CN112862602A - User request determining method, storage medium and electronic device - Google Patents
User request determining method, storage medium and electronic device Download PDFInfo
- Publication number
- CN112862602A CN112862602A CN202110336662.6A CN202110336662A CN112862602A CN 112862602 A CN112862602 A CN 112862602A CN 202110336662 A CN202110336662 A CN 202110336662A CN 112862602 A CN112862602 A CN 112862602A
- Authority
- CN
- China
- Prior art keywords
- user
- inventory
- incremental
- liability
- stock
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000004422 calculation algorithm Methods 0.000 claims description 40
- 239000011159 matrix material Substances 0.000 claims description 20
- 238000011156 evaluation Methods 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 13
- 230000006399 behavior Effects 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 8
- 230000014509 gene expression Effects 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000013523 data management Methods 0.000 abstract description 8
- 230000009466 transformation Effects 0.000 description 18
- 230000008569 process Effects 0.000 description 6
- 238000013210 evaluation model Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000007257 malfunction Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011002 quantification Methods 0.000 description 2
- 208000035126 Facies Diseases 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000012854 evaluation process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention provides a method for determining a user request, which is used for determining the security level of a user based on the determined common debt level score of the user and determining whether to pass the user request of the user based on the security level so as to avoid the condition that the data management system fails due to unreasonable passing of the user request.
Description
Technical Field
The present invention relates to the field of data management, and in particular, to a method for determining a user request, a storage medium, and an electronic device.
Background
With the ever-increasing volume of consumer credit markets, the opportunity to consume finance is self-evident, but currently the subject of various markets participates in which the greatest challenge is the risk of multiple commons. The same customer, facing more and more easy loan channels, is very vulnerable to the chain of funds and risk resistance when each institution gives credit. The financial institution's risk management capability determines the client's choice and trust, as well as how to perform risk pricing and client location, and even for the risks and challenges that may be encountered over long periods of time.
Therefore, based on the relevant information of the client, the machine learning algorithm is utilized to quantify the common debt degree of the client, which provides an important basis for how a bank reasonably trusts to maintain data security and normal operation of a trusting system. Currently, the evaluation of the common debt degree of the customers in financial industries such as banking industry and the like still stays in a more traditional mode, and generally only external data information of the customers (for example, loan information of the customers on other financial platforms) is combined, but the information of the customers is often lost in the process, the risk is used as a target for modeling, and finally the output result is discrete (0,1), so that the obtained result is not very accurate, and an inaccurate result can cause the data management system which grants the user request to malfunction or operate abnormally through the user request (for example, the loan request), so that a new more accurate and reasonable technical scheme is urgently needed to evaluate the common debt degree of the customers more accurately, so as to avoid the situation that the data management system malfunctions due to unreasonable user request.
Disclosure of Invention
The embodiment of the invention provides a method for determining a user request, a storage medium and an electronic device, which are used for at least solving the problem of data management system failure caused by unreasonable user request in the related art.
According to an embodiment of the present invention, there is provided a method for determining a user request, including:
when a user to be determined of the balance level is a stock user, obtaining a balance level score of the stock user based on user characteristics of the stock user and a stock balance model algorithm, wherein the stock user is a user after a preset performance period;
when a user to be determined with a common debt level is an incremental user, obtaining a common debt level score of the incremental user based on the user characteristics of the incremental user, the common debt level score evaluation result of an inventory user by the inventory common debt model algorithm and the incremental common debt model algorithm, wherein the common debt level score evaluation result of the inventory user by the inventory common debt model algorithm is used for adjusting the weight of the user characteristics of the incremental user, and the incremental user is a new user who has not passed through a preset expression period;
determining a security level of the user based on the determined liability level score of the user, and determining whether to pass the user request of the user based on the security level.
According to a further embodiment of the present invention, a computer-readable storage medium is also provided, in which a computer program is stored, wherein the computer program is configured to carry out the steps of any of the above-described method embodiments when executed.
According to yet another embodiment of the present invention, there is also provided an electronic device, including a memory and a processor, the memory having a computer program stored therein, the processor being configured to execute the computer program to implement the steps in any of the above method embodiments.
According to the embodiment of the invention, the safety level of the user is determined based on the determined common debt level score of the user, and whether the user request of the user passes or not is determined based on the safety level, so that the condition that the data management system fails due to unreasonable passing of the user request is avoided.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
FIG. 1 is a flow chart of a method of determining a user request according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a modeling process according to an exemplary embodiment of the present invention;
fig. 3 is a schematic diagram of weight adjustment according to an exemplary embodiment of the present invention.
Detailed Description
The invention will be described in detail hereinafter with reference to the accompanying drawings in conjunction with embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
Example 1
The method provided by the first embodiment of the present application may be executed in a mobile terminal, a computer terminal, a server, or a similar computing device. These devices have some structures that are known at present and will not be described again.
In the present embodiment, a method for determining a user request running on the computing device is provided, and fig. 1 is a flowchart of a method for determining a user request according to an embodiment of the present invention, as shown in fig. 1, the flowchart includes the following steps:
step S101, when the user to be determined the balance level is the stock user, the balance level score of the stock user is obtained based on the user characteristics of the stock user and the stock balance model algorithm, and the stock user is the user after the preset expression period;
when the user to be determined with the common debt level is an incremental user, obtaining the common debt level score of the incremental user based on the user characteristics of the incremental user, the common debt level score evaluation result of the inventory user by the inventory common debt model algorithm and the incremental common debt model algorithm, wherein the common debt level score evaluation result of the inventory user by the inventory common debt model algorithm is used for adjusting the weight of the user characteristics of the incremental user, and the incremental user is a new user who has not passed through a preset expression period;
step S103, determining the security level of the user based on the determined liability level score of the user, and determining whether to pass the user request of the user based on the security level.
According to the embodiment of the invention, the safety level of the user is determined based on the determined common debt level score of the user, and whether the user request of the user passes or not is determined based on the safety level, so that the condition that the data management system fails due to unreasonable passing of the user request is avoided. Wherein the security level may be determined according to the liability level score, if the liability level score indicates that the liability level of the user is high and the loan repayment capacity is low, the security level may be set to indicate that the user request of the user is not passed or the user request of the user is cautious passed, and the alarm may be given based on the user request. Vice versa, if the liability level score indicates that the user has a low liability level and a high loan repayment capacity, the security level may be set to indicate a user request by the user or a user request by the user more loosely, and a prompt may be made, for example, to prompt the user request.
In one example, the deriving the balance level score of the inventory consumer based on the consumer characteristics of the inventory consumer and an inventory balance model algorithm comprises:
acquiring the user characteristics of the stock users, and preprocessing the user characteristics of the stock users, wherein the user characteristics of each stock user correspond to a preprocessing mode;
based on the preprocessed user characteristics of the stock users, carrying out vector standardization to obtain a standardized matrix;
determining a liability level score for the inventory user based on the normalization matrix.
In one example, the determining the liability level score for the inventory user based on the standardized matrix comprises:
determining the distance between the stock user and the optimal scheme and the worst scheme by using a best-worst solution distance method according to the standardized matrix;
determining a liability level score for the inventory user based on the distance.
In one example, determining the distance between the inventory user and the optimal solution and the worst solution using a best solution method based on the normalization matrix comprises:
determining the distance between the inventory user and the optimal solution and the worst solution according to the following formula:
wherein i represents the ith sample client, j represents the jth index of the m indexes,indicating the distance of the client i from the optimal solution,indicates the distance, w, of the customer i from the worst casejIn order to obtain the weight of the index,is the value of the jth index in the optimal solution,is the value of the jth index in the worst case, zijIs the value of the jth index for client i;
determining a liability level score for the inventory user based on the distance, comprising:
determining a liability level score for the inventory consumer by the following formula:
wherein, CiRepresents the liability score of the customer i,indicating the distance of the client i from the optimal solution,indicating the distance of customer i from the worst case.
In one example, the index weight is determined by:
calculating the entropy value of each index by the following formula:
wherein x isijValue of j index representing i client, ejEntropy representing the jth index;
the index weight is obtained by the following formula:
wherein e isjEntropy, w, representing the j indexjThe index weight of the jth index is represented.
In one example, the obtaining the corporate level score of the incremental user based on the user characteristics of the incremental user, the evaluation result of the corporate level score of the inventory user by the inventory corporate model algorithm and the incremental corporate model algorithm comprises:
using the balance level score of the inventory user as a target variable of the incremental balance model algorithm;
adjusting the training sample weight of the incremental liability model using the liability level score of the inventory user;
and obtaining the liability level score of the incremental user by using the incremental liability model algorithm based on the user characteristics of the incremental user.
In one example, the method further comprises:
and verifying the balance level score of the stock user obtained by the stock balance model algorithm, and/or verifying the balance level score of the incremental user obtained by the incremental balance model algorithm.
In one example, the user characteristics of the inventory user include: the user attribute of the stock user and the behavior characteristic of the stock user; the user characteristics of the incremental user include: the user attributes of the incremental user.
Example embodiments
The following further explains the embodiment of the present invention with reference to a specific implementation scenario, and specifically explains the method for quantifying the risk of mutual debt of a client (also referred to as a user) provided in the embodiment by taking a bank scenario as an example, and it should be noted that the method provided in the embodiment is not only applicable to the bank scenario but also applicable to other financial institution scenarios, and details are not described below. The embodiment of the invention provides a method for quantifying risk of common debt of a client by combining unsupervised learning and supervised learning, which is used for risk identification control of a financial bank. The method provided by the embodiment adopts a supervised learning method in the pre-loan link and an unsupervised learning method in the middle loan link, and the two methods are combined to jointly quantify the common debt level of the customer. It should be added that "client" in this document, i.e., "user", is not described in detail below.
In one exemplary embodiment, an inventory liability model may be established for inventory customers of a bank. The new customers of the bank can become stock customers after a preset performance period, the customers can have more performance of the bank behavior in the performance period, relevant characteristic variables are extracted based on the attributes of the customers and the performance of the bank behavior, a comprehensive evaluation model is established through an unsupervised learning algorithm, and the common debt score is calculated for the common debt level of each customer. The attribute of the client may be basic information of the user, such as gender, age, industry occupation, education background, income, and the like, and may also include information of a credit inquiry result (e.g., a credit rating information) of the user.
In an exemplary embodiment, in the comprehensive evaluation process, expert experience is introduced to correct the model result, a small sample is extracted, the artificial judgment is carried out by front-end experts who give credit, examine and approve and the like according to experience, the judgment result of the experts and the result of the model are compared with each other, and the result of the stock common debt model is verified. That is, the result of scoring the liability model provided in the present embodiment may be compared with the result of manual determination for verification.
In an exemplary embodiment, for a newly added customer at a bank, an incremental liability model may be established; for the incremental common-bond model, a supervised learning algorithm can be used, the result of the stock common-bond model is provided for the incremental common-bond model training, the client common-bond scores obtained by the comprehensive evaluation model are classified into two categories and then serve as target variables of the pre-loan link, the sample weight of the incremental client model is adjusted based on the stock common-bond scores, and the incremental common-bond model is trained by using an xgboost algorithm.
The technical scheme adopted by the invention can greatly improve the identification capability of the risk of the common debt client, wherein the method combining unsupervised learning and supervised learning enables the quantitative result to be more objectively close to the actual common debt level of the client, thereby avoiding the unreasonable request of a data management system through a user.
The embodiment of the invention relates to a method for quantifying the risk of common debt of a client by combining unsupervised learning and supervised learning, which can be applied to the quantification of the common debt level of the client before, during and after loan and the like. Defining a grading threshold value of the common debt customer according to the distribution of the customer common debt grading, for example, directly giving rejection to the customer with high common debt level in credit strategies such as adjustment and management; otherwise, the customer is accepted, i.e., a low liability level is accepted.
In an exemplary embodiment, the process of establishing the model is mainly divided into the establishment of an inventory balance model (also referred to as "inventory client balance model") and the establishment of an incremental model (also referred to as "incremental client balance model") as shown in fig. 2.
1) The related index features of the stock clients (such as the attribute information of the client and the behavior features of the client at the bank in the above embodiment) are extracted, and particularly the behavior (i.e., the behavior features) of the client at the bank is extracted, so as to establish the feature variables of the stock balance model. The characteristic variables (i.e., "correlation index characteristics") are preprocessed by methods such as isotropic transformation (which may also be referred to as isotropic transformation), quantile transformation, and logarithmic transformation to construct a normalization matrix. The attribute information of the client may be basic information of the user, such as gender, age, industry occupation, education background, income, and the like, and may also include information of a credit inquiry result (e.g., a credit rating information) of the user. The behavior characteristics of the client at the bank can be transaction information, borrowing information, deposit information, amount information and the like, but not limited to the above, and the behavior characteristics can also comprise other characteristic indexes suitable for analyzing the common debt level of the client. It should be further noted that, the characteristic variables are preprocessed, one algorithm may be used to preprocess the characteristic variables, and multiple algorithms may also be used to preprocess one characteristic variable;
2) through an unsupervised learning method such as a good-poor solution method, an inventory client common debt degree score (also called as "inventory client common debt score") is obtained by establishing an inventory common debt model and considering the difference between a client and an optimal scheme and a worst scheme, wherein one client corresponds to one common debt degree score (also called as "common debt score").
3) And extracting the related attribute characteristics of the newly added client, adjusting the weight of the training sample by using the total debt score (also called as a "total debt degree score") of the stock client obtained by the total debt model of the stock client, and classifying the total debt score obtained by the total debt model of the stock client into two categories to be used as target variables of the newly added client.
4) And establishing an incremental common-liability model by using an xgboost algorithm based on the target variable and the training sample after the weight is adjusted, wherein the incremental common-liability model is used for quantifying the common-liability of the newly added client.
Wherein, the stock customer characteristic processing can use the following ideas:
basic information of the client, card information, credit inquiry data and the like are extracted, namely attribute information of the client is extracted. Since the scales and meanings of the indexes of the clients are different from each other, when the comprehensive evaluation method is used, the fact is that the distance scale is used for quantifying the difference between the clients. Using the distance scale, the index attribute is scaled.
Some of the data may use a homonymization transformation, which is applied to data in which the index value itself varies divergently from the liability level, such as the income index, and the higher the income index value, the higher the ability of the customer to pay back liabilities further indicates that the liability level of the customer may be smaller. On the other hand, for the liability index, the smaller the liability index value is, the less liability of the client is represented, so in order to unify the meaning of the index, we adopt the following processing for the anisotropic index such as income:
wherein x represents data before transformation (example)Such as revenue values), x' represents processed data (e.g., processed revenue values).
In addition, some information can be processed by using quantile transformation, for example, data distribution of part of indexes presents long tail characteristics, while in a model which is calculated by taking distance as a core, such as a comprehensive evaluation method, the distance scale is greatly influenced by an extreme value, the extreme value of part of customers can make the model unstable, and for the purpose of increasing robustness of the model and reducing influence of abnormal values on the model, quantile transformation is performed on the indexes presenting obvious long tail distribution of data, for example, 95 quantiles are adopted as maximum values, and abnormal data are removed (values greater than 95% of the scores are assigned as 95% quantiles of the variables):
x'=min(x,x0.95) Where x denotes data before transformation and x' denotes data after transformation.
In addition, some information can be processed by using logarithmic transformation, for example, the data distribution range of different indexes is different, for example, the dimensions of loan amount and loan number are different, and for this, logarithmic transformation is performed on indexes of amount and balance class (for example, various loan amounts and balances of customers), and the transformation can reduce the influence of different dimensions on the data as a whole to a certain extent:
x 'denotes log (1+ x), where x denotes data before transformation and x' denotes data after transformation.
Then, based on these transformed data, a normalized matrix is constructed, and a specific method is exemplified as follows:
let us have n customers, each customer has k indexes (i.e. the extracted basic information of the customer (including the attribute information of the customer and the behavior characteristics of the customer at the bank) is processed to obtain data, and one kind of data corresponds to one kind of index), then the original customer index matrix is:
the metric is vector normalized, i.e., each column element is divided by the L2 norm of the current column vector,
xija value representing the jth index of the ith customer,l2 norm, z representing the jth indexijAnd the j index of the ith client is represented as a value obtained by vector normalization transformation.
A normalized normalization matrix Z is thus obtained:
each client corresponds to one row in the matrix, for example, the first row is the index value of the first client, the nth row is the index value of the nth client, the row corresponds to the several clients, and the column corresponds to the several indexes.
In an exemplary embodiment, further, based on the normalized matrix Z obtained above, the distance between a certain customer and the optimal solution or the worst solution is obtained by using a best-and-worst solution distance method. The good-bad solution distance method is also called TOPSIS, and is a comprehensive evaluation method, which can fully utilize the information of original data and accurately reflect the difference between evaluation schemes. The basic process is based on the normalized homodromous feature matrix, an optimal scheme and a worst scheme in the limited schemes are found out by adopting a cosine method, then the distances between each evaluation object and the optimal scheme and the worst scheme are respectively calculated, the relative approach degree of each evaluation object and the optimal scheme is obtained and is used as a basis for evaluating the advantages and the disadvantages, and the specific algorithm flow is as follows:
inventory balance model algorithm flow:
procedure liability model (calculation Procedure):
carrying out syntropy transformation on the original index into X1;
scaling X1 to X2 within 0 to 95% quantile;
carrying out logarithmic transformation on the X2 to obtain X3;
constructing a normalized matrix Z after vector normalization, wherein the normalized matrix Z is { Z1, Z2, … …, zn };
each column Zi, do of For Z;
the ith dimension of the worst scheme Z-, the minimum value of Zi elements;
the ith dimension of the worst scheme Z +, and the maximum value of Zi elements;
End for
For zi∈Z do
End for
According to CiSorting the sizes of the components;
End procedure;
and outputting the comprehensive evaluation result of each client.
The normalized feature matrix has been described above as to how to determine the best solution and the worst solution, the best solution being composed of the maximum values of each column of elements in Z, which means the ideal case with the lowest degree of liabilities, which does not refer to a specific client, but the ideal case with the lowest degree of liabilities.
The worst case is composed of the minimum value of each column of elements in Z, which means the ideal case with the highest degree of common debt, and as with the best case, this case does not refer to a specific client, but the ideal case with the highest degree of common debt.
Calculating the distance between a certain customer and the optimal scheme and the worst scheme based on the index variable and the weight of the customer:
wherein i represents the ith sample client, j represents the jth index of the m indexes,indicating the distance of the client i from the optimal solution,indicates the distance, w, of the customer i from the worst casejThe weight of the j index determined by the entropy method and the like,is the value of the jth index in the optimal solution,is the value of the jth index in the worst case, zijIs the value of the jth index for client i;
and finally, calculating the comprehensive common debt degree of the client:
Cirepresents the liability score of the customer i,indicating the distance of the client i from the optimal solution,indicating the distance of customer i from the worst case.
CiThe degree of common debt of the client i is measured, and through the comprehensive evaluation method, the abstract degree of common debt is quantified, the quantification standard is not based on a certain fixed standard, but the various behavior characteristic information of the client is fully utilized, and the result can accurately reflect the facies between the clientsFor the liability level.
In an exemplary embodiment, determining the index weight based on the entropy weight method refers to: according to the basic principle of information theory, information is a measure of the degree of system order; while entropy is a measure of the degree of disorder of the system. The entropy weight method is an objective weighting method, and in a specific using process, the entropy weight method calculates the entropy weight of each index by using information entropy according to the variation degree of each index, and then corrects the weight of each index through the entropy weight, so that objective index weight is obtained.
Calculating the entropy value of each index:
xijvalue of j index representing i client, ejRepresenting the entropy value of the j-th index.
Further, we normalize the entropy values as weights for each index:
ejentropy, w, representing the j indexjRepresenting the entropy value of the j-th index after normalization.
wjThe larger the index is, the larger the amount of information represented by the index is, the stronger the ability to discriminate the degree of liability is.
In one exemplary embodiment, expert experience may be used to verify the effectiveness of the supervised learning approach. In order to verify the effectiveness of the supervised learning algorithm, an expert experience judgment method is introduced to correct the result of the comprehensive evaluation model. And (3) manually judging the extracted data according to experience by front-end experts such as credit granting, examination and approval and the like, for example, the experts label 1050 cases in total, wherein each case adopts a pairwise cross judgment mode. And adopting a mode of judging again for the case with large individual difference.
For the newly added customers, an incremental common bond model is established to evaluate the common bond level of the customers. Firstly, extracting relevant index characteristics (including attribute information of the client) of the newly added client, such as basic information (such as gender, age, industry occupation, education background, income and the like) and data of a pedestrian credit inquiry result and the like, and establishing a corresponding data set.
The method uses the xgboost as an increment common-liability model, the xgboost is an optimized distributed Gradient increment learning framework, the method realizes a machine learning algorithm under the Gradient Boosting idea, provides parallel tree promotion, and can quickly and accurately solve a plurality of data science problems. A highly extensible end-to-end lifting tree system is designed and constructed, a theoretically reasonable weighting quantile sketch map is provided for calculating a candidate set, a novel sparse perception algorithm is introduced for parallel tree learning, and missing values are enabled to have default directions.
xgboost is a linear addition model that sums the results of K trees as the final predictor, i.e.:
the first term to the right of the equality sign of the above equation is the loss function, where n is the number of training function samples, l is the loss to a single sample, assuming it is a convex function, yiThe true label value of the sample is trained for the mode,for the model to the predicted value of the training sample, ft(xi) A model representing the ith tree of the ith customer.
Second term Ω (f) of the above formulat) Is a regular term, controls the complexity of the tree, and prevents overfitting. The regularization term defines the complexity of the model:
wherein gamma and lambda are regularization parameters set manually, omega is a vector formed by all leaf node values of the decision tree, and T is the number of the leaf nodes.
The combination of the inventory and the incremental model, the expert experience, the segmentation of the model score of the inventory common-liability model, the judgment of whether the client is a common-liability client or not and the judgment of whether the client is a target variable of the incremental common-liability model, so that the incremental model of the newly added client can learn the characteristics of the inventory client with a large amount of behavioral expressions, and the predictability of the incremental common-liability model is improved.
In this embodiment, the target variable of the model is an index that the model intends to predict or evaluate, and in this embodiment, the target variable is: whether the customer is a corporate client.
The model is trained with input target variables, i.e., the model training phase, which needs to determine what the training samples are, the model using phase, which needs to determine what the model input is, and what the output result is. For example, in the training phase: the new clients become stock clients after a certain presentation period, and the characteristic indexes of the clients in the stock stage are selected as the input of the stock balance model; and selecting the characteristic indexes of the clients in the application stage as the input of the newly added common debt model. In the model use stage: and the characteristic index of the new client in application is used as the input of the newly added common debt model. In this embodiment, not only the model score of the stock liability model is used as the target variable of the incremental liability model, but also the model score is used to adjust the weight of the training sample of the incremental liability model, as shown in fig. 3:
for customers with inventory liability scores greater than the median, their sample weight x:
x=(d-d0.5)/(1-d0.5) D is the stock balance score of the sample, d0.5Is the median of the total incremental bond model sample inventory bond scores.
For customers with inventory liability scores less than the median, their sample weights are:
x=(d-d0.5)/d0.5d is the stock balance score of the sample, d0.5Is the median of the total incremental common debt model sample stock common debt score。
Because the goal of modeling is to identify customers with very high and very low liability levels, the sample weights are adjusted by the stock liability score. The inventory balance scores were high and low for customers, with the sample weights being greater. This allows samples with inventory liability scores that deviate more from the median or average to have more weight, which in turn allows the xgboost algorithm to more easily identify customers with very high and very low liability levels.
In the embodiment, an xgboost algorithm is used for establishing an incremental common-liability model and predicting the common-liability level of a newly-added client, so that an inventory common-liability model is established for inventory clients by using an unsupervised learning method, an incremental common-liability model is established for the newly-added client by using a supervised learning method, and an evaluation system for different stages of the whole life process of the client is established by combining supervision and unsupervised learning, and the robustness of the evaluation of the overall common-liability level can be improved more.
Through the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation mode in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.
Embodiments of the present invention also provide a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to implement the steps in any of the above method embodiments when executed. Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation manners, and this embodiment is not described herein again.
Optionally, in this embodiment, the storage medium may include, but is not limited to: various media capable of storing computer programs, such as a usb disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk.
Embodiments of the present invention also provide an electronic device comprising a memory having a computer program stored therein and a processor configured to execute the computer program to implement the steps of any of the above method embodiments.
Optionally, the electronic apparatus may further include a transmission device and an input/output device, wherein the transmission device is connected to the processor, and the input/output device is connected to the processor.
Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation manners, and this embodiment is not described herein again.
It will be apparent to those skilled in the art that the modules or steps of the present invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of multiple computing devices, and alternatively, they may be implemented by program code executable by a computing device, such that they may be stored in a storage device and executed by a computing device, and in some cases, the steps shown or described may be performed in an order different than that described herein, or they may be separately fabricated into individual integrated circuit modules, or multiple ones of them may be fabricated into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for determining a user request, comprising:
when a user to be determined of the balance level is a stock user, obtaining a balance level score of the stock user based on user characteristics of the stock user and a stock balance model algorithm, wherein the stock user is a user after a preset performance period;
when a user to be determined with a common debt level is an incremental user, obtaining a common debt level score of the incremental user based on the user characteristics of the incremental user, the common debt level score evaluation result of an inventory user by the inventory common debt model algorithm and the incremental common debt model algorithm, wherein the common debt level score evaluation result of the inventory user by the inventory common debt model algorithm is used for adjusting the weight of the user characteristics of the incremental user, and the incremental user is a new user who has not passed through a preset expression period;
determining a security level of the user based on the determined liability level score of the user, and determining whether to pass the user request of the user based on the security level.
2. The method of claim 1, wherein said deriving a balance level score for said inventory subscribers based on said subscriber characteristics of said inventory subscribers and an inventory balance model algorithm comprises:
acquiring the user characteristics of the stock users, and preprocessing the user characteristics of the stock users, wherein the user characteristics of each stock user correspond to a preprocessing mode;
based on the preprocessed user characteristics of the stock users, carrying out vector standardization to obtain a standardized matrix;
and determining the common debt level score of the inventory user according to the standardized matrix.
3. The method of claim 2, wherein said determining a liability level score for said inventory subscribers from said standardized matrix comprises:
determining the distances between the stock user and the optimal scheme and the worst scheme by using a best-worst solution distance method according to the standardized matrix;
and determining the common debt level score of the inventory user according to the distance.
4. The method of claim 3, wherein determining the distance between the inventory user and the optimal solution and the worst solution using a best-and-worst solution method according to the normalization matrix comprises:
determining the distance between the inventory user and the optimal solution and the worst solution according to the following formula:
wherein i represents the ith sample client, j represents the jth index of the m indexes,indicating the distance of the client i from the optimal solution,indicates the distance, w, of the customer i from the worst casejIn order to obtain the weight of the index,is the value of the jth index in the optimal solution,is the value of the jth index in the worst case, zijIs the value of the jth index for client i;
determining a liability level score for the inventory user based on the distance, comprising:
determining a liability level score for said inventory consumer by the formula:
5. The method of claim 4, wherein the metric weight is determined by:
calculating the entropy value of each index by the following formula:
wherein x isijValue of j index representing i client, ejEntropy representing the jth index;
the index weight is obtained by the following formula:
wherein e isjEntropy, w, representing the j indexjThe index weight of the jth index is represented.
6. The method for determining a user request according to any one of claims 1 to 5, wherein the obtaining of the liability level score of the incremental user based on the user characteristics of the incremental user, the assessment result of the liability level score of the inventory user by the inventory liability model algorithm and the incremental liability model algorithm comprises:
using the balance level score of the inventory user as a target variable of the incremental balance model algorithm;
adjusting training sample weights for the incremental liability model using the liability level scores of the inventory users;
and obtaining the liability level score of the incremental user by using the incremental liability model algorithm based on the user characteristics of the incremental user.
7. The method of determining a user request according to claim 1, further comprising:
and verifying the balance level score of the stock user obtained by the stock balance model algorithm, and/or verifying the balance level score of the incremental user obtained by the incremental balance model algorithm.
8. The method of claim 1, wherein the user characteristics of the inventory users comprise: the user attribute of the stock user and the behavior characteristic of the stock user; the user characteristics of the incremental user include: user attributes of the incremental user.
9. A computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to carry out the method of any one of claims 1 to 8 when executed.
10. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, and wherein the processor is arranged to execute the computer program to implement the method of any one of claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110336662.6A CN112862602B (en) | 2021-03-29 | 2021-03-29 | User request determining method, storage medium and electronic device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110336662.6A CN112862602B (en) | 2021-03-29 | 2021-03-29 | User request determining method, storage medium and electronic device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112862602A true CN112862602A (en) | 2021-05-28 |
CN112862602B CN112862602B (en) | 2024-05-24 |
Family
ID=75993132
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110336662.6A Active CN112862602B (en) | 2021-03-29 | 2021-03-29 | User request determining method, storage medium and electronic device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112862602B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108648074A (en) * | 2018-05-18 | 2018-10-12 | 深圳壹账通智能科技有限公司 | Loan valuation method, apparatus based on support vector machines and equipment |
CN109727123A (en) * | 2019-01-04 | 2019-05-07 | 平安科技(深圳)有限公司 | User's collage-credit data construction method, device and computer readable storage medium |
CN110135972A (en) * | 2019-04-23 | 2019-08-16 | 上海淇玥信息技术有限公司 | A kind of method, apparatus, system and recording medium for improving user and moving branch rate |
CN110728301A (en) * | 2019-09-09 | 2020-01-24 | 北京镭文科技有限公司 | Credit scoring method, device, terminal and storage medium for individual user |
-
2021
- 2021-03-29 CN CN202110336662.6A patent/CN112862602B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108648074A (en) * | 2018-05-18 | 2018-10-12 | 深圳壹账通智能科技有限公司 | Loan valuation method, apparatus based on support vector machines and equipment |
CN109727123A (en) * | 2019-01-04 | 2019-05-07 | 平安科技(深圳)有限公司 | User's collage-credit data construction method, device and computer readable storage medium |
CN110135972A (en) * | 2019-04-23 | 2019-08-16 | 上海淇玥信息技术有限公司 | A kind of method, apparatus, system and recording medium for improving user and moving branch rate |
CN110728301A (en) * | 2019-09-09 | 2020-01-24 | 北京镭文科技有限公司 | Credit scoring method, device, terminal and storage medium for individual user |
Also Published As
Publication number | Publication date |
---|---|
CN112862602B (en) | 2024-05-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kozodoi et al. | Fairness in credit scoring: Assessment, implementation and profit implications | |
Maldonado et al. | Integrated framework for profit-based feature selection and SVM classification in credit scoring | |
Verbraken et al. | Development and application of consumer credit scoring models using profit-based classification measures | |
US20180260891A1 (en) | Systems and methods for generating and using optimized ensemble models | |
US20220122171A1 (en) | Client server system for financial scoring with cash transactions | |
Bravo et al. | Granting and managing loans for micro-entrepreneurs: New developments and practical experiences | |
Callejón et al. | A System of Insolvency Prediction for industrial companies using a financial alternative model with neural networks | |
CN112102073A (en) | Credit risk control method and system, electronic device and readable storage medium | |
CN115205011B (en) | Bank user portrait model generation method based on LSF-FC algorithm | |
Eddy et al. | Credit scoring models: Techniques and issues | |
CN112836750A (en) | System resource allocation method, device and equipment | |
CN112232950A (en) | Loan risk assessment method and device, equipment and computer-readable storage medium | |
CN116468273A (en) | Customer risk identification method and device | |
CN116596659A (en) | Enterprise intelligent credit approval method, system and medium based on big data wind control | |
CN111061948A (en) | User label recommendation method and device, computer equipment and storage medium | |
CN113344692B (en) | Method for establishing network loan credit risk assessment model with multi-information-source fusion | |
CN117575773A (en) | Method, device, computer equipment and storage medium for determining service data | |
Ray et al. | Bayesian rapid optimal adaptive design (broad): Method and application distinguishing models of risky choice | |
CN113177733B (en) | Middle and small micro enterprise data modeling method and system based on convolutional neural network | |
CN112862602B (en) | User request determining method, storage medium and electronic device | |
CN116128339A (en) | Client credit evaluation method and device, storage medium and electronic equipment | |
CN115271442A (en) | Modeling method and system for evaluating enterprise growth based on natural language | |
CN114170000A (en) | Credit card user risk category identification method, device, computer equipment and medium | |
CN113052512A (en) | Risk prediction method and device and electronic equipment | |
US20230222378A1 (en) | Method and system for evaluating fairness of machine learning model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |