CN116843340A

CN116843340A - Trusted evaluation model construction and application method and device and electronic equipment

Info

Publication number: CN116843340A
Application number: CN202310684979.8A
Authority: CN
Inventors: 王彩亮; 贺红亮
Original assignee: Sany Group Co Ltd
Current assignee: Sany Group Co Ltd
Priority date: 2023-06-09
Filing date: 2023-06-09
Publication date: 2023-10-03

Abstract

The application provides a method, a device and an electronic device for constructing and applying a trust evaluation model. Because the sample user portrait comprises a plurality of dimensional features, namely basic features, asset features, credit investigation features and behavior features, a high-precision and high-stability target credit investigation evaluation model can be obtained by carrying out model construction, training and evaluation based on the sample user portrait and the sample label, and credit risk evaluation is carried out on the target client based on the target credit investigation evaluation model, so that the accuracy of risk evaluation and credit investigation decision can be effectively improved.

Description

Trusted evaluation model construction and application method and device and electronic equipment

Technical Field

The application relates to the technical field of financial risk control, in particular to a trust evaluation model construction and application method and device and electronic equipment.

Background

Capital-intensive industries, such as the engineering machinery industry, require significant capital investment and credit support due to high product prices and long sales cycles, and in such market environments, banks or other financial institutions need to perform credit risk assessment (i.e., credit assessment) on an engineering machinery enterprise to make credit decisions, i.e., determine relevant conditions such as credit amount and interest rate, while providing support for financing or loaning to the engineering machinery enterprise.

Credit scoring is one of the important methods for credit risk assessment by financial institutions such as banks and credit card companies. The traditional credit scoring method is that a financial analyst carries out routine financial analysis based on experience, and credit risk assessment is carried out on credit records, personal assets, professional backgrounds and the like of clients by adopting a plurality of fixed scoring standards and rules, so as to obtain credit scoring used for representing the credit risk of the clients.

However, the traditional credit rating method depends on experience and judgment of a financial analyst, is easily subjectively influenced by the financial analyst, causes deviation of an evaluation result, has lower accuracy of credit risk evaluation on clients, and has lower accuracy of credit decision making.

Disclosure of Invention

Based on the defects and shortcomings of the prior art, the application provides a trust evaluation model construction and application method, a trust evaluation model construction and application device and electronic equipment, which can perform model training and evaluation based on a multi-dimensional sample user portrait and a sample label to obtain a high-precision stable target trust evaluation model, and perform credit risk evaluation on a target client based on the target trust evaluation model, so that the accuracy of credit risk evaluation and trust decision can be effectively improved.

According to a first aspect of an embodiment of the present application, there is provided a trust evaluation model construction method, including:

acquiring a sample data set; the sample dataset includes a sample user representation and a sample label; the sample user portraits comprise basic features, asset features, credit features and behavioral features of the target client; the sample label comprises comprehensive credit rating, purchasing behavior rating and/or credit limit of the target client;

constructing a preset credit evaluation model based on the sample user portrait;

training and evaluating the preset trust evaluation model based on the sample user portrait and the sample label to obtain a target trust evaluation model; the target credit evaluation model is used for carrying out credit risk evaluation on the target client based on the real-time user image to obtain real-time comprehensive credit scores, real-time purchase behavior scores and/or real-time credit limit of the target client.

According to a second aspect of the embodiment of the present application, there is provided a trust evaluation model application method, including:

acquiring a real-time user image of a target client; the real-time user image comprises basic characteristics, asset characteristics, credit investigation characteristics and behavior characteristics;

based on the real-time user portrait and a target credit trust evaluation model, performing credit risk evaluation on the target client to obtain a real-time evaluation result of the target client; the real-time evaluation result comprises a real-time comprehensive credit rating, a real-time purchasing behavior rating and/or a real-time credit limit.

According to a third aspect of the embodiment of the present application, there is provided a trust evaluation model construction apparatus, including:

the acquisition module is used for acquiring a sample data set; the sample dataset includes a sample user representation and a sample label; the sample user portraits comprise basic features, asset features, credit features and behavioral features of the target client; the sample label comprises comprehensive credit rating, purchasing behavior rating and/or credit limit of the target client;

the construction module is used for constructing a preset credit evaluation model based on the sample user portrait;

the training and evaluating module is used for training and evaluating the preset credit evaluation model based on the sample user portrait and the sample label to obtain a target credit evaluation model; the target credit evaluation model is used for carrying out credit risk evaluation on the target client based on the real-time user image to obtain real-time comprehensive credit scores, real-time purchase behavior scores and/or real-time credit limit of the target client.

According to a fourth aspect of the embodiment of the present application, there is provided a trust evaluation model application apparatus, including:

the acquisition module is used for acquiring real-time user images of target clients; the real-time user image comprises basic characteristics, asset characteristics, credit investigation characteristics and behavior characteristics;

the determining module is used for carrying out credit risk assessment on the target client based on the real-time user portrait and the target credit trust assessment model to obtain a real-time assessment result of the target client; the real-time evaluation result comprises a real-time comprehensive credit rating, a real-time purchasing behavior rating and/or a real-time credit limit.

According to a fifth aspect of an embodiment of the present application, there is provided an electronic device including a memory and a processor;

the memory is connected with the processor and is used for storing a computer program;

the processor is configured to implement the trust evaluation model construction method according to the first aspect or the trust evaluation model application method according to the second aspect by running a computer program stored in the memory.

According to a sixth aspect of the embodiment of the present application, there is provided a storage medium, on which a computer program is stored, the computer program implementing the trust evaluation model construction method according to the first aspect or the trust evaluation model application method according to the second aspect when being executed by a processor.

In the method, the device and the electronic equipment for constructing and applying the trust evaluation model, the preset trust evaluation model can be constructed by acquiring the sample data set based on the sample user portrait in the sample data set, and training and evaluating the preset trust evaluation model based on the sample user portrait and the sample tag to obtain the target trust evaluation model. Because the sample user portrait comprises a plurality of dimensions such as basic characteristics, asset characteristics, credit investigation characteristics, behavior characteristics and the like, a high-precision and high-stability target credit investigation model can be obtained by carrying out model construction, training and evaluation based on the sample user portrait and the sample label, and credit risk investigation is carried out on the target client based on the target credit investigation model, so that the accuracy of risk investigation and credit investigation decision can be effectively improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present application, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow chart of a trust evaluation model construction method according to an embodiment of the present application;

FIG. 2 is a schematic diagram of a basic credit line according to an embodiment of the present application;

FIG. 3 is a schematic flow chart of a trust evaluation model application method according to an embodiment of the present application;

fig. 4 is a schematic structural diagram of a trust evaluation model construction device according to an embodiment of the present application;

fig. 5 is a schematic structural diagram of an application device of a trust evaluation model according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

SUMMARY

As described in the background art, when credit risk assessment is performed on a client, the conventional credit scoring method relies on experience and judgment of a financial analyst, is easily subjectively affected by the financial analyst, and causes deviation of an assessment result, so that the accuracy of credit risk assessment on the client is low, and the accuracy of credit decision-making is low.

Based on the basis, the inventor finds out through further research that based on a multi-dimensional sample user portrait, namely a sample user portrait comprising basic features, asset features, credit investigation features and behavior features, a preset credit investigation evaluation model is constructed, and based on the sample user portrait and a sample label, the preset credit investigation evaluation model is trained and evaluated, so that a target credit investigation evaluation model with higher accuracy and reliability can be obtained, and a target client is subjected to credit investigation based on the target credit investigation evaluation model, so that the accuracy of credit investigation and credit investigation decision can be effectively improved, and the problem that the manual credit investigation is easily subjectively influenced, and the accuracy of credit investigation decision are lower is solved.

Based on the above conception, the embodiment of the present specification provides a trust evaluation model construction and application method, and the trust evaluation model construction and application method will be described in an exemplary manner with reference to the accompanying drawings.

Exemplary method

Referring to fig. 1, in an exemplary embodiment, a trust evaluation model construction method is provided and is applied to any device. As shown in fig. 1, the trust evaluation model construction method includes steps S101-S103:

S101: a sample dataset is acquired.

Wherein the sample dataset includes sample user portraits and sample labels.

Specifically, the sample user representation includes basic features, asset features, credit features and behavioral features of the target client.

The number of the characteristic items in the basic characteristics is at least one, the number of the characteristic items in the asset characteristics is at least one, the number of the characteristic items in the credit characteristics is at least one, and the number of the characteristic items in the behavior characteristics is at least one.

In addition, the target client may be an individual or a company. In the embodiment of the application, the technical scheme of the application is introduced by taking the target client as an enterprise-level client, namely the client at the end B as an example.

In short, the B-side client is an enterprise or organization. Illustratively, the B-side clients include national enterprises, stock market enterprises, stock system non-market enterprises, group all system enterprises, private enterprises, affiliated enterprises, individual enterprises, and the like.

S102: and constructing a preset trust evaluation model based on the sample user portrait.

The preset credit assessment model is used for carrying out credit risk assessment on the target client based on the sample user portrait to obtain comprehensive credit score, purchase behavior score and/or credit limit of the target client.

Specifically, based on feature items in the user portrait, positions of the feature items in a preset model are determined, and preset super parameters are configured into the preset model to obtain a corresponding preset trust evaluation model.

Wherein, since different data types and data amounts are applicable to different models, the preset model can be selected based on the data types and data amounts of the sample user portraits and the sample tags.

The pre-set model may be a common model, such as a logistic regression model, a decision tree model, a support vector machine model, and the like.

S103: training and evaluating a preset trust evaluation model based on the sample user portrait and the sample label to obtain a target trust evaluation model.

The target credit evaluation model is used for carrying out credit risk evaluation on the target client based on the real-time user portrait to obtain real-time comprehensive credit scores, real-time purchase behavior scores and/or real-time credit limits of the target client.

Training and evaluating the preset credit evaluation model based on the sample user portrait and the sample label, and determining the trained preset credit evaluation model as a target credit evaluation model if the trained preset credit evaluation model passes the evaluation.

In particular, the sample data set is divided into a training data set and a test data set. The training data set is used for training the model, and the test data set is used for verifying generalization capability and accuracy of the model. And then, carrying out repeated iterative training on the preset credit evaluation model based on the sample user portrait and the sample label in the training data set, and adjusting preset super parameters in the preset credit evaluation model to obtain the trained preset credit evaluation model. And then, based on the sample user portrait and the sample label in the test data set, evaluating the generalization capability and the accuracy of the trained preset credit evaluation model. And finally, determining the trained preset trust evaluation model after evaluation as a target trust evaluation model.

Correspondingly, if the trained preset credit evaluation model fails to pass the evaluation, performing iterative training on the trained preset credit evaluation model again by adopting the training data set, or re-acquiring a sample data set, and performing iterative training on the credit evaluation model for a plurality of times by adopting the re-acquired training data set in the sample data set.

Typically, the training data set is 70% -80% of the sample data set and the test data set is 20% -30% of the sample data set. For example, the training dataset may comprise 70% of the sample dataset, while the test dataset may comprise 30% of the sample dataset.

More specifically, when evaluating the trained preset trust evaluation model, firstly inputting a sample user portrait in the test data set into the trained preset trust evaluation model, comparing the output of the trained preset trust evaluation model with a corresponding sample label, calculating the indexes such as accuracy, recall rate F1 and the like of the authorization evaluation model, drawing the charts such as ROC curve, PR curve and the like of the authorization evaluation model, and comprehensively evaluating the performance of the trust evaluation model. Of course, the output of the trained preset credit assessment model is the comprehensive credit score, purchase behavior score and/or credit limit corresponding to the sample user image determined by the model.

In this embodiment, the obtained sample data set includes a sample user image and a sample tag, where the sample user image includes multiple dimensions of features, that is, basic features, asset features, credit features, and behavior features. Therefore, based on the multi-dimensional sample user image and the sample label, a target credit evaluation model with higher accuracy and stability can be obtained through model construction, training and evaluation, credit risk evaluation is carried out on a target client based on the target credit evaluation model, the accuracy of credit evaluation can be effectively improved, and the accuracy of credit decision making is further improved.

In some embodiments, when acquiring a sample data set, sample raw data is acquired first, and the sample raw data is preprocessed to obtain sample data. And then, extracting features of the sample data to obtain features to be selected, and screening the features to be selected to obtain target features. Finally, a sample user representation is generated based on the target features. Meanwhile, a sample label corresponding to the sample user image is acquired.

In particular, sample raw data is obtained from aspects including basic information data of the target customer, transaction data (including revenue status, consumption status, transaction amount, transaction frequency, etc.), social media data, and the like. Such data may be obtained from various sources, such as related applications, transaction records, and dynamics on social media platforms. Thus, the user portrait generated based on the data can better know the requirements and risks of the clients.

Specifically, preprocessing is performed on sample raw data obtained from multiple aspects, including data cleaning, deduplication, missing value filling (i.e., deficiency), outlier processing, and the like, to obtain sample data. The sample data comprises basic data, asset data, credit data and behavior data of the target client. Thus, the integrity and the accuracy of the sample data can be effectively ensured, and a high-quality sample data set can be obtained.

In addition, the pretreatment may further include normalization, and the like. In this way, the data of different types in the sample data can be compared and comprehensively analyzed.

In particular, feature extraction may be performed on sample data using machine learning and data mining techniques. It will be appreciated that these features may reflect the credit status and preferences of the target customer, etc. from a variety of angles.

For example, various feature extraction algorithms, such as principal component analysis PCA, linear discriminant analysis LDA decision tree, etc., may be employed to perform feature extraction on sample data. The PCA algorithm can convert high-dimensional sample data into low-dimensional feature vectors, so that classification and clustering can be performed better.

The sample data may be divided into static data and dynamic data, which are relatively static and dynamic. The period of change of static data is longer than that of dynamic data. In general, the static data includes asset data, credit information data and basic data, and the asset characteristics, the credit information characteristics and the basic characteristics can be extracted from the static data; the dynamic data includes behavior data, and behavior features can be extracted from the dynamic data.

Specifically, after the feature to be selected is extracted, the feature to be selected is analyzed, the feature to be selected is screened based on whether the feature to be selected has correlation with the credit condition of the target client or not and the magnitude of the correlation, and the feature to be selected which has correlation with the credit condition of the target client and has larger correlation is reserved as the target feature. Thus, through feature screening, the complexity and the operation time for generating the sample user portrait can be reduced well, and the accuracy and the stability of the generated sample user portrait are improved.

Illustratively, the asset features, credit features, base features, behavioral features extracted from the sample data are screened. Taking the asset characteristics as an example, the asset characteristics comprise characteristics a-f, wherein the characteristics a-d are related with the credit status of the target client and have larger relevance. And reserving the characteristics a-d in the asset characteristics as asset characteristics in the target characteristics through characteristic screening.

Specifically, target features are combined to generate a sample user representation. The sample user portrait can comprehensively and accurately reflect the characteristics, behavior habits and the like of the target client from multiple dimensions and provides basis for subsequent basic evaluation. Of course, the sample user representation may be visually presented, thus facilitating the viewing and analysis by business personnel.

The target features include basic features, asset features, credit features and behavior features.

In this embodiment, the obtained sample client raw data is preprocessed, so that relatively complete and accurate high-quality sample client data can be obtained, the target features are obtained by screening the to-be-selected features extracted from the sample client data, and the number of target features for generating the user portrait can be reduced on the basis of retaining the features related to the credit condition of the client, so that the complexity and the operation time of generating the user portrait are reduced, and the preset credit evaluation model constructed based on the user portrait and the complexity and the operation time of the finally obtained target credit evaluation model are further reduced.

In some embodiments, the preset credit assessment model includes a preset credit assessment model, a preset purchase behavior scoring model, and/or a preset credit limit assessment model.

At this time, based on the sample user portrayal, a preset trust evaluation model is constructed, including:

and constructing a preset comprehensive credit rating model based on the characteristic items contained in the sample user portrait. And/or constructing a preset purchasing behavior scoring model based on the feature items contained in the sample user portrait. And/or constructing the preset credit limit assessment model based on the feature items, the preset comprehensive credit rating model and the preset purchasing behavior rating model contained in the sample user portrait.

The method comprises the steps that a preset credit limit assessment model is used for assessing the credit limit of a target client based on a sample user portrait, a preset purchasing behavior scoring model is used for assessing purchasing behavior scores of the target client based on the sample user portrait, and a preset comprehensive credit scoring model is used for assessing the comprehensive credit scores of the target client based on the sample user portrait.

Correspondingly, the target credit rating model comprises a target comprehensive credit rating model, a target purchasing behavior rating model and/or a target credit limit rating model.

The following describes several cases involved in this embodiment, respectively:

case 1: in the sample user representation, the behavior features comprise working behavior features, the asset features comprise inherent asset feature items and equipment asset feature items, and the credit feature comprises an internal credit feature item.

Based on the feature items contained in the sample user portrait, a preset comprehensive credit rating model is constructed, which comprises the following steps: and constructing a preset comprehensive credit rating model based on the basic features, the inherent asset feature items, the equipment asset feature items, the internal credit investigation feature items and the working behavior feature items contained in the sample user portrait and the first preset super parameters.

Specifically, based on basic features, inherent asset feature items, equipment asset feature items, internal credit feature items and all feature items in a working behavior feature item in a sample user portrait, a preset model is selected, the positions of the feature items in the preset model are determined, and then a first preset super-parameter is configured into the preset model to obtain a preset comprehensive credit rating model.

Correspondingly, training and evaluating the preset comprehensive credit rating model based on basic features, inherent asset feature items, equipment asset feature items, internal credit feature items and working behavior feature items in the sample user portraits of the sample data set and comprehensive credit rating in sample labels corresponding to the sample user portraits in the sample data set, and adjusting first preset super parameters in the preset comprehensive credit rating model to obtain a target comprehensive credit rating model.

The classification algorithm used for the preset comprehensive credit rating model may be various, such as logistic regression, support Vector Machine (SVM), decision tree, etc., that is, the preset model used for constructing the preset comprehensive credit rating model may be various, such as logistic regression model, support Vector Machine (SVM) model, decision tree model, etc. For example, a logistic regression algorithm may be used to predict the composite trust score based on feature items for each dimension in the sample user representation.

Illustratively, the logistic regression algorithm is formulated as

$$h_{\theta}(x)＝\frac{1}{1+e^{-\theta^Tx}}$$

Where $ y$ represents the predicted composite credit score, $\theta$ represents the first predicted hyper-parameter, $x$ represents the feature items (or feature vectors) in the various dimensions of the sample user representation.

In addition, the first preset super-parameters comprise first preset weight parameters and scores corresponding to all feature items in the features of different dimensions. The first preset weight parameters correspond to features of different dimensions, namely, the first preset weight has a corresponding relation with basic features, asset features, credit investigation features and working behavior features.

The first preset weight parameter may be determined based on an actual scene, so that a constructed preset comprehensive credit score model approximates to a comprehensive credit score in an actual situation, that is, a corresponding sample label, based on a comprehensive credit score obtained by credit risk assessment of the target client by the sample user portrait.

Illustratively, the inherent asset characteristic item includes, for example, an inherent asset type, an inherent asset value, and the device asset characteristic item in the asset characteristic includes a device asset value. In general, physical forms of the inherent assets include houses and buildings, office equipment, machinery, transportation equipment, and the like.

Illustratively, the internal credit features include an overdue amount ratio, an average overdue amount, a balance of the good ratio, and an average pay-per-view ratio. The overdue amount accounts for the ratio of the current overdue amount to the total amount of money, the average overdue amount is the ratio of the accumulated overdue amount of the host to the number of the purchasing devices, and the balance ratio of the amount of money is the ratio of the amount of money to be received to the amount of money to be purchased.

Illustratively, the feature items in the base feature include tax credits, company types, and registered capital.

Illustratively, the working behavior characteristic items comprise characteristic items of four aspects of starting, oil consumption, equipment fluidity and customer letter increasing. In the aspect of starting work, the working behavior characteristic items comprise a starting work rate and a daily average working time length; in the aspect of oil consumption, the working behavior characteristic items comprise month average oil consumption and accumulated oil consumption; in the aspect of equipment mobility, the working behavior characteristic items comprise an average value of equipment fluctuation times and an average value of equipment numbers in an area where equipment is located; in terms of customer augmentation, the work behavior feature items include collaborative company type and engineering magnitude.

The operating rate is a mean value of operating rates of each device in a first preset time period, and for each device, the operating rate of the device in the first preset time period is a ratio of days, in which the working time of the device in the first preset time period is not 0, to total days in the first preset time period. The average daily working time length is the average value of the average daily working time length of each device in the second preset time period. And the average fuel consumption of the month is the average value of the total fuel consumption of all the equipment in each month in the third preset time period. And the accumulated oil consumption is the total oil consumption of all the equipment in the fourth preset time period. The equipment change times are the average value of the equipment position (longitude and latitude) change times in a fifth preset time period. The average value of the number of devices in the area where the device is located is the average value of the number of devices in the area (for example, county level) where the current device is located.

Illustratively, the first preset weight parameters corresponding to the basic feature, the asset feature (intrinsic asset feature item and equipment asset feature item), the credit feature (internal credit feature item) and the behavior feature (working behavior feature item) are respectively 10%, 40%, 25% and 25%. In the dimension of the basic feature, the maximum score corresponding to the registered capital is 30, the maximum score corresponding to the tax payment credit is 40, and the maximum score corresponding to the company type is 30; in the internal credit feature item of the dimension of the credit feature, the overdue balance accounts for 30 corresponding maximum scores, the maximum score corresponding to the average overdue number of the equipment is 30, the loan balance accounts for 20 corresponding maximum scores, and the maximum score corresponding to the average first payment proportion is 20; in the dimension of the asset characteristics, the maximum score corresponding to the inherent asset characteristic item is 20, and the maximum score corresponding to the equipment asset characteristic item is 80; in the working behavior characteristic items of the dimension of the behavior characteristic, the maximum score of the characteristic item corresponding to the starting aspect is 30, the maximum score of the characteristic item corresponding to the oil consumption aspect is 30, the maximum score of the characteristic item corresponding to the equipment fluidity aspect is 20, and the maximum score of the characteristic item corresponding to the customer letter increasing aspect is 20. It is known that, in this example, the maximum score of the integrated credit score is 100.

It will be appreciated that for any feature item, the score corresponding to that feature item when in a different condition is different.

More specifically, in the above example, in this dimension of the basic feature, the different conditions of the registered capital a include 500 ten thousand or less (a <500 ten thousand), 500 ten thousand-1000 ten thousand (500 ten thousand < = a < = 1000 ten thousand), 1000 ten thousand-3000 ten thousand (1000 ten thousand < a < = 3000 ten thousand), 3000 ten thousand-5000 ten thousand (3000 ten thousand < a < = 5000 ten thousand), 5000 ten thousand or more (a >5000 ten thousand), the corresponding scores being 0, 5, 10, 20, 30, respectively; different conditions of tax payment include class A tax payers and common tax payers, and corresponding scores are respectively 40 and 20; different conditions of the company types include national enterprises, stock-making enterprises without marketing, collective ownership systems, private, affiliated, individual and the like, the national enterprises corresponding to the stock-making enterprises are 30 in score, the stock-making enterprises without marketing, the collective ownership systems, the private and affiliated are 20 in score, and the individual corresponding score is 10.

In the dimension of the asset characteristics, the inherent asset types are divided into different conditions according to the physical forms of the owned inherent assets, wherein the conditions comprise possession of houses and buildings, no houses and buildings, office equipment, no houses and buildings and no office equipment, and the corresponding scores are respectively 10, 5 and 0; different conditions of the intrinsic asset value b include more than 5 hundred million (b >5 hundred million), 1 hundred million-5 hundred million (1 hundred million < = b < = 5 hundred million), 5000 ten thousand-1 hundred thousand (5000 ten thousand < = b <1 hundred million), 5000 ten thousand (b <5000 ten thousand), the corresponding scores are 10, 5, 2, 0 respectively; different conditions of the equipment asset value c include more than 1 hundred million (c >1 hundred million), 5000-1 hundred million (1 hundred million < = c < = 5000 ten thousand), 1000-5000 ten thousand (1000 ten thousand < = c <5000 ten thousand), 300-1000 ten thousand (300 ten thousand < = c <1000 ten thousand), 300 ten thousand (c <300 ten thousand), and the corresponding scores are 80, 60, 40, 20, 10 respectively.

In the dimension of the credit sign feature, the host overdue amount d is divided into four conditions of 0 (d=0), 0-5% (0<d < =5%), 5% -10% (5% < d < =10%), more than 10% (d > 10%) according to the ratio, and the corresponding scores are 30, 20, 10 and 0 respectively; the loan balance ratio e is divided into three conditions of less than 30 percent (e <30 percent), 30% -50 percent (30% <=e < =50 percent) and more than 50 percent (e >50 percent) according to the ratio, and the corresponding scores are respectively 20, 10 and 0; the average overdue number f is divided into 4 cases of 0 (f=0), 0-0.5 (0<f < =0.5), 0.5-1 (0.5 < f < 1), 1 (f=1) or more according to the ratio, and the corresponding scores are 30, 20, 10, 0 respectively; the average first-payment proportion g is divided into three conditions of less than 30% (g < 30%), 30% -50% (30% <=g < =50%), more than 50% (g > 50%) according to the ratio, and the corresponding scores are respectively 0, 10 and 20.

In the dimension of the behavioral characteristics, the operating rate h is, for example, the average value of the operating rate of each equipment in one month, different conditions are divided according to the ratio, wherein the operating rate comprises less than 20 percent (h < 20%), 20% -40% (20% <=h < =40%), 40% -60% (40% < h < =60%), and more than 60% (h > 60%), and the corresponding scores are 3, 6, 10 and 15 respectively; the daily working time length i is, for example, the average value of the daily working time length of each device in one month, different conditions are divided according to the time length, wherein the conditions comprise less than 2h (i <2 h), 2h-5h (2 h < = i < = 5 h), 5h-8h (5 h < = 8 h), and more than 8h (i >8 h), and the corresponding scores are 3, 6, 10 and 15 respectively; dividing different conditions according to the numerical values, wherein the average fuel consumption j is the average value of the total fuel consumption of all equipment in three months, and the average fuel consumption j comprises less than 5L (j < 5L), 5L-15L (5L < = j < = 15L), 15L-30L (15L < = 30L), and more than 30L (j > 30L), and the corresponding scores are 3, 6, 10 and 15 respectively; the accumulated oil consumption k is the total oil consumption of all equipment in one month, different conditions are divided according to the numerical values, wherein the total oil consumption is below 5L (k < 5L), 5L-15L (5L < = k < = 15L), 15L-30L (15L < = 30L), and more than 30L (k > 30L), and the corresponding scores are 3, 6, 10 and 15 respectively; dividing different conditions according to numerical values by using an average value l of equipment position change times, such as an average value of longitude and latitude change times of a single equipment within one month, wherein the average value l is less than 5 times (l <5 times), 5 times-10 times (5 times < = l < = 10 times), and more than 10 times (l >10 times), and the corresponding scores are respectively 10, 5 and 2; dividing the average value m of the equipment number in the area where the current equipment is positioned into different conditions according to the numerical value, wherein the conditions comprise more than 10 (m is greater than 10), 5-10 (5 < m < = 10), 1-5 (1 < m < = 5), and 1 (m=1), and the corresponding scores are respectively 10, 5, 2 and 0; different conditions of the type of the cooperative company comprise national enterprises, stock marketing enterprises, stock system non-marketing enterprises, collective ownership systems, private systems, individuals and affiliated systems, wherein the corresponding scores of the national enterprises and the stock marketing enterprises are 10, the corresponding scores of the stock system non-marketing enterprises, the collective ownership systems, the private systems and the affiliated systems are 5, and the corresponding scores of the individuals are 0; the engineering magnitude is divided into different conditions according to engineering funds n, wherein the engineering magnitude comprises less than 500 ten thousand (n <500 ten thousand), 500 ten thousand-2000 ten thousand (500 ten thousand < = n < = 2000 ten thousand), 2000 ten thousand-5000 ten thousand (2000 ten thousand < n < = 5000 ten thousand), 5000 ten thousand-1 hundred million (5000 ten thousand < n < = 1 hundred million), more than 1 hundred million (n >1 hundred million), and the corresponding scores are respectively 0, 2, 4, 8 and 10.

Of course, the first preset super-parameters further include learning rate, regularization parameters, loss function, evaluation indexes and the like. The evaluation index is the above-mentioned index such as accuracy, recall F1, etc., and the chart such as ROC curve and PR curve. More, the first preset super-parameters further comprise training data batch size, iteration times, optimization algorithm and the like. It will be appreciated that other super-parameters may be included in the first preset super-parameters, or fewer super-parameters may be included, based on the needs of the actual model construction, training and evaluation.

Case 2: the behavioral characteristics include purchasing behavioral characteristics items.

Based on feature items contained in the sample user portrait, a preset purchasing behavior scoring model is constructed, which comprises the following steps: and constructing a preset purchasing behavior scoring model based on purchasing behavior feature items contained in the sample user portrait and the second preset super parameters.

Specifically, a preset model is selected based on each feature item in the purchase behavior feature items, the positions of each feature item in the preset model are determined, and then a second preset super parameter is configured to the preset model to obtain a preset purchase behavior scoring model.

Correspondingly, based on the purchasing behavior feature items in the sample user portraits of the sample data set and the purchasing behavior scores in the sample labels corresponding to the sample user portraits in the sample data set, the second preset super parameters in the preset purchasing behavior score model are adjusted, namely the preset purchasing behavior score model is trained and evaluated, and the second preset super parameters in the preset purchasing behavior score model are adjusted to obtain the target purchasing behavior score model.

The classification algorithm used for the preset purchase behavior scoring model may be various, for example, logistic regression, support Vector Machine (SVM), decision tree, etc., that is, the preset model used for constructing the preset purchase behavior scoring model may be various, for example, logistic regression model, support Vector Machine (SVM) model, decision tree model, etc. For example, a logistic regression algorithm may be used to predict purchase behavior scores based on the purchase behavior feature items of the sample user portraits.

Illustratively, the logistic regression algorithm is formulated as

$$h_{\theta}(x)＝\frac{1}{1+e^{-\theta^Tx}}$$

Where $ y $ represents the predicted purchase behavior score, $\theta $ represents the second predicted hyper-parameter, $x $ represents a feature item (or feature vector) in the purchase behavior feature of the sample user representation.

In addition, the second preset super-parameters comprise second preset weight parameters and scores corresponding to all the characteristic items in the purchasing behavior characteristic items.

The second preset weight parameters correspond to the feature items in the purchasing behavior feature items. The second preset weight parameter has a corresponding relationship with the feature item in the purchase behavior feature item, that is, the purchase amount coverage ratio, the last purchase duration, the average number of purchases, and the number of cash-in-charge purchases.

The second preset weight parameter may be determined based on an actual scenario, so that a purchase behavior score obtained by performing credit risk assessment on the sample client by the constructed preset purchase behavior score model is close to a purchase behavior score in an actual situation, that is, a corresponding sample label.

Illustratively, the purchase behavior feature items include a purchase amount coverage ratio, a last purchase duration, a monthly purchase amount, and a cash-in-place purchase amount.

The purchasing quantity coverage ratio is an average value of the ratio of the purchasing quantity to the demand quantity of each month in a sixth preset time period, the last purchasing time is the time between the last purchasing time and the current time, and the average purchasing quantity of each month is an average value of the purchasing quantity of each month in a seventh preset time period.

Illustratively, the second preset weight parameters corresponding to the purchase amount coverage ratio, the last purchase duration, the average purchase amount, and the pay-as-you-go purchase amount are all 1. The purchase amount coverage ratio is, for example, a purchase amount coverage ratio of approximately 3 months, the corresponding maximum score is 20, the last purchase time corresponds to a maximum score of 20, the average number of purchases per month is, for example, the average number of purchases per month of approximately 3 months, the corresponding maximum score is 30, and the present-and-the-charge number is, for example, the present-and-charge number of approximately half a year, the corresponding maximum score is 30.

More specifically, in the above example, the purchase quantity coverage ratio o divides different conditions according to the ratio, including 80% or more (o > 80%), 50% -80% (50% <=o < =80%), 30% -50% (30% <=o < 50%), 0-30% (0 < =o < 30%), and the corresponding scores are 20, 10, 6, 3, respectively; the last purchase time p is divided into different conditions according to time length, wherein the different conditions comprise 3 months (p <3 months), 3-6 months (3 months < = p < = 6 months), 6-12 months (6 months < p < = 12 months) and more than 12 months (p >12 months), and the corresponding scores are respectively 20, 10, 6 and 3; the number of the average monthly purchases is divided into 4 sections according to the box type diagram, the scores corresponding to the 4 sections are 5, 10, 20 and 30 respectively, and the score corresponding to the larger the number of the average monthly purchases is, the higher the score is; the pay-as-you-go purchase amount is divided into 4 sections according to the box diagram, and the scores corresponding to the 4 sections are 5, 10, 20 and 30 respectively, and the score corresponding to the larger the pay-as-you-go purchase amount is, the higher the corresponding score is.

Of course, similar to the first preset super-parameters, the second preset super-parameters further include learning rate, regularization parameters, loss function, evaluation index, and the like. The evaluation index is the above-mentioned index such as accuracy, recall F1, etc., and the chart such as ROC curve and PR curve. More, the second preset super parameters further comprise training data batch size, iteration times, optimization algorithm and the like. It will be appreciated that other super-parameters may be included in the first preset super-parameters, or fewer super-parameters may be included, based on the needs of the actual model construction, training and evaluation.

Case 3: the asset characteristics include a special asset characteristic term that characterizes whether the target client has a special asset.

Wherein the particular asset may be, for example, a blender station apparatus. In general, a great deal of funds are consumed for the establishment of the stirring station, and a great deal of benefits are brought by the existence of the stirring station, so that clients with the stirring station have relatively great strength and less credit risk.

When the preset credit assessment model is built based on characteristic items, a preset comprehensive credit scoring model and a preset purchasing behavior scoring model contained in the sample user portrait, the preset credit rating model is built based on special asset characteristic items, the preset comprehensive credit scoring model, the preset purchasing behavior scoring model and a third preset super parameter in the sample user portrait.

In this case, the preset comprehensive credit rating model and the preset purchase behavior rating model do not exist or are used alone, but are used as feature items: and integrating the credit rating and the purchasing behavior rating for constructing a preset credit limit evaluation model.

Specifically, after the preset comprehensive credit rating model and the preset purchasing behavior rating model are determined based on the mode, the preset model is selected based on the characteristic items of the special asset, the preset comprehensive credit rating model and the preset purchasing behavior rating model in the sample user portrait, and the positions of the characteristic items of the special asset, the comprehensive credit rating and the purchasing behavior rating in the preset model are determined. And then, configuring a third preset hyper-parameter to the preset model to obtain a preset credit evaluation model.

The regression algorithm used in the preset credit evaluation model may be various, such as linear regression, decision tree regression, etc. That is, the preset model used for constructing the preset credit evaluation model may be various, such as a linear regression model, a decision tree regression model, and the like. For example, a linear regression algorithm may be used to predict credit limits based on particular asset characteristic items in a sample user representation.

Illustratively, the linear regression algorithm is formulated as

$$y＝\theta^Tx$$

Wherein $ y $ represents the predicted value, i.e., the predicted credit, $\theta $ represents the model parameters, i.e., the third preset super parameters, $x $ represents the feature vector, i.e., the special asset feature item in the sample user representation.

The method comprises the steps of selecting a preset model based on basic characteristics, inherent asset characteristic items, equipment asset characteristic items, internal credit characteristic items, working behavior characteristics, purchasing behavior characteristic items and special asset characteristic items contained in a sample user portrait, determining the positions of the characteristic items in the preset model, and directly configuring a first preset super-parameter, a second preset super-parameter and a third preset super-parameter to the preset model to obtain a preset credit evaluation model.

Case 4: in the sample user portrait, the behavior features comprise working behavior features and purchasing behavior feature items, the asset features comprise inherent asset feature items and equipment asset feature items, and the credit feature comprises an internal credit feature item.

For the construction of the preset comprehensive credit rating model and the preset purchasing behavior rating model, reference can be made to the above.

Case 5: in the sample user representation, the behavior features comprise working behavior features, the asset features comprise inherent asset feature items, equipment asset feature items and special asset feature items, and the credit feature comprises an internal credit feature item.

And determining a preset comprehensive credit rating model and a preset purchasing behavior rating model based on the mode. And then, constructing a preset target credit limit assessment model based on special asset characteristic items in the sample user portrait, and on comprehensive credit scores obtained by a preset comprehensive credit assessment model, a preset purchasing behavior score model and a third preset super parameter.

In this case, the preset purchasing behavior score model does not exist or is used alone, but as a feature item: and the purchase behavior score is used for constructing a preset credit limit evaluation model.

Specifically, a preset model is selected based on a special asset feature item, a comprehensive credit score and a preset purchasing behavior score model in a sample user portrait, and the positions of the special asset feature item, the comprehensive credit score and the preset purchasing behavior score model in the preset model are determined. And then, configuring a third preset hyper-parameter to the preset model to obtain a preset credit evaluation model. The characteristic item of the comprehensive credit score is determined based on a preset comprehensive credit score model.

Or, based on the purchase behavior feature item, the special asset feature item and the feature item of the comprehensive credit rating contained in the sample user portrait, a preset model is selected, the position of each feature item in the preset model is determined, and the second preset super parameter and the third preset super parameter are directly configured to the preset model to obtain a preset credit rating assessment model.

Case 6: the behavioral characteristics include purchasing behavioral characteristics, and the asset characteristics include special asset characteristics.

And determining a preset comprehensive credit rating model and a preset purchasing behavior rating model based on the mode. And then, constructing a preset target credit limit assessment model based on special asset feature items in the sample user portrait, a preset comprehensive credit rating model, a purchase behavior score obtained based on the preset purchase behavior scoring model and a third preset super parameter.

In this case, the preset comprehensive credit score model does not exist or is used alone, but is used as a feature item: the comprehensive credit rating is used for constructing a preset credit limit evaluation model.

Specifically, a preset model is selected based on a special asset feature item, a preset comprehensive credit rating model and a purchasing behavior score in a sample user portrait, and the positions of the special asset feature item, the preset comprehensive credit rating model and the purchasing behavior score in the preset model are determined. And then, configuring a third preset super parameter to the preset model to obtain a preset credit rating assessment model. The feature item of the purchase behavior score is determined based on a preset purchase behavior score model.

Or, based on the feature items in the dimensions of the basic feature, the asset feature, the credit feature, the working behavior feature and the special asset feature item contained in the sample user portrait and the feature item of the preset purchasing behavior scoring model, a preset model is selected, the position of each feature item in the preset model is determined, and the first preset super-parameter and the third preset super-parameter are directly configured to the preset model to obtain the preset credit rating assessment model.

Case 7: in the sample user portrait, the behavior features comprise working behavior features and purchasing behavior feature items, the asset features comprise inherent asset feature items, equipment asset feature items and special asset feature items, and the credit feature comprises an internal credit feature item.

And determining a preset comprehensive credit rating model and a preset purchasing behavior rating model based on the mode. And then, constructing a preset target credit rating evaluation model based on special asset feature items in the sample user portrait, comprehensive credit rating obtained based on a preset comprehensive credit rating model, purchase behavior rating obtained based on a preset purchase behavior rating model and a third preset super parameter.

In this case, both the preset comprehensive credit rating model and the preset purchasing behavior rating model may exist alone to be usable.

Specifically, a preset model is selected based on the characteristic items of the special asset, the comprehensive credit score and the purchase behavior score in the sample user portrait, and the positions of the characteristic items of the special asset, the comprehensive credit score and the purchase behavior score in the preset model are determined. And then, configuring a third preset hyper-parameter to the preset model to obtain a preset credit evaluation model.

In one embodiment, when the preset credit rating system is constructed based on the special asset feature item, the preset comprehensive credit rating model, the preset purchasing behavior rating model and the third preset super parameter, a two-dimensional matrix is constructed based on the preset comprehensive credit rating model and the preset purchasing behavior rating model, and then the preset credit rating system is constructed based on the third preset super parameter, the two-dimensional matrix and the special asset feature item.

The selected preset model for constructing the preset credit limit evaluation model is determined based on matrix operation and linear rules. In the preset credit evaluation model, a basic credit is determined through matrix operation based on a two-dimensional matrix, then other credit is determined based on special asset characteristic items, and the sum of the basic credit and the other credit is determined as the credit of a sample client and is output.

It should be noted that, when the preset credit evaluation model is constructed, the following design logic is required to be followed: the high-quality large client threshold is low, the medium and small clients threshold is high, namely even if the purchase behavior score of the target client with high comprehensive credit rating is low, the corresponding basic credit limit is not 0, and the purchase behavior score of the target client with low comprehensive credit rating is higher, the corresponding basic credit limit is not 0.

Illustratively, the underlying credit line may be as shown in FIG. 2 under different purchase behavior scores and different composite credit scores. In general, no matter whether the target client has purchasing behavior, the credit limit can be determined to be not 0 under the condition that the comprehensive credit score is higher. As shown in fig. 2, even if the purchase behavior score of the target client is 0, when the comprehensive credit score is high, for example, 70 is reached, the basic credit limit of the target client may be determined to be 30. In fig. 2, the higher the purchase score, the higher the credit line, and in addition, the higher the overall credit score, the higher the credit line. It should be noted that, in general, the comprehensive credit rating is taken as an important loop in credit risk assessment, and the target client with a lower comprehensive credit rating may have a higher credit risk, a worse credit or a worse operation, so that the credit rating of the target client is determined to be not 0 only when the comprehensive credit rating is higher, for example, higher than 40, so that the credit risk can be better reduced. In addition, in general, when the comprehensive credit score is higher than 40, the customer has more equipment assets, the equipment asset value is higher, and the overdue amount of the host is smaller.

For example, if the special asset characteristic item characterizes that the sample client has a special asset, it is determined that the other credit line is 50 ten thousand, and based on the basic credit line, for example 250 ten thousand, the other credit line is added, so that the credit line of the sample client is 300 ten thousand.

In several cases given in the foregoing embodiments, the preset comprehensive credit rating model may be replaced with a comprehensive credit rating obtained based on the preset comprehensive credit rating model, and/or the preset purchase behavior rating model may be replaced with a purchase behavior rating obtained based on the preset purchase behavior rating model, which may all construct the preset credit rating evaluation model according to the foregoing manner.

The method comprises the steps of constructing a two-dimensional matrix based on a preset comprehensive credit rating model or comprehensive credit rating, a preset purchasing behavior rating model or purchasing behavior rating, and then constructing a preset credit limit assessment model based on a third preset super-parameter, the two-dimensional matrix and special asset characteristic items.

Referring to fig. 3, in an exemplary embodiment, a trust evaluation model application method is provided, which is applied to any device. As shown in fig. 3, the trust evaluation model application method includes steps S301-S302:

S301: and acquiring a real-time user image of the target client.

Wherein the real-time user image includes dimensions of basic features, asset features, credit features, and behavioral features. Of course, other dimensions may be included in the user representation according to different actual requirements.

Similarly, reference is made to the acquisition and introduction of sample user portraits described above with respect to the acquisition and introduction of real-time user portraits of target customers.

S302: and carrying out credit risk assessment on the target client based on the real-time user portrait and the target credit trust assessment model to obtain a real-time assessment result of the target client.

The real-time evaluation result comprises a real-time comprehensive credit rating, a real-time purchasing behavior rating and/or a real-time credit limit.

And inputting each characteristic item in the real-time user image of the target client into a target credit assessment model, and carrying out credit risk assessment on the target client to obtain the real-time comprehensive credit rating, the real-time purchase behavior rating and/or the real-time credit rating of the target client.

It can be understood that the target trust evaluation model herein is the target trust evaluation model obtained in the above trust evaluation model construction method.

In this embodiment, a real-time user portrait of a target client with multiple dimensions including basic features, asset features, credit investigation features, behavior features and the like is obtained, and credit risk is estimated for the target client based on the real-time user portrait with multiple dimensions and the target credit investigation estimation model with higher accuracy and stability, so that the accuracy of risk estimation can be effectively improved, and the accuracy of credit investigation decision can be further improved.

In some embodiments, the real-time user image of the target client further includes an external credit assessment feature item, and before the target client is subjected to credit risk assessment based on the real-time user image of the target client by applying the target credit assessment model, whether the target client is out of trust is determined based on the external credit assessment feature item, and whether to suspend credit to the target client is determined based on whether the target client is out of trust.

If so, determining that the target client is out of trust based on the external credit investigation characteristic item, suspending credit investigation to the target client, and stopping credit risk assessment to the target client. If not, based on the external credit feature item, the target client is determined not to be lost, the credit risk assessment is continued to be carried out on the target client by continuing to give credit to the target client.

Specifically, if the external credit investigation feature item does not meet the basic credit investigation condition, determining that the target client loses credit. Correspondingly, if the external credit feature item meets the basic credit condition, determining that the target client is not lost.

The external credit feature includes whether the host is overdue, restricted from consumption, determined to be a trusted executor, frozen by the asset, and credit overdue.

Accordingly, the underlying trust conditions include host no expiration, unrestricted consumption, no determination as a trusted executor, no freezing by the asset, and no expiration of the credit.

Credit sales generally refers to credit transactions in which an enterprise sells goods or services to units or individuals by means of installments, deferred payments, and the like. The credit has no overdue, i.e. the customer does not overdue when buying goods or services by means of installment or deferred payment.

For example, a charge free expiration may refer to a periodic repayment without an expiration occurring during the customer's use of a product provided by the business, such as an oil loan.

That is, if the plurality of external credit feature items are not overdue, not restricted from consumption, not determined to be a trusted executor, not frozen by the asset, and not overdue by credit, the external credit feature items satisfy the basic credit condition, and it can be determined that the target client is not trusted. Accordingly, if the external credit feature item does not meet the basic credit condition, the target client can be determined to be out of credit.

In addition, in the case of determining that the target client is not trusted, a trusted service, such as oil credit, is not provided to the client, and at this time, it is meaningless to perform credit risk assessment on the target client based on the target trusted assessment model. Therefore, when the target client is determined not to be out of trust based on the external credit investigation characteristic item and the credit risk assessment is carried out on the target client based on the target credit investigation assessment model when the target client is continuously credited to the target client, when the target client is determined to be out of trust based on the external credit investigation characteristic item and the credit risk assessment is not continuously carried out on the target client when the credit investigation to the target client is stopped. In this way, unnecessary computing resource consumption may be reduced.

In this embodiment, if the external credit feature item in the real-time user image of the target client represents that the target client loses credit, that is, the external credit feature item does not meet the basic credit condition, it indicates that the credit of the target client is poor, and at this time, the credit giving to the target client is suspended, so that unnecessary loss can be avoided, and the credit giving risk is reduced.

Specifically, in some embodiments, the target credit rating model includes a target comprehensive credit rating model, a target purchase behavior rating model, and/or a target credit rating model.

Based on the real-time user portrait and the target credit trust evaluation model, credit risk evaluation is carried out on the target client to obtain a real-time evaluation result of the target client, and the method comprises the following steps: and inputting the characteristic items in the real-time user image into a target comprehensive credit rating model, a target purchasing behavior rating model and/or a target credit limit evaluation model to obtain a real-time comprehensive credit rating, a real-time purchasing behavior rating and/or a real-time credit limit.

Reference is made to the above process of constructing a preset trust evaluation model:

and inputting the basic characteristics, the inherent asset characteristic items, the equipment asset characteristic items, the internal credit investigation characteristic items and the working behavior characteristic items in the real-time user image into a target comprehensive credit investigation evaluation model to obtain the real-time comprehensive credit investigation score of the target client.

And inputting each characteristic item in the purchasing behavior characteristic items in the real-time user image into a target purchasing behavior scoring model to obtain the purchasing behavior score of the target client.

And inputting special asset characteristic items, comprehensive credit rating and purchasing behavior rating in the real-time user image into a target credit rating system evaluation model to obtain the credit rating system of the target client. Or inputting the basic characteristics, the inherent asset characteristic items, the equipment asset characteristic items, the internal credit investigation characteristic items, the working behavior characteristic items, the special asset characteristic items and the purchasing behavior scores in the real-time user image into a target credit line assessment model to obtain the credit line of the target client. Or, inputting each characteristic item, the special asset characteristic item and the comprehensive credit rating in the purchase behavior characteristic items in the real-time user image into the target credit rating system evaluation model to obtain the credit rating of the target client.

In some embodiments, after the real-time integrated credit score is derived based on the target integrated credit score model, credit to the target customer is suspended or continued based on the real-time integrated credit score.

Under the condition that the higher the comprehensive credit score is, the lower the credit risk is, if the real-time comprehensive credit score is lower than the preset comprehensive credit score, the credit to the target client is suspended; otherwise, continuing to trust the target client.

Of course, under the condition that the comprehensive credit score is lower and the credit risk is lower, if the real-time credit score is higher than the preset comprehensive credit score, the credit to the target client is suspended; otherwise, continuing to trust the target client.

In this embodiment, since the level of the comprehensive credit rating of the target client reflects the level of the credit risk of the target client, if the real-time comprehensive credit rating is lower than the preset comprehensive credit rating, the credit to the target client is suspended, so that the situation that the credit service is provided for the target client under the condition that the credit risk of the target client is higher can be effectively avoided, and the risk is reduced.

In some embodiments, a real-time user representation of the target customer is periodically acquired and credit risk assessment is performed on the target customer based on a target trust assessment model.

And when the comprehensive credit rating, the purchasing behavior rating and the credit limit of the target client are changed, recording the latest comprehensive credit rating, the purchasing behavior rating and the credit limit.

TABLE 1

Illustratively, as shown in the above table 1, after credit risk assessment is performed on the clients a-e by using the target trust assessment model given in the above embodiment, the assessment result is: the purchase behavior scores of a-e are 75, 70 and 70 respectively, the comprehensive credit rating is 85.5, 79.5, 87.85, 78.75 and 61 respectively, and the credit limit is 200, 150, 100 and 50 respectively. And the credit limit of a-e determined by other modes is respectively 100, 50 and 38.06. In contrast, in this embodiment, the corresponding credit limit can be better determined based on the credit risk of the target client, so as to reduce the credit risk.

In some embodiments, after credit risk assessment is performed on the target client, the comprehensive credit rating, the purchase behavior rating and/or the credit limit of the target client are obtained, the target client may be screened based on the comprehensive credit rating, the purchase behavior rating and/or the credit limit of the target client when needed, and relevant information of the target client with the comprehensive credit rating, the purchase behavior rating and/or the credit limit meeting the needs is checked.

It will be appreciated that, based on the credit line, the target clients may be distributed in a pyramid shape, i.e. the higher the credit line, the fewer the number of corresponding target clients. Illustratively, the number of target clients not subject to credit is 296, wherein 24 target clients overdue or losing credit is performed, the number of target clients with credit rating of 30 ten thousand is 48, the number of target clients with credit rating of 50 ten thousand is 33, the number of target clients with credit rating of 80 ten thousand is 19, the number of target clients with credit rating of 100 ten thousand is 16, the number of target clients with credit rating of 150 ten thousand is 10, the number of target clients with credit rating of 200 ten thousand is 9, and the number of target clients with credit rating of 250 ten thousand is 2.

In some embodiments, after the target comprehensive credit rating model and the target purchase behavior rating model are determined, the comprehensive credit rating and the purchase behavior rating of the target client can be determined based on the target comprehensive credit rating model and the target purchase behavior rating model, and the punctual distribution or the bar graph distribution of the target client under different purchase behavior ratings and different comprehensive credit ratings is checked in a chart form. Based on sample labels in the test data set in the same way, the punctual distribution or the bar graph distribution of the target clients under different purchase behavior scores and comprehensive credit giving scores is checked in the form of a chart. By comparing the punctuation distribution or the bar graph distribution of the target clients under the two conditions, the accuracy and the stability of the target comprehensive credit rating model and the target purchasing behavior rating model can be evaluated.

Exemplary apparatus

Correspondingly, the embodiment of the application also provides a trusted evaluation model construction device, which comprises an acquisition module 401, a construction module 402 and a training and evaluation module 403.

Wherein, the liquid crystal display device comprises a liquid crystal display device,

an acquisition module 401 for acquiring a sample dataset; the sample dataset includes a sample user representation and a sample label; the sample user portraits comprise basic features, asset features, credit features and behavioral features of the target client; the sample label comprises comprehensive credit rating, purchasing behavior rating and/or credit limit of the target client;

a construction module 402, configured to construct a preset trust evaluation model based on the sample user representation;

the training and evaluating module 403 is configured to train and evaluate the preset trust evaluation model based on the sample user portrait and the sample tag, so as to obtain a target trust evaluation model; the target credit evaluation model is used for carrying out credit risk evaluation on the target client based on the real-time user image to obtain real-time comprehensive credit scores, real-time purchase behavior scores and/or real-time credit limit of the target client.

The trust evaluation model construction device provided by the embodiment belongs to the same application conception as the trust evaluation model construction method provided by the embodiment of the application, can execute the method provided by any embodiment of the application, and has the corresponding functional module and beneficial effects of the execution method. Technical details not described in detail in this embodiment may refer to specific processing content of the trust evaluation model construction method provided in the foregoing embodiment of the present application, and will not be described herein.

Correspondingly, the embodiment of the application also provides a trust evaluation model construction device, which comprises an acquisition module 501 and an evaluation module 502.

an acquisition module 501, configured to acquire a real-time user image of a target client; the real-time user image comprises basic characteristics, asset characteristics, credit investigation characteristics and behavior characteristics;

the evaluation module 502 is configured to perform credit risk evaluation on the target client based on the real-time user portrait and the target trust evaluation model, so as to obtain a real-time evaluation result of the target client; the real-time evaluation result comprises a real-time comprehensive credit rating, a real-time purchasing behavior rating and/or a real-time credit limit.

The trust evaluation model application device provided by the embodiment belongs to the same application conception as the trust evaluation model application method provided by the embodiment of the application, can execute the method provided by any embodiment of the application, and has the corresponding functional module and beneficial effects of the execution method. Technical details not described in detail in this embodiment may refer to specific processing content of the trust evaluation model application method provided in the foregoing embodiment of the present application, and will not be described herein.

The functions implemented by the above acquisition module 401, the construction module 402, the training and evaluation module 403, the acquisition module 501, and the evaluation module 502 may be implemented in the form of invoking software by the same or different processors, respectively, and the embodiment of the present application is not limited.

The embodiment of the application also provides a trust evaluation system which comprises a trust evaluation model construction device and a trust evaluation model application device. The trust evaluation model constructing device is used for executing the trust evaluation model constructing method and sending the target trust evaluation model to the trust evaluation model application device, and the trust evaluation model application device is used for executing the trust evaluation model application method. The trust evaluation model construction device and the trust evaluation model application device may be located in the same equipment or may be located in different equipment.

Exemplary electronic device

Another embodiment of the present application also proposes an electronic device, referring to fig. 6, including: a memory 600 and a processor 610.

Wherein the memory 600 is connected to the processor 610, and is used for storing a program;

the processor 610 is configured to implement the trust evaluation model construction and application method disclosed in any one of the foregoing embodiments by running a program stored in the memory 600.

Specifically, the electronic device may further include: a bus, a communication interface 620, an input device 630, and an output device 640.

The processor 610, the memory 600, the communication interface 620, the input device 630, and the output device 640 are connected to each other by a bus. Wherein:

a bus may comprise a path that communicates information between components of a computer system.

The processor 610 may be a general-purpose processor, such as a general-purpose Central Processing Unit (CPU), microprocessor, etc., or may be an application-specific integrated circuit (ASIC), or one or more integrated circuits, for controlling the execution of programs in accordance with aspects of the present application. But may also be a Digital Signal Processor (DSP), application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components.

The processor 610 may include a main processor, and may also include a baseband chip, a modem, and the like.

The memory 600 stores programs for implementing the technical scheme of the present application, and may also store an operating system and other critical services. In particular, the program may include program code including computer-operating instructions. More specifically, memory 600 may include read-only memory (ROM), other types of static storage devices that may store static information and instructions, random access memory (random access memory, RAM), other types of dynamic storage devices that may store information and instructions, disk storage, flash, and the like.

The input device 630 may include means for receiving data and information entered by a user, such as a keyboard, mouse, camera, scanner, light pen, voice input means, touch screen, pedometer, or gravity sensor, among others.

Output device 640 may include means such as a display screen, printer, speakers, etc. that allow information to be output to a user.

The communication interface 620 may include devices using any transceiver or the like to communicate with other devices or communication networks, such as ethernet, radio Access Network (RAN), wireless Local Area Network (WLAN), etc.

The processor 610 executes programs stored in the memory 600 and invokes other devices that may be used to implement the various steps of any of the trust evaluation model construction and application methods provided by the above-described embodiments of the present application.

It will be appreciated by those skilled in the art that the structure shown in fig. 6 is merely a block diagram of a portion of the structure associated with the present inventive arrangements and is not limiting of the electronic device to which the present inventive arrangements are applied, and that a particular electronic device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.

The embodiment of the application also provides a chip which comprises a processor and a data interface, wherein the processor reads and runs a program stored in a memory through the data interface so as to execute the trust evaluation model construction and application method introduced in any embodiment, and the specific processing process and the beneficial effects thereof can be introduced by referring to the embodiment of the trust evaluation model construction and application method.

In addition to the methods and apparatus described above, embodiments of the present application provide a computer program product comprising computer program instructions which, when executed by a processor, cause the processor to perform the steps in the trust evaluation model construction and application method according to the various embodiments of the present application described in the "exemplary methods" section of this specification.

The computer program product may write program code for performing operations of embodiments of the present application in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server.

Furthermore, embodiments of the present application also propose a storage medium on which a computer program is stored, the computer program being executed by a processor to perform the steps in the trust evaluation model construction and application method according to the various embodiments of the present application described in the above section of the "exemplary method" of the present description.

It will be appreciated that the specific examples herein are intended only to assist those skilled in the art in better understanding the embodiments of the present description and are not intended to limit the scope of the present description.

It should be understood that, in various embodiments of the present disclosure, the sequence number of each process does not mean that the execution sequence of each process should be determined by the function and the internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure.

It will be appreciated that the various embodiments described in this specification may be implemented either alone or in combination, and are not limited in this regard.

Unless defined otherwise, all technical and scientific terms used in the embodiments of this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this specification belongs. The terminology used in the description is for the purpose of describing particular embodiments only and is not intended to limit the scope of the description. The term "and/or" as used in this specification includes any and all combinations of one or more of the associated listed items. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It will be appreciated that the processor of the embodiments of the present description may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method embodiments may be implemented by integrated logic circuits of hardware in a processor or instructions in software form. The processor may be a general purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), an off-the-shelf programmable gate array (Field Programmable gate array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components. The methods, steps and logic blocks disclosed in the embodiments of the present specification may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the embodiments of the present specification may be embodied directly in hardware, in a decoded processor, or in a combination of hardware and software modules in a decoded processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory, and the processor reads the information in the memory and, in combination with its hardware, performs the steps of the above method.

It will be appreciated that the memory in the embodiments of this specification may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory. The nonvolatile memory may be a read-only memory (ROM), a Programmable ROM (PROM), an Erasable Programmable ROM (EPROM), an Electrically Erasable Programmable ROM (EEPROM), or a flash memory, among others. The volatile memory may be Random Access Memory (RAM). It should be noted that the memory of the systems and methods described herein is intended to comprise, without being limited to, these and any other suitable types of memory.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present specification.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system, apparatus and unit may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.

In the several embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiment.

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

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present specification may be essentially or portions contributing to the prior art or portions of the technical solutions may be embodied in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present specification. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a read-only memory (ROM), a random-access memory (RAM), a magnetic disk, or an optical disk, etc.

The foregoing is merely specific embodiments of the present disclosure, but the scope of the disclosure is not limited thereto, and any person skilled in the art who is skilled in the art can easily think about variations or substitutions within the scope of the disclosure of the present disclosure, and it is intended to cover the variations or substitutions within the scope of the disclosure. Therefore, the protection scope of the present specification shall be subject to the protection scope of the claims.

Claims

1. The trust evaluation model construction method is characterized by comprising the following steps:

2. The trust evaluation model construction method of claim 1, wherein the obtaining a sample dataset comprises:

acquiring sample original data, and preprocessing the sample original data to obtain sample data; the sample data comprises basic data, asset data, credit data and behavior data of the target client;

Extracting the characteristics of the sample data to obtain characteristics to be selected;

screening the features to be selected to obtain target features; the target features include the base feature, the asset feature, the credit feature, and the behavioral feature;

generating the sample user representation based on the target feature;

and acquiring the sample label.

3. The method for constructing a trust evaluation model according to claim 1, wherein,

the preset credit rating model comprises a preset comprehensive credit rating model, a preset purchasing behavior rating model and/or a preset credit limit rating model;

the constructing a preset trust evaluation model based on the sample user portrait comprises the following steps:

constructing the preset comprehensive credit rating model based on characteristic items contained in the sample user portrait; the preset comprehensive credit rating model is used for evaluating the comprehensive credit rating of the target client based on the sample user portrait;

and/or the number of the groups of groups,

constructing the preset purchasing behavior scoring model based on characteristic items contained in the sample user portrait; the preset purchasing behavior scoring model is used for evaluating the purchasing behavior score of the target client based on the sample user portrait;

And/or the number of the groups of groups,

constructing a preset credit limit assessment model based on feature items contained in the sample user portrait, the preset comprehensive credit score model and the preset purchasing behavior score model; and the preset credit limit assessment model is used for assessing the credit limit of the sample client based on the sample user portrait.

4. The method for constructing a trust evaluation model according to claim 3, wherein,

the behavior characteristics comprise working behavior characteristic items;

the asset characteristics include an intrinsic asset characteristic item and an equipment asset characteristic item;

the credit standing feature comprises an internal credit standing feature item;

the constructing the preset comprehensive credit rating model based on the characteristic items contained in the sample user portrait comprises the following steps:

and constructing the preset comprehensive credit rating model based on the basic features, the inherent asset feature items, the equipment asset feature items, the internal credit feature items, the working behavior feature items and the first preset super parameters in the sample user portrait.

5. The method for constructing a trust evaluation model according to claim 3, wherein,

the behavior feature comprises a purchasing behavior feature item;

The constructing the preset purchasing behavior scoring model based on the feature items contained in the sample user portrait includes:

and constructing the preset purchasing behavior scoring model based on the purchasing behavior feature items in the sample user portrait and a second preset super parameter.

6. The method for constructing a trust evaluation model according to claim 3, wherein,

the asset characteristics include special asset characteristic items for characterizing whether the target client has a special asset;

the constructing the preset credit limit assessment model based on the feature items contained in the sample user portrait, the preset comprehensive credit scoring model and the preset purchasing behavior scoring model comprises the following steps:

and constructing the preset credit rating system assessment model based on the special asset characteristic item, the preset comprehensive credit rating model, the preset purchasing behavior rating model and a third preset super parameter.

7. The trust evaluation model construction method according to claim 6, wherein the constructing the trust quota evaluation model based on the special asset feature item, the preset comprehensive trust score model, the preset purchase behavior score model and a third preset super parameter comprises:

Constructing a two-dimensional matrix based on the preset comprehensive credit rating model and the preset purchasing behavior rating model;

and constructing the preset credit limit assessment model based on the third preset super parameter, the two-dimensional matrix and the special asset characteristic item.

8. A trust evaluation model application method, the method comprising:

9. The method for applying a trust evaluation model of claim 8,

the credit standing feature comprises an external credit standing feature item;

before the credit risk assessment is performed on the target client based on the real-time user portrait and target credit trust assessment model, the method further comprises:

determining whether the target client is out of trust based on the external credit feature item in the real-time user image;

If yes, halting the trust of the target client;

if not, continuing to trust the target client.

10. The method for applying a trust evaluation model of claim 8,

the target credit rating model comprises a target comprehensive credit rating model, a target purchasing behavior rating model and/or a target credit rating model;

the credit risk assessment is carried out on the target client based on the real-time user portrait and the target credit trust assessment model to obtain a real-time assessment result of the target client, and the method comprises the following steps:

and inputting the characteristic items in the real-time user image into the target comprehensive credit rating model, the target purchasing behavior rating model and/or the target credit rating model to obtain the real-time comprehensive credit rating, the real-time purchasing behavior rating and/or the real-time credit rating.

11. The trust evaluation model application method of claim 8, further comprising:

if the real-time comprehensive credit score is lower than a preset comprehensive credit score, halting credit to the target client; otherwise, continuing to trust the target client.

12. A trust evaluation model building apparatus, the apparatus comprising:

13. A trust evaluation model application apparatus, the apparatus comprising:

14. An electronic device, comprising: a processor and a memory;

wherein the memory is connected with the processor and is used for storing a computer program;

the processor is configured to implement the trust evaluation model construction method according to any one of claims 1 to 7 or the trust evaluation model application method according to any one of claims 8 to 11 by running a computer program stored in the memory.

15. A storage medium having stored thereon a computer program which, when executed by a processor, implements the trust evaluation model construction method of any one of claims 1-7 or the trust evaluation model application method of any one of claims 8-11.