip system includes at least one processor for supporting the implementation of any one of the above-mentioned first to second aspects and any possible implementation manners of the first to second aspects The functions involved, for example, receiving or processing the data involved in the above methods, etc.

In a possible design, the chip system further includes a memory for storing program instructions and data, and the memory is located inside the processor or outside the processor.

The chip system can be composed of chips, and can also include chips and other discrete devices.

In a sixth aspect, the present application provides a computer-readable storage medium on which a computer program (also referred to as code, or instruction) is stored, and when the computer program is executed by a processor, causes the The method described in any one of possible implementations of the first aspect to the second aspect and the first aspect to the second aspect is performed.

In a seventh aspect, the present application provides a computer program product, the computer program product comprising: a computer program (also referred to as code, or instructions), when the computer program is executed, the above-mentioned first aspect to the first The method described in the second aspect and any one of the possible implementations of the first aspect to the second aspect is performed.

It should be understood that the third to seventh aspects of the present application correspond to the technical solutions of the first and second aspects of the present application, and the beneficial effects obtained by each aspect and the corresponding feasible implementation manner are similar, and will not be repeated.

It should also be understood that the method for predicting the default time and the related device provided by the present application can be applied to the field of artificial intelligence, and can also be applied to other fields. This application does not limit this.

Description of drawings

1 is a schematic diagram of an application scenario applicable to the method provided by the embodiment of the present application;

2 is a schematic flowchart of a method for predicting a default time provided by an embodiment of the present application;

3 is another schematic flowchart of the method for predicting default time provided by an embodiment of the present application;

4 is a schematic block diagram of a server provided by an embodiment of the present application;

FIG. 5 is another schematic block diagram of a server provided by an embodiment of the present application.

Detailed ways

The following detailed description of the embodiments of the application provided in the accompanying drawings is not intended to limit the scope of the application as claimed, but is merely representative of selected embodiments of the application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

In order to facilitate understanding of the method for predicting the default time provided by the embodiment of the present application, the following describes the application scenarios applicable to the embodiment of the present application. It is understandable that the application scenarios described in the embodiments of the present application are for the purpose of illustrating the technical solutions of the embodiments of the present application more clearly, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application.

FIG. 1 is a schematic diagram of an application scenario applicable to the method provided by the embodiment of the present application. As shown in FIG. 1 , the technician may input relevant data for credit risk assessment of the customer through the electronic device 110 . The electronic device 110 may be connected in communication with the server 120 , and the server 120 may present a user interface through the electronic device 110 . The user interface provides an interface for the technician to interact with the server 120, and the technician can send data or information to the server 120 by inputting or selecting operations on the user interface. Accordingly, the server 120 may present the customer's credit risk assessment result through the electronic device 110 based on the data or information input by the technician.

It should be understood that the scenario shown in FIG. 1 is only an example, and the server 120 may be one physical device, or may be a server cluster composed of multiple physical devices.

With the rapid development of my country's economy and finance, consumer credit is gradually accepted by people. In recent years, various credit businesses such as auto loans, education loans, petty cash loans, and beauty loans have flourished. For credit business, financial institutions' credit risk assessment of customers is crucial. In the era of big data, although the data used for credit risk assessment is becoming more and more abundant, it also brings many challenges to credit risk assessment.

The current mainstream credit risk assessment method is to use statistical models to predict whether a customer will default or calculate the customer's default probability, but simply predicting the customer's default probability is not comprehensive enough for financial institutions' risk control, and the efficiency of risk control is not sufficient. high.

In order to solve the above-mentioned problems, the present application provides a method for predicting the default time. The obtained data of the first customer is input into the first model to obtain the default probability of the customer. When the default probability is greater than the preset value, In this case, the second model is used to predict the time interval to which the customer's default time belongs, that is, how long the customer may default in the future. Each time interval corresponds to a third model that predicts the customer's default time, and is further based on the prediction. The customer's default time belongs to the time interval, and the specific default time of the customer is predicted through the corresponding third model, so that the financial institution can formulate more reasonable risk control measures based on the default time.

The technical solutions of the present application and how the technical solutions of the present application solve the above-mentioned technical problems will be described in detail below with specific examples. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.

FIG. 2 is a schematic flowchart of a method 200 for predicting a default time provided by an embodiment of the present application. The method 200 for predicting the default time shown in FIG. 2 may include step 210 and step 240 . Each step in the method 200 is described in detail below.

It should be understood that the method 200 shown in FIG. 2 uses the server as the execution body, but the execution body of the method should not constitute any limitation. As long as the program that records the code of the method provided by the present application can be executed, the implementation of the present application can be executed. method provided by the example. For example, the server can also be replaced with components (eg, chips, chip systems, etc.) configured in the server, or other functional modules capable of calling programs and executing programs.

It should also be understood that the above-mentioned server is configured with a first model, a second model and at least one third model, wherein the first model is used to predict the default probability of the customer, and the second model is used to predict the time interval to which the customer's default time belongs, That is, it is used to predict how long a customer may be in default, at least one third model corresponds to at least one time interval, and each third model is used to predict the number of defaults and the default time of the customer in the corresponding time interval. .

Step 210: Acquire data of the first customer, where the data of the first customer includes parameters used to reflect the credit risk of the first customer.

The first customer may be an incremental customer, that is, a new customer, or an existing customer, that is, an old customer, which is not limited in this embodiment of the present application. The historical data of old customers can be used to train the first model, the second model and at least one third model.

A possible implementation is that the server obtains the data of the first client in response to the input operation of the technician, that is, the technician can input the data reflecting the credit risk of the first client through the user interface, and accordingly, the server The data of the first customer is acquired to predict the default time of the first customer.

Another possible implementation is that the server can obtain the data of the first customer from its own platform or the partner platform, wherein the own platform or the partner platform stores the data of the first customer, and the own platform or the partner platform stores the data of the first customer. Can communicate with the server. The server can trigger the server to obtain the data of the first customer from the above-mentioned platform in response to the operation such as the click of the technical staff, that is, the technical staff can click to predict the default time of the first customer through the user interface, and then trigger the server to obtain the data of the first customer from the above-mentioned platform. A customer's data.

Optionally, the above data includes one or more of the following: industry category, execution interest rate, loan amount, number of loan periods, gender, age, education, annual household income, employment situation, unit type, residence situation, job title, social security sign and customer level.

Among them, the customer level can be used to reflect the importance of the customer. For example, the higher the customer rank, the more important the customer is.

In one example, the server may, in response to the input operation of the technician, obtain the industry category, execution interest rate, loan amount, loan period, gender, age, education level, annual family income, employment status, unit type, residence status, One or more items of job title, social security mark, and customer level, so that the server can predict the default probability of the first customer based on various influences.

For another example, the server may obtain the industry category, execution interest rate, loan amount, loan period, gender, age, education level, annual family income, employment status, unit type, and residence status of the first customer from its own platform or a partner platform. One or more of , job title, social security logo, and customer level, where the above-mentioned data of the first customer is stored in the self-owned platform or the partner platform, the self-owned platform or the cooperative platform can communicate with the server, and the server can respond The server is triggered to acquire the above-mentioned data of the first client from the above-mentioned platform due to the operation such as the click of the technician.

It should be understood that in this application, the collection, storage, use, processing, transmission, provision and disclosure of financial data or user personal data and other information involved are in compliance with relevant laws and regulations, and do not violate public order and good customs.

By providing the above-mentioned various data that may affect the default situation of the customer, the embodiment of the present application comprehensively considers the impact of various factors on the default situation of the customer, which is beneficial to improve the accuracy of the predicted default probability of the customer.

Optionally, the above-mentioned first model is a Logistic model, the second model is a GMM, and the third model is an ARIMA model.

The following will introduce the Logistic model, GMM and ARIMA model in detail.

1. Logistic model

其中，

表示客户i的违约概率，

表示客户i的数据，

是需要训练的参数。The logistic model has a form similar to the following:

in,

represents the default probability of customer i,

represents the data of customer i,

are the parameters to be trained.

表示客户i发生违约，不应对本申请实施例构成任何限定。例如，

还可以表示客户i不发生违约。相应地，当客户不违约的概率大于预设值时，表示客户的信用良好，可以认为该客户在贷款期内守约，换言之，服务器无需进一步预测该客户的违约时间；当客户不违约的概率小于或等于预设值时，表示客户的信用一般，可以认为该客户在贷款期内将会违约(可以称其为高风险客户)，换言之，服务器需要进一步预测该客户的违约时间。It is understood that the above

It means that the customer i has breached the contract, which shall not constitute any limitation to the embodiments of this application. E.g,

It can also mean that customer i does not default. Correspondingly, when the probability of the customer not defaulting is greater than the preset value, it means that the customer's credit is good, and the customer can be considered to be in compliance during the loan period. In other words, the server does not need to further predict the default time of the customer; when the probability of the customer not defaulting When it is less than or equal to the preset value, it indicates that the customer's credit is average, and it can be considered that the customer will default within the loan period (it can be called a high-risk customer). In other words, the server needs to further predict the customer's default time.

2. GMM

其中，N(x|μ,σ)表示概率，σ表示标准差，μ表示均值，即期望，x表示客户的数据，如家庭收入、或年龄等。上述公式表示在μ附近的概率。可以理解，距离μ越近，即σ越小，其概率越大。The 1D Gaussian distribution has a form similar to the following:

Among them, N(x|μ,σ) represents the probability, σ represents the standard deviation, μ represents the mean, that is, the expectation, and x represents the customer's data, such as family income, or age. The above formula expresses the probability around μ. It can be understood that the closer the distance μ, that is, the smaller the σ, the greater the probability.

The form of d (d is a positive integer greater than 1) dimensional Gaussian distribution is similar to the following:

其中，d表示x的维数，∑表示d*d的协方差矩阵，|∑|为协方差的行列式的值，μ表示均值，即期望，x表示客户的数据，如家庭收入、年龄等。

Among them, d represents the dimension of x, ∑ represents the covariance matrix of d*d, |∑| is the value of the determinant of the covariance, μ represents the mean, that is, the expectation, and x represents the customer's data, such as family income, age, etc. .

其中，p(x|C_j)＝N(x|μ_j,∑_j)是类别或分组j(服从高斯分布)的条件概率密度，p(C_j)≥0，p(C_j)是类别j的权重参数(w_j＝p(C_j))，并且

模型的参数为p(C_j)，μ_j，∑_j，其中j＝1,…,k，k是类别或分组的总数(总共有k个类别或分组)，k为正整数。μ_j是类别j的均值，∑_j是类别j的协方差。The Gaussian mixture model is to mix multiple Gaussian models together, and use the weight parameter to adjust the mixture ratio of different Gaussian models (representing the categories in the data sample). The form of the Gaussian mixture model is similar to the following:

Among them, p(x|C _j )=N(x|μ _j ,∑ _j ) is the conditional probability density of category or group j (obey Gaussian distribution), p(C _j )≥0, p(C _j ) is the category weight parameter for j (w _j =p(C _j )), and

The parameters of the model are p(C _j ), μ _j , Σ _j , where j=1,...,k, k is the total number of categories or groups (there are k categories or groups in total), and k is a positive integer. μ _j is the mean of category j and ∑ _j is the covariance of category j.

In the embodiment of the present application, the category represents the time interval of the customer's default time, that is, how long in the future the customer will default. For example, several categories are: default within one year, default within two years, default within three years, and default within four years. Based on the above Gaussian mixture model, the posterior probability of the first customer can be calculated and divided into one of the Gaussian models, that is, the category to which the first customer belongs, that is, the default time of the first customer belongs to. time interval.

It can be understood that before using the above-mentioned Gaussian mixture model, parameter estimation needs to be performed to obtain a Gaussian mixture model for predicting the category to which the first customer belongs. For example, an expectation maximization (expectation maximization, EM) algorithm can be used to perform parameter estimation on the Gaussian mixture model, to obtain a Gaussian mixture model for predicting the category to which the first customer belongs. The process of parameter estimation is described in detail below.

其中，p(x|C_j)＝N(x|μ_j,Σ_j)。两边取对数，获得GMM的对数似然函数：

目标也就是使该对数似然函数最大化，故使用EM算法。EM算法包括初始化步骤和迭代步骤。初始化步骤为：初始化K个集群：C₁,…,C_k，对于每个集群j，都有参数(μ_j,∑_j)和p(C_j)。迭代步骤为：估计每个数据点的所属集群p(C_j|x_j)(期望步骤)，计算似然函数的期望；重新估计每个集群j的参数(μ_j,∑_j)和p(C_j)(最大化步骤)。The goal of fitting a GMM is to find p(C _j ), μ _j , Σ _j such that maximizing

Wherein, p(x|C _j )=N(x|μ _j ,Σ _j ). Take the logarithm of both sides to obtain the log-likelihood function of GMM:

The goal is to maximize the log-likelihood function, so the EM algorithm is used. The EM algorithm includes an initialization step and an iterative step. The initialization steps are: Initialize K clusters: C ₁ , . . . , C _k , for each cluster j, there are parameters (μ _j , ∑ _j ) and p(C _j ). The iterative steps are: estimate the cluster p(C _j |x _j ) to which each data point belongs (expectation step), calculate the expectation of the likelihood function; re-estimate the parameters (μ _j ,∑ _j ) and p( C _j ) (maximization step).

The specific process of the EM algorithm:

Step 1: Let z ₁ ,...,z _n denote the true source (ie category) of the corresponding data x ₁ ,...,x _n . Each _zi is a discrete variable with values between j=1,...,k, where k is the number of categories. has the following log-likelihood function:

来代替。Step 2: Use θ to represent the model parameters, and use logp(X, z|θ) as its expectation

instead.

Step 3: The distribution of logp(X,z|θ ^(t) ) of the current parameter estimation needs to be given, according to the Bayes rule, it satisfies

Step 4: p(z _i =c| _xi ,θ ^(t) ) is called the “responsibility” of cluster c for data point i, r _ic =p(z _i =c| _xi ,θ ^(t) ) .

Step 5: Desired step of GMM, denote x ₁ ,...,x _n as X,

其中，

Step 6: The maximization step of GMM, take Q(θ,θ ^(t) ) for each parameter and take its corresponding partial derivatives, and set these partial derivatives to zero, so as to obtain new parameter estimates

in,

It can be understood that for a more specific description of the GMM and EM algorithms, reference may be made to known technologies, and details are not repeated here.

3. ARIMA model

其中，y_t是当前值，μ是常数项，p是阶数，γ_i是自相关系数，e_t是误差。移动平均模型关注的是自回归模型中的误差项的累加，公式定义如下：

自回归模型和移动平均模型相结合，就得到了自回归移动平均模型ARMA(p,q)，计算公式如下：

将自回归模型、移动平均模型和差分法结合，就得到了ARIMA(p,d,q)，其中d是需要对数据进行差分的阶数。The autoregressive model first needs to determine an order p, which means that the historical value of several periods is used to predict the current value. The formula of the p-order autoregressive model is defined as:

where y _t is the current value, μ is a constant term, p is the order, γ _i is the autocorrelation coefficient, and e _t is the error. The moving average model is concerned with the accumulation of error terms in the autoregressive model, and the formula is defined as follows:

Combining the autoregressive model and the moving average model, the autoregressive moving average model ARMA(p,q) is obtained. The calculation formula is as follows:

Combining the autoregressive model, the moving average model, and the differencing method yields ARIMA(p,d,q), where d is the order at which the data needs to be differenciated.

It should be understood that the above descriptions of the Logistic model, the GMM and the ARIMA model are only for the purpose of making the methods provided by the embodiments of the present application more clearly, and should not constitute any limitation to the embodiments of the present application. For a more detailed introduction, reference may be made to known technologies. It will not be repeated here.

In the embodiment of the present application, the first model uses the Logistic model to predict the default probability of customers, which is beneficial to preliminarily screen out some high-risk customers, that is, customers with a higher default probability, and further facilitates financial institutions to focus on evaluating such customers. Possibly reduce the losses of financial institutions. The second model adopts GMM, which can be used to identify more complex distributions. By using GMM, it can comprehensively consider the influence of customers' default situation by industry, income, employment situation, etc., which is beneficial to improve the predicted default time of customers. The accuracy of the time interval to which it belongs. The third model uses the ARIMA model to predict the customer's default time, which can improve the accuracy of the prediction.

Step 220: Input the above data into the first model to obtain the default probability of the first customer.

After acquiring the data of the first customer, the server inputs the data of the first customer into the first model to obtain the default probability of the first customer.

Exemplarily, the server inputs the industry category, execution interest rate, loan amount, loan period, gender, age, education, annual family income, employment status, unit type, residence status, position, social security mark and customer level of the first customer. In the Logistic model, the default probability of the first customer is obtained.

It should be understood that the above description takes obtaining the default probability of the first customer as an example, but should not constitute any limitation to the embodiments of the present application. For example, the server may also calculate the probability that the first customer does not default based on the logistic model. Correspondingly, when the probability of the customer not defaulting is greater than the preset value, it means that the customer's credit is good, and the customer can be considered to be in compliance within the loan period. In other words, the server does not need to further predict the default time of the customer; when the probability of the customer not defaulting When it is less than or equal to the preset value, it means that the customer's credit is average, and it can be considered that the customer will default within the loan period (it can be called a high-risk customer). In other words, the server needs to further predict the customer's default time.

Step 230 , when the default probability of the first customer is greater than the preset value, predict the time interval to which the default time of the first customer belongs by using the second model.

After the server obtains the probability of default of the first customer, if the probability of default of the first customer is less than or equal to the preset value, it can be considered that the credit of the first customer is good, that is, the first customer keeps the contract during the loan period, in other words , the server does not need to further predict the default time of the customer.

When the default probability of the first customer is greater than the preset value, the server uses the second model to predict the time interval to which the default time of the first customer belongs, that is, how long in the future the first customer may default.

Exemplarily, when the default probability of the first customer is greater than the preset value, the server predicts the time interval to which the default time of the first customer belongs by using GMM. Breach occurs.

Step 240 , based on the time interval to which the default time of the first customer predicted by the second model belongs, predict the default time of the first customer through the corresponding third model.

Each time interval corresponds to a third model. For example, a customer may default within one year, and it corresponds to a third model, which is used to predict the specific time when the customer defaults in one year, such as the tenth month of one year. Breach of contract.

After predicting the time interval to which the default time of the first customer belongs, the server predicts the default time of the first customer through the corresponding third model.

Exemplarily, the server predicts through the GMM that the first customer may default within one year after the loan, and further, through its corresponding ARIMA model, predicts that the first customer may default in the tenth month after the loan, In turn, it is convenient for financial institutions to formulate more reasonable risk control measures.

It should be understood that different types of customers have different economic levels, and their probability of default is relatively high. Therefore, the server can select the corresponding first sub-model, second sub-model and third sub-model according to the customer type to which the first customer belongs. submodel. The following will describe in detail how the server predicts the default time of the customer after the customer is classified according to the customer type.

Optionally, the first model includes a plurality of first sub-models, the plurality of first sub-models correspond to different customer types, and the customer type is determined according to the age group or region to which the customer belongs; data is input into the first model, and the first sub-model is obtained. The default probability of a customer includes: determining a first sub-model corresponding to the customer type of the first customer from a plurality of first sub-models; inputting the data of the first customer into the first sub-model to obtain the first customer’s Default probability.

Exemplarily, the first model includes a first sub-model 1 and a first sub-model 2, the customer type corresponding to the first sub-model 1 is a southern customer, the customer type corresponding to the first sub-model 2 is a northern customer, and the server is based on the first sub-model 2. The type of the customer determines which first sub-model to input its data into. If the first customer belongs to the southern customer, the server inputs the data of the first customer into the first sub-model 1 to obtain the default probability of the first customer.

Optionally, the second model includes a plurality of second sub-models, the plurality of second sub-models correspond to different customer types, and the customer type is determined according to the age group or region to which the customer belongs; the second model is used to predict the default time of the first customer. The time interval to which it belongs includes: determining a second sub-model corresponding to the customer type of the first customer from a plurality of second sub-models; and predicting the time interval to which the default time of the first customer belongs by using the second sub-model.

Exemplarily, the second model includes a second sub-model 1 and a second sub-model 2, the client type corresponding to the second sub-model 1 is a southern client, the client type corresponding to the second sub-model 2 is a northern client, and the server is based on the first The type of the customer to which it belongs determines which second sub-model to input its data into. For example, if the first customer belongs to the southern customer, if the default probability of the first customer predicted based on the first sub-model 1 is greater than the preset value, the server will The data of the first customer is input into the second sub-model 1 to predict the time interval to which the default time of the first customer belongs.

Optionally, each third model includes a plurality of third sub-models, and any two third sub-models in the plurality of third sub-models correspond to different customer types, and the customer types are determined according to the age group or region to which the customer belongs; Based on the time interval to which the default time of the first customer predicted by the second model belongs, predicting the default time of the first customer by using the corresponding third model, including: in the time interval to which the default time of the first customer predicted by the second model belongs In the corresponding third model, a third sub-model corresponding to the customer type of the first customer is determined; the default time of the first customer is predicted through the third sub-model.

Exemplarily, the first customer belongs to the southern customer, and the server predicts the time interval to which the default time of the first customer belongs based on the second sub-model 1, and different time intervals correspond to different third sub-models, This year corresponds to the third sub-model 1, two years after the loan corresponds to the third sub-model 2, and three years after the loan corresponds to the third sub-model 3. Assuming that the first customer's default time falls within one year after the loan, the server will pass The third sub-model 1 predicts the default time of the first customer, for example, the first customer may default in the tenth month after the loan.

Optionally, before acquiring the data of the first customer, the method shown in FIG. 2 further includes: acquiring a training set, where the training set includes historical data of multiple customers; based on the training set, the first model, the second model and the At least one third model is trained.

Exemplarily, the server may obtain the historical data of multiple clients, and divide the historical data of the multiple clients into training data and verification data, that is, a part of the client's data is used for training the model, and a part of the client's data is used for the verification model, to obtain the trained first model, second model and at least one third model.

In this embodiment of the present application, the first model, the second model, and the at least one third model are respectively trained through the training set, which is beneficial to improve the accuracy of the first model, the second model, and the at least one third model, and further helps to improve the prediction the accuracy of the customer's default time.

Optionally, the method shown in FIG. 2 also includes: grouping the training set based on the customer types corresponding to the multiple customers respectively, to obtain multiple sets of training sets, the customer types corresponding to the multiple sets of training sets are different, and the customer types are based on the customer’s belonging. The age group or region of the A third model is trained to obtain a trained first sub-model, a second sub-model and a plurality of trained third sub-models.

The server can group multiple customers according to the region where the customer is located or the age group the customer belongs to, that is, the training set is divided into multiple sets of training sets, each set of training sets corresponds to a type of customer, and based on each group The training set trains the first model, the second model and the at least one third model for training to obtain a trained first sub-model, a second sub-model and a plurality of trained third sub-models.

Exemplarily, the server obtains data of 100 customers, of which 40 customers belong to the south and 60 customers belong to the north. Due to the difference in economic level between the north and the south, the 100 customers are divided into two groups, for example, 40 customers are used. For data belonging to customers in the south, the first model, the second model and at least one third model are respectively trained to obtain a trained first sub-model, a second sub-model and a plurality of trained third sub-models; And, use the data of 60 customers belonging to the north to train the first model, the second model and at least one third model respectively, and obtain a trained first sub-model, second sub-model and a plurality of trained sub-models. The third submodel. In this way, two first sub-models, two second sub-models, and several third sub-models (the number of third sub-models is twice the number of third models) can be obtained. For a new client, the server may use its corresponding first sub-model, second sub-model and third sub-model to make predictions based on the type of the client.

In the embodiment of the present application, classifying customers according to their types is beneficial to reduce the influence of economic differences of different types of customers on predicting the default probability of customers.

Optionally, the method shown in FIG. 2 further includes: updating the training set according to a preset period to obtain an updated training set; The model is trained.

In other words, the server may periodically update the training set, and based on the updated training set, train the first model, the second model and the at least one third model, respectively. For example, the server may periodically acquire historical data of different clients, and train the first model, the second model and the at least one third model based on the acquired data.

In the embodiment of the present application, by periodically updating the training set and training the model based on the updated training set, that is, training the model multiple times, it is beneficial to improve the accuracy of the model, thereby improving the accuracy of the predicted customer default time.

FIG. 3 is another schematic flowchart of a method for predicting a default time provided by an embodiment of the present application.

As shown in FIG. 3, in step 310, the server is started.

Step 320, the server maintains or updates the model and configures model parameters. For example, the server maintains or updates the first model, the second model, and at least one third model, and configures the parameters of the above models.

Step 330, the server enables the model of client risk monitoring. For example, the server enables the first model described above.

In step 340, the server classifies and clusters the clients. Exemplarily, the server determines whether the customer is a high-risk customer or a compliance customer based on the first model. For example, the server predicts the customer's default probability based on the first model. If the customer's default probability is greater than a preset value, the server considers the customer to be a high-risk customer, and needs to further predict the time interval and specific default time to which the default time belongs. For a detailed process, reference may be made to the related description of FIG. 2 , which will not be repeated here.

In step 350, the server displays the client classification and clustering results. In this way, the technical staff can be prompted which high-risk customers are for the business staff to focus on reviewing. For example, high-risk customers can further predict their default time.

In step 360, the server performs calculation based on the historical records, and displays the model monitoring effect. In other words, the server inputs the customer's historical data into the model, determines the default time, and judges whether the customer's default time is accurate. For example, the customer's actual default time is the same as the default time calculated based on the model, indicating that the model monitoring effect is better.

Understandably, based on the model monitoring effect, the developer can instruct the server to adjust the model to improve the accuracy of the model.

Based on the above technical solution, the server inputs the acquired data of the first customer into the first model to obtain the default probability of the customer, and when the default probability is greater than the preset value, predicts the customer's default probability through the second model The time interval to which the default time belongs, that is, how long the customer may default in the future. Further, based on the time interval to which the predicted default time of the customer belongs, the specific default time of the customer is predicted through the corresponding third model. It can not only predict the default probability of customers, but also further predict the default time for customers who may default, so more comprehensive information can be obtained, which is convenient for financial institutions to formulate more reasonable risk control measures based on this, which is conducive to improving financial The efficiency of an institution's credit risk control.

Optionally, the embodiment of the present application further provides a model training method, and the method can be executed by a server. The server is configured with a first model, a second model and at least one third model, the first model is used to predict the default probability of the customer, the second model is used to predict the time interval to which the customer's default time belongs, at least one third model and At least one time interval corresponds, and each third model is used to predict the number of defaults and the default time of the customer within the corresponding time interval.

Exemplarily, the method includes: acquiring data of a plurality of customers, where the data of each customer includes a parameter used to reflect the credit risk of the customer; inputting the data of the plurality of customers into the first model to obtain the default of the plurality of customers. Probability; when the customer's default probability is greater than the preset value, the second model is used to predict the time interval to which the customer's default time belongs; the time interval to which the customer's default time is predicted based on the second model is determined through the corresponding first The three-model predicts the customer's default time.

For the specific process of training the first model, the second model and the at least one third model, reference may be made to the relevant description of the embodiment shown in FIG. 2 , which will not be repeated here.

Based on the above technical solution, the obtained data of multiple customers is input into the first model to obtain the default probability of multiple customers. For customers whose default probability is greater than the preset value, the second model is used to predict the default time of the customer. The time interval, that is, how long the customer may default in the future, and further based on the time interval to which the predicted default time of the customer belongs, the specific default time of the customer is predicted through the corresponding third model. The first model, the second model and the at least one third model are trained on the data of each customer, which is beneficial to improve the accuracy of the model.

FIG. 4 is a schematic block diagram of a server provided by an embodiment of the present application.

As shown in FIG. 4 , the apparatus 400 may include: an acquisition unit 410 , an input unit 420 and a processing unit 430 . The server 400 can be used to implement the method described in the embodiment shown in FIG. 2 or FIG. 3 .

Exemplarily, when the apparatus 400 is used to implement the method described in the embodiment shown in FIG. 2 , the obtaining unit 410 is used to obtain the data of the first customer, and the data of the first customer includes the credit used to reflect the first customer. The parameter of risk; the input unit 420 is used to input the data into the first model to obtain the default probability of the first customer; the processing unit 430 is used to pass the second model when the default probability of the first customer is greater than the preset value Predict the time interval to which the default time of the first customer belongs; the processing unit 430 is further configured to predict the default time of the first customer through the corresponding third model based on the time interval to which the default time of the first customer belongs to predicted by the second model.

Optionally, the first model is a Logistic model, the second model is a GMM, and the third model is an ARIMA model.

Optionally, the above data includes one or more of the following: industry category, execution interest rate, loan amount, number of loan periods, gender, age, education, annual household income, employment situation, unit type, residence situation, job title, social security sign and customer level.

Optionally, the first model includes a plurality of first sub-models, and the plurality of first sub-models correspond to different customer types, and the customer type is determined according to the age group or region to which the customer belongs; A first sub-model corresponding to the customer type of the first customer is determined in the first sub-model; the data of the first customer is input into the first sub-model to obtain the default probability of the first customer.

Optionally, the second model includes a plurality of second sub-models, the plurality of second sub-models correspond to different customer types, and the customer type is determined according to the age group or region to which the customer belongs; the processing unit 430 is specifically configured to select from the plurality of second sub-models. A second sub-model corresponding to the customer type of the first customer is determined in the sub-model; the time interval to which the default time of the first customer belongs is predicted by the second sub-model.

Optionally, each third model includes a plurality of third sub-models, and any two third sub-models in the plurality of third sub-models correspond to different customer types, and the customer types are determined according to the age group or region to which the customer belongs; The processing unit 430 is specifically configured to determine a third sub-model corresponding to the customer type of the first customer in the third model corresponding to the time interval to which the default time of the first customer predicted by the second model belongs; through the third sub-model Predict the default time of the first customer.

Optionally, the processing unit 430 is further configured to acquire a training set, where the training set includes historical data of multiple customers; based on the training set, the first model, the second model and at least one third model are trained respectively.

Optionally, the processing unit 430 is further configured to group the training set based on the customer types corresponding to the multiple customers respectively, to obtain multiple sets of training sets, the customer types corresponding to the multiple sets of training sets are different, and the customer types are based on the age groups to which the customers belong. or region determination; and, the processing unit 430 is specifically configured to respectively train the first model, the second model and the at least one third model based on each group of training sets to obtain a trained first sub-model and a second sub-model and multiple trained third sub-models.

Optionally, the processing unit 430 is further configured to update the training set according to a preset period to obtain an updated training set; based on the updated training set, respectively train the first model, the second model and at least one third model .

It should be understood that the division of units in the embodiments of the present application is illustrative, and is only a logical function division, and there may be other division manners in actual implementation. In addition, each functional unit in each embodiment of the present application may be integrated into one processor, or may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware, or can be implemented in the form of software function modules.

Exemplarily, the server may include a data management unit, a model management unit, and a risk monitoring unit. Among them, the data management unit is used to store or update loan customer data, and the loan customer data is used for machine learning model training and verification. The model management unit is used to configure and update machine learning models for data analysis and risk analysis, such as logistic model, GMM and ARIMA model. The risk monitoring unit is used to apply machine learning models, perform cluster analysis and risk prediction on loan customer data, etc., and display credit risk, that is, the customer's default time and/or default probability.

The data management unit can also be specifically divided into a loan customer data module, a customer group data module and a forecast record data module.

Among them, the loan customer data module is used to store loan customer data. The data includes industry category, execution interest rate, loan amount, number of loan periods, gender, age, education, family annual income, employment situation, unit type, residence situation, position, social security mark, customer level, etc. The customer group data module is used to generate customer group data (that is, a collection of data of customers belonging to the same category) through cluster analysis. The forecast record data module is used to store the customer risk forecast result and record the data.

The model management unit can also be specifically divided into a model instance module, a model parameter module and a model update module.

The model instance module is used to configure machine learning models for classification and cluster analysis, including logistic models and GMMs. The Model Parameters module is used to configure and update the parameters of a machine learning model. For the parameters of the GMM, the Metropolis-Hastings algorithm is used to update the model parameters to achieve convergence. The model update module is used to update existing models, or delete old models, or add new models.

The risk monitoring unit can also be specifically divided into an enabling model module, a customer clustering result module and a model monitoring effect module.

Among them, the enable model module is used to enable or disable machine learning model processing risk monitoring. The customer clustering result module is used to use the model to classify and cluster the loan customers, and display the results. If it is determined to be a high-risk customer, it will prompt the default duration prediction result. For example, customer A will default within one year, with a probability of 83 %. The model monitoring effect module is used to synthesize the historical data of loan customers to display the model risk monitoring effect, such as the correct classification ratio, error ratio, recall rate and other indicators, for model developers to analyze and optimize the model.

FIG. 5 is another schematic block diagram of a server provided by an embodiment of the present application.

The server 500 can be used to implement the method described in the embodiment shown in FIG. 2 or FIG. 3 above. The server 500 may be a chip system. In this embodiment of the present application, the chip system may be composed of chips, or may include chips and other discrete devices.

As shown in FIG. 5 , the server 500 may include at least one processor 510 .

Exemplarily, the processor 510 may be configured to obtain data of the first customer, where the data of the first customer includes parameters used to reflect the credit risk of the first customer; input the data into the first model to obtain the default of the first customer; probability; when the default probability of the first customer is greater than the preset value, the second model is used to predict the time interval to which the default time of the first customer belongs; the time interval to which the default time of the first customer is predicted based on the second model belongs, The default time of the first customer is predicted by the corresponding third model. For details, refer to the detailed description in the method example, which is not repeated here.

The server 500 may also include at least one memory 520, which may be used to store program instructions and/or data. Memory 520 is coupled to processor 510 . The coupling in the embodiments of the present application is an indirect coupling or communication connection between devices, units or modules, which may be in electrical, mechanical or other forms, and is used for information exchange between devices, units or modules. Processor 510 may cooperate with memory 520 . Processor 510 may execute program instructions stored in memory 520 . At least one of the at least one memory may be included in the processor.

The server 500 may also include a communication interface 530 for communicating with other devices through a transmission medium, so that the server 500 can communicate with other devices. The communication interface 530 may be, for example, a transceiver, an interface, a bus, a circuit, or a device capable of implementing a transceiving function. The processor 510 can use the communication interface 530 to send and receive data and/or information, and be used to implement the method described in the embodiment shown in FIG. 2 or FIG. 3 .

The specific connection medium between the processor 510, the memory 520, and the communication interface 530 is not limited in the embodiments of the present application. In the embodiment of the present application, in FIG. 5 , the processor 510 , the memory 520 , and the communication interface 530 are connected through a bus 540 . The bus 540 is represented by a thick line in FIG. 5 , and the connection manners between other components are only for schematic illustration and are not intended to be limiting. The bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of presentation, only one thick line is used in FIG. 5, but it does not mean that there is only one bus or one type of bus.

The present application further provides a chip system, where the chip system includes at least one processor for implementing the method described in the embodiment shown in FIG. 2 or FIG. 3 above.

In a possible design, the chip system further includes a memory for storing program instructions and data, and the memory is located inside the processor or outside the processor.

The chip system can be composed of chips, and can also include chips and other discrete devices.

The present application also provides a computer program product, the computer program product includes: a computer program (also referred to as code, or instructions), when the computer program is run, the computer executes as shown in FIG. 2 or FIG. 3 . The method described in the examples.

The present application also provides a computer-readable storage medium storing a computer program (also referred to as code, or instructions). When the computer program is executed, the computer is caused to perform the method described in the embodiment shown in FIG. 2 or FIG. 3 .

It should be noted that the default time prediction method and related device provided in the embodiments of the present application can be applied to the field of artificial intelligence, and can also be applied to any field other than the field of artificial intelligence, which is not limited in this application.

It should be understood that the processor in this embodiment of the present application may be an integrated circuit chip, which has a signal processing capability. In the implementation process, each step of the above method embodiments may be completed by a hardware integrated logic circuit in a processor or an instruction in the form of software. The above-mentioned processor may be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable Logic devices, discrete gate or transistor logic devices, discrete hardware components. The methods, steps, and logic block diagrams disclosed in the embodiments of this application can be implemented or executed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in conjunction with the embodiments of the present application may be directly embodied as executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media mature in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware.

It should also be understood that the memory in the embodiments of the present application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically programmable Erase programmable read-only memory (electrically EPROM, EEPROM) or flash memory. Volatile memory may be random access memory (RAM), which acts as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, ESDRAM), synchronous link dynamic random access memory (synchlink DRAM, SLDRAM) And direct memory bus random access memory (directrambus RAM, DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but not be limited to, these and any other suitable types of memory.

The terms "unit", "module" and the like used in this specification may be used to refer to a computer-related entity, hardware, firmware, a combination of hardware and software, software, or software in execution.

Those of ordinary skill in the art will appreciate that the various illustrative logical blocks and steps described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware accomplish. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application. In the several embodiments provided in this application, it should be understood that the disclosed apparatuses, devices and methods may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or modules, and may be in electrical, mechanical or other forms.

The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, may be located in one place, or may be distributed to multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically alone, or two or more units may be integrated into one module.

In the above embodiments, the functions of each functional module may be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented in software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer may be a general purpose computer, special purpose computer, computer network, or other programmable device. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be downloaded from a website site, computer, server, or data center Transmission to another website site, computer, server, or data center by wire (eg, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (eg, infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that includes an integration of one or more available media. The available media may be magnetic media (eg, floppy disk, hard disk, magnetic tape), optical media (eg, digital video disc (DVD)), or semiconductor media (eg, solid state disk (SSD)) )Wait.

The functions, if implemented in the form of software functional units and sold or used as independent products, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a removable hard disk, a ROM, a RAM, a magnetic disk, or an optical disk and other mediums that can store program codes.

The above are only specific embodiments of the present application, but the protection scope of the present application is not limited to this. should be covered within the scope of protection of this application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims (13)

1. A method for predicting default time is applied to a server, the server is configured with a first model, a second model and at least one third model, the first model is used for predicting default probability of a customer, the second model is used for predicting a time interval to which default time of the customer belongs, the at least one third model corresponds to the at least one time interval, and each third model is used for predicting default times and default time of the customer in the corresponding time interval, the method comprises the following steps:

obtaining data of a first customer, the data of the first customer comprising a parameter for reflecting credit risk of the first customer;

inputting the data into the first model to obtain a default probability of the first customer;

under the condition that the default probability of the first customer is larger than a preset value, predicting a time interval to which default time of the first customer belongs through the second model;

and predicting the default time of the first customer through a corresponding third model based on the time interval to which the default time of the first customer predicted by the second model belongs.

2. The method of claim 1, wherein the first model is a multivariate logical Logistic model, the second model is a gaussian mixture model GMM, and the third model is an autoregressive moving average ARIMA model.

3. The method of claim 1, wherein the data comprises one or more of: industry category, interest rate of performance, amount of loan, number of loan terms, gender, age, academic history, family annual income, employment status, unit type, occupancy status, job title, social security label, and customer rating.

4. The method of claim 1, wherein the first model comprises a plurality of first sub-models corresponding to different customer types, the customer types being determined according to the age group or region to which the customer belongs;

said inputting said data into said first model to obtain a probability of breach by said first customer, comprising:

determining a first sub-model corresponding to a customer type of the first customer from the plurality of first sub-models;

and inputting the data of the first customer into the first submodel to obtain the default probability of the first customer.

5. The method of claim 1, wherein the second model comprises a plurality of second submodels corresponding to different customer types, the customer types being determined according to the age group or region to which the customer belongs;

the predicting, by the second model, a time interval to which the default time of the first customer belongs includes:

determining a second sub-model corresponding to the customer type of the first customer from the plurality of second sub-models;

and predicting a time interval to which the default time of the first customer belongs through the second submodel.

6. The method of claim 1, wherein each third model comprises a plurality of third submodels, any two third submodels in the plurality of third submodels correspond to different customer types, and the customer types are determined according to the age groups or regions to which the customers belong;

the predicting the default time of the first customer through a corresponding third model based on the time interval to which the default time of the first customer predicted by the second model belongs comprises:

determining a third sub-model corresponding to the customer type of the first customer in a third model corresponding to a time interval to which the default time of the first customer predicted by the second model belongs;

predicting, by the third submodel, a default time for the first customer.

7. The method of claim 1, wherein the method further comprises:

acquiring a training set, wherein the training set comprises historical data of a plurality of clients;

training the first model, the second model, and the at least one third model, respectively, based on the training set.

8. The method of claim 7, wherein the method further comprises:

grouping the training sets based on the client types respectively corresponding to the clients to obtain a plurality of groups of training sets, wherein the client types corresponding to the training sets are different, and the client types are determined according to the age groups or the regions of the clients; and the number of the first and second groups,

the training the first model, the second model, and the at least one third model, respectively, based on the training set, includes:

and training the first model, the second model and the at least one third model respectively based on each group of training sets to obtain a trained first sub-model, a trained second sub-model and a plurality of trained third sub-models.

9. The method of claim 7 or 8, wherein the method further comprises:

updating the training set according to a preset period to obtain an updated training set;

training the first model, the second model, and the at least one third model, respectively, based on the updated training set.

10. A server, wherein the server is configured with a first model, a second model and at least one third model, the first model is used for predicting default probability of a customer, the second model is used for predicting a time interval to which default time of the customer belongs, the at least one third model corresponds to at least one time interval, each third model is used for predicting number of times of default and default time of the customer in the corresponding time interval, the server comprises:

an acquisition unit for acquiring data of a first customer, the data of the first customer comprising a parameter for reflecting credit risk of the first customer;

the input unit is used for inputting the data into the first model to obtain the default probability of the first customer;

the processing unit is used for predicting a time interval to which default time of the first customer belongs through the second model under the condition that the default probability of the first customer is greater than a preset value;

the processing unit is further used for predicting the default time of the first customer through a corresponding third model based on the time interval to which the default time of the first customer predicted by the second model belongs.

11. A server comprising a processor and a memory communicatively coupled to the processor;

the memory stores computer-executable instructions;

the processor executes computer-executable instructions stored by the memory to implement the method of any of claims 1 to 9.

12. A computer-readable storage medium, comprising a computer program which, when run on a computer, causes the computer to perform the method of any one of claims 1 to 9.

13. A computer program product, comprising a computer program which, when executed, causes a computer to perform the method of any one of claims 1 to 9.

Images