User rating model creation method, device and electronic device

Abstract

Disclosed are a user rating model creation method, a user rating model creation device, and an electronic device. The user rating model creation method includes: designing a basic model and a scene sub-model; creating a basic model and a scene sub-model that conform to the design based on model data; determining the weights of the basic model and the scene sub-model; and combining the basic model and the scene sub-model based on the weights to construct a user rating model, the user rating model being used to predict a user score from user data. In this way, by creating a combined user rating model by combining a general basic model and a subdivided sub-model under a scene with a predetermined weight, the development, update, and use effects of the user rating model are improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and more specifically, to a method for creating a user rating model, a device for creating a user rating model, and an electronic device. Background Art

[0002] With the development of information technology, data processing technology for user scoring based on user data has made great progress. For example, in the financial field, personal credit scoring technology has developed significantly. In addition, with the standardization of credit information and the access to more external data, the data has become richer and more reliable, which has also enabled user scoring models, such as user credit risk measurement models, to develop from single models to combined models with better accuracy, better robustness, and faster optimization and adjustment.

[0003] At present, in the process of modeling models that actually use user data for user rating, more research is on how to segment the model to improve the model's predictive performance, but this also brings corresponding problems, that is, the more segmented models there are, the smaller the sample size of each segmented model is, which is not conducive to ensuring the stability of the model. On the other hand, with the continuous introduction of external data sources, in the process of modeling using multi-source data, the main method is to group the data according to its dimensional characteristics and then establish sub-models separately, and re-model the modeling results of the sub-models as input variables. However, this will also be affected by the predictive performance and stability of the segmented model.

[0004] Therefore, it is desirable to provide an improved way of creating a user rating model. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a user rating model creation method, a user rating model creation device and an electronic device, which create a combined user rating model by combining a general basic model and a subdivided sub-model under a scenario with a predetermined weight, thereby improving the development, update and use of the user rating model.

[0006] According to one aspect of the present application, a method for creating a user rating model is provided, comprising: designing a base model and a scene sub-model; creating a base model and a scene sub-model that conform to the design based on model data; determining weights of the base model and the scene sub-model; and, combining the base model and the scene sub-model based on the weights to construct a user rating model, wherein the user rating model is used to predict a user score from user data.

[0007] In the above-mentioned user rating model creation method, the basic model includes one or more of a demographic information model, a credit investigation model, and a third-party information model; and the scenario sub-model includes one or more of a provident fund sub-model, a social security sub-model, a personal tax sub-model, and other channel sub-models.

[0008] In the above-mentioned user rating model creation method, creating a basic model and a scenario sub-model that conforms to the design based on the model data includes: dividing the model data into a training set and a test set according to a predetermined ratio; creating a basic model and a scenario sub-model that conforms to the design by WOE transformation and logistic regression, WOE is used to represent the difference between the proportion of good customers in the current group to all good customers and the proportion of bad customers in the current group to all bad customers; and, calibrating the created basic model and scenario sub-model.

[0009] In the above-mentioned user scoring model creation method, the scoring calibration of the created basic model and scene sub-model includes: scoring calibration of the created basic model and scene sub-model through three parameters: standard score, standard probability and the score increase required to double the probability.

[0010] In the above-mentioned user rating model creation method, determining the weights of the basic model and the scene sub-model includes: determining the initial weights of the basic model and the scene sub-model; constructing an effect evaluation objective function of the combined model composed of the basic model and the scene sub-model; and obtaining the weights of the basic model and the scene sub-model by finding the optimal solution to the effect evaluation objective function.

[0011] In the above-mentioned user rating model creation method, the effect evaluation objective function is the KS value.

[0012] In the above-mentioned user rating model creation method, it further includes: obtaining updated user data; obtaining an updated user rating from the user rating model based on the updated user data; calculating the difference between the updated user rating and the actual user rating based on the updated user data; and, in response to the difference being greater than a predetermined threshold, updating the user rating model.

[0013] In the above-mentioned user rating model creation method, updating the user rating model includes: performing deviation correction on the output score of the user rating model based on the actual user rating.

[0014] According to another aspect of the present application, a user rating model creation device is provided, comprising: a model design unit, for designing a basic model and a scene sub-model; a model creation unit, for creating a basic model and a scene sub-model that conform to the design based on model data; a weight determination unit, for determining the weights of the basic model and the scene sub-model; and a model combination unit, for combining the basic model and the scene sub-model based on the weights to construct a user rating model, wherein the user rating model is used to predict a user score from user data.

[0015] According to another aspect of the present application, an electronic device is provided, comprising: a processor; and a memory, in which computer program instructions are stored, and when the computer program instructions are executed by the processor, the processor executes the user rating model creation method as described above.

[0016] According to another aspect of the present application, a computer-readable medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor executes the user rating model creation method as described above.

[0017] The user rating model creation method, user rating model creation device and electronic device of the present application create a combined user rating model by combining a general basic model and a segmented sub-model under a scenario with predetermined weights. In a new business scenario, the general basic model can be first referred to, and then the segmented sub-model under the scenario can be added, thereby achieving rapid development of the model.

[0018] Furthermore, after completing the creation of the basic model and the segmented sub-model, the user rating model creation method, user rating model creation device and electronic device of the present application do not need to repeat the modeling. They can be put online for application only through optimal weighting. If the cooperation scenario or the data party is unexpectedly interrupted, the relevant sub-model can be removed without affecting other models, thereby facilitating the instant updating of the model.

[0019] In addition, the user rating model creation method, user rating model creation device and electronic device of the present application can ensure the performance of the model by weighting the general basic model and the sub-models under the scenario with optimal weights. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 A flowchart of a method for creating a user rating model according to an embodiment of the present application is shown.

[0022] Figure 2 A schematic diagram illustrating a specific design framework of a model in a method for creating a user rating model according to an embodiment of the present application is illustrated.

[0023] Figure 3 A schematic diagram of determining the optimal weight in the method for creating a user rating model according to an embodiment of the present application is illustrated.

[0024] Figure 4 The figure illustrates a schematic diagram of a model score correction process in a method for creating a user rating model according to an embodiment of the present application.

[0025] Figure 5 The flowchart of an exemplary application example of the method for creating a user rating model according to an embodiment of the present application is illustrated.

[0026] Figure 6 The figure illustrates a block diagram of a device for creating a user rating model according to an embodiment of the present application.

[0027] Figure 7 A block diagram of an electronic device according to an embodiment of the present application is illustrated. DETAILED DESCRIPTION

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] Exemplary Methods

[0030] Figure 1 The figure illustrates a flow chart of a method for creating a user rating model according to an embodiment of the present application.

[0031] like Figure 1 As shown, the method for creating a user rating model according to an embodiment of the present application includes the following steps.

[0032] Step S110, designing a basic model and a scenario sub-model. As described above, with the development of the business scenarios of user rating, more and more segmented scenarios are introduced, and in each segmented scenario, a corresponding segmented sub-model is developed.

[0033] Taking the loan business scenario of financial institutions as an example, with the development of the Internet business of financial institutions, more and more scenario data are introduced in the process of loan business. When the data volume is sufficient, the distinguishing ability of the model developed according to scenario segmentation is undoubtedly the best. However, there is often not enough scenario data support in the early stage of business launch, or even if there is, the data is not in line with the risk preferences of the financial institution.

[0034] Therefore, by adopting the user scoring model creation method according to the embodiment of the present application, a modeling method that is more suitable for the current financial scenario can be achieved by combining the general basic model and the sub-model under the segmented scenario to construct a combined model. That is, in the financial scenario, the design of the basic model starts from the perspective of the general model, and considers the development of the basic model on a large customer group with similar risk characteristics (such as: consumer customer group, business customer group). In addition, the basic model includes demographic information models, credit investigation models, and three-party data multi-head models. Usually, most of the data of financial institutions can meet the development of data-driven basic models. The scenario sub-model mainly considers the use of scenario data for model development. Based on the way financial institutions cooperate with the scenario party, the scenario sub-model may be the development method of expert cards or the development method of data-driven cards.

[0035] Specifically, the design of the basic model uses the idea of ​​a general model, and the basic model includes a population information model, a credit investigation model, a three-party data multi-head model, etc. Here, when designing the basic model, a certain degree of accuracy can be sacrificed on the basis of ensuring sufficient sample size and universal model. The basic model is mainly segmented from dimensions such as customer type (operation, consumption, etc.) and region, which can ensure the adequacy of the sample size of data-driven model development, the stability of the model, and the universality of future models in new businesses.

[0036] The design of the scenario sub-model utilizes the idea of ​​model segmentation, focusing on the characteristics of customer groups in different scenarios and scenario-specific data. Figure 2 A schematic diagram illustrating a specific design framework of a model in a method for creating a user rating model according to an embodiment of the present application is illustrated.

[0037] Therefore, in the user rating model creation method according to the embodiment of the present application, the basic model includes one or more of a demographic information model, a credit model, and a third-party information model; and the scenario sub-model includes one or more of a provident fund sub-model, a social security sub-model, a personal tax sub-model, and other channel sub-models.

[0038] Step S120, creating a basic model and a scene sub-model that conform to the design based on the model data. That is, after completing the design of the basic model and the scene sub-model, it is necessary to carry out actual development of the model based on the model data.

[0039] First, the data set is divided into a training set and a test set according to a certain ratio. That is, two types of samples are divided in different model data of the basic model and the scene sub-model, namely, development samples corresponding to the training set and verification samples corresponding to the test set.

[0040] Then, the basic model and scenario sub-model are developed. For example, in the loan business scenario of the financial institution as described above, the basic model and scenario sub-model are developed by WOE transformation and logistic regression. Here, the model candidate indicators are the indicators after WOE transformation after the original variables have been subdivided into sub-boxes and coarse bins (here, sub-boxes and coarse bins are common actions in the model development process. First, the variables are processed in sub-boxes according to certain rules (such as the method of equal proportion division, the data is divided into 20 equal parts), and then the overdue rate performance of each sub-box group is combined to form a coarse bin). WOE is the difference between "the proportion of good customers in the current group to all good customers" and "the proportion of bad customers in the current group to all bad customers", for example, expressed as:

[0041] f G (i) = GOOD%, f B (i) = BAD%

[0042]

[0043] Among them, f G (i) is the ratio of good customers to all good customers in the current group, f B (i) is the ratio of bad customers in the current group to all bad customers.

[0044] The results of the final model were fitted using logistic regression:

[0045]

[0046] Where g(x) is the sigmoid function, which converts the predicted value of the linear regression model into a probability output that conforms to the {0, 1} classification data. Here, θ T X is a linear function, and X is the variable after WOE conversion.

[0047] In addition, after the development of the basic model and scenario sub-model for user scoring is completed, scoring calibration is also required. Specifically, scoring calibration can set the parameters of standard score calibration, such as three parameters: standard score, standard probability (odds) and PDO (Points to Double the Odds: the score required to double the probability). The mapping process is as follows:

[0048]

[0049] Among them, the standard score of the score calibration scale is set to base_score, the standard probability is Odds0, In(Odds0) is the logarithm of Odds0, the intercept of the score calibration scale is A, and the slope is B; p(Y=Bad|X) represents the probability of a bad event when the value of X is known, and p(Y=Good|X) represents the probability of a good event when the value of X is known. Here, A and B represent the assumption that the calibrated score is in a linear relationship with In(Odds0), then A represents the intercept of the linear relationship, and B represents the slope.

[0050] Therefore, in the user scoring model creation method according to the embodiment of the present application, creating a basic model and a scenario sub-model that conform to the type based on the model data includes: dividing the model data into a training set and a test set according to a predetermined ratio; creating a basic model and a scenario sub-model that conform to the type through WOE transformation and logistic regression, WOE is used to represent the difference between the proportion of good customers in the current group to all good customers and the proportion of bad customers in the current group to all bad customers; and, calibrating the created basic model and scenario sub-model.

[0051] Furthermore, in the user scoring model creation method according to the embodiment of the present application, scoring calibration of the created basic model and scene sub-model includes: scoring calibration of the created basic model and scene sub-model through three parameters: standard score, standard probability and the score increase required to double the probability.

[0052] Step S130, determining the weights of the basic model and the scene sub-model. That is, when combining the basic model and the scene sub-model to create the final user rating model, the weights of the combined basic model and scene sub-model need to be determined.

[0053] Specifically, in the embodiment of the present application, the initial weights of the base model and the scene sub-model in the combined model are first set. For example, the initial weight is assigned to i=0.7, and the combined model is expressed as:

[0054] Score_comb=Score_main×i+Score_sub×(1-i)

[0055] Among them, Score_comb is the combination model, Score_main is the base model, and Score_sub is the sub-model.

[0056] Substituting the above defined A, B and odds into the above formula, we can get:

[0057] Score_comb = [AB × ln (odds main )]×i+[AB×ln(oddssub )]×(1-i)

[0058] Right now,

[0059] Then, an effect evaluation objective function of the combined model is constructed. For example, the effect evaluation objective function may be an indicator for testing the model, such as a KS value, which is defined as:

[0060]

[0061] Here, n represents the binning sequence number of the combined model score, CumB n CumG is the cumulative number of bad events in each bin. n The number of good events accumulated for each bin, Bad total is the total number of bad events, Good total is the total number of good events.

[0062] KS = max (cumulative event percentage - cumulative non-event percentage)

[0063] Next, the optimal weight is solved according to the objective function optimization:

[0064] CumB n =f(Score_comb bad ,i) n

[0065] CumG n =f(Score_comb good ,i) n

[0066]

[0067] The whole process is to find the optimal solution to the function. When KS reaches the optimal value, the optimal weight i can be obtained.

[0068] That is, in the embodiment of the present application, by finding the optimal solution to the KS function to obtain the optimal weight, the rapid development of the model can be achieved. That is, after completing the creation of the basic model and the sub-model, there is no need to repeat the modeling. Only by weighting with the optimal weight, the application can be put online, and if the cooperation scenario or data party is unexpectedly interrupted, the relevant sub-model can be removed without affecting other models, thereby promoting the instant update of the model. In addition, by calibrating the scores of the basic model and the scenario sub-model, the standard of the model scale can be guaranteed. Figure 3 A schematic diagram of determining the optimal weight in the method for creating a user rating model according to an embodiment of the present application is illustrated.

[0069] Therefore, in the method for creating a user scoring model according to an embodiment of the present application, determining the weights of the base model and the scenario sub-model includes: determining the initial weights of the base model and the scenario sub-model; constructing an effect evaluation objective function for the combined model formed by combining the base model and the scenario sub-model; and obtaining the weights of the base model and the scenario sub-model by finding the optimal solution of the effect evaluation objective function.

[0070] Moreover, in the above method for creating a user scoring model, the effect evaluation objective function is the KS value.

[0071] Step S140, combining the base model and the scenario sub-model based on the weights to construct a user scoring model, which is used to predict user scores from user data. That is, the final user scoring model is formed by combining the base model and the scenario sub-model. In this way, if a new product needs to be quickly launched, the base model part can be referred to first, and then the risk factors of the scenario module can be added in the way of expert experience, enabling rapid model development. Also, if the cooperation scenario or data provider is unexpectedly interrupted, the relevant sub-model can be removed without affecting other models.

[0072] In addition, in the actual application process, if there is a difference between the user score obtained through the user scoring model and the actual user score, it is also necessary to correct the predicted score of the model, that is, perform prediction probability calibration.

[0073] That is, after the scorecard scale transformation, a linear relationship between ln(Odds) and Score can be obtained (as described above, this is the premise assumption of score conversion). If there is a deviation in the default rates of the development sample and the recent sample, they are two intersecting straight lines and need to be corrected.

[0074]

[0075] Score = A - B * Ln(Odds) (2)

[0076] Where p is the proportion of the number of bad samples to the total number of samples; w is the coefficient of the model, that is, the target to be obtained; x is the development sample; Odds expect is the bad ratio of the development sample; Odds actual is the bad ratio of the actual sample; Odds calibrate is Odds actual compared with Odds expect ; x1 is the actual sample; and w1 is the model coefficient obtained after the actual sample is incorporated into the model.

[0077] Moreover, (1) is the calibration equation, and (2) is the relationship between Ln(Odds) and Score.

[0078] Figure 4 FIG. 1 is a schematic diagram of a model score correction process in a method for creating a user scoring model according to an embodiment of the present application. Figure 4 As shown in the figure, the dark solid dot lines are development sample data, and the light hollow dot lines are actual verification sample data. Since the development sample data during modeling is often a sampling sample, there will be some deviation in odds from the actual sample data used after the model is launched, so the error caused by the deviation between sample data should be corrected.

[0079] Therefore, in the user rating model creation method according to the embodiment of the present application, it further includes: obtaining updated user data; obtaining an updated user rating from the user rating model based on the updated user data; calculating the difference between the updated user rating and the actual user rating based on the updated user data; and, in response to the difference being greater than a predetermined threshold, updating the user rating model.

[0080] Furthermore, in the above-mentioned user rating model creation method, updating the user rating model includes: performing deviation correction on the output score of the user rating model based on the actual user rating.

[0081] Figure 5 FIG. 1 is a flowchart illustrating an exemplary application example of the method for creating a user rating model according to an embodiment of the present application. Figure 5 As shown in the figure, when designing a combined model, first design the basic model and sub-model, then develop the model and calibrate the scores for the designed basic model and sub-model respectively. Next, perform combined weighting, construct the objective function, and determine the model by finding the optimal weight. Finally, calibrate the model for prediction probability.

[0082] Exemplary Devices

[0083] Figure 6 The figure illustrates a block diagram of a device for creating a user rating model according to an embodiment of the present application.

[0084] like Figure 6 As shown, according to an embodiment of the present application, a user rating model creation device 200 includes: a model design unit 210, used to design a basic model and a scene sub-model; a model creation unit 220, used to create a basic model and a scene sub-model that conform to the design based on model data; a weight determination unit 230, used to determine the weights of the basic model and the scene sub-model; and, a model combination unit 240, used to combine the basic model and the scene sub-model based on the weights to construct a user rating model, and the user rating model is used to predict user scores from user data.

[0085] In one example, in the above-mentioned user rating model creation device 200, the basic model includes one or more of a demographic information model, a credit model, and a third-party information model; and the scenario sub-model includes one or more of a provident fund sub-model, a social security sub-model, a personal tax sub-model, and other channel sub-models.

[0086] In one example, in the above-mentioned user rating model creation device 200, the model creation unit 220 is used to: divide the model data into a training set and a test set according to a predetermined ratio; create a basic model and a scenario sub-model that conform to the design through WOE transformation and logistic regression, WOE is used to represent the difference between the proportion of good customers in the current group to all good customers and the proportion of bad customers in the current group to all bad customers; and, perform scoring calibration on the created basic model and scenario sub-model.

[0087] In one example, in the above-mentioned user scoring model creation device 200, the model creation unit 220 performs scoring calibration on the created basic model and scene sub-model, including: calibrating the created basic model and scene sub-model through three parameters: standard score, standard probability and the score increase required to double the probability.

[0088] In one example, in the above-mentioned user rating model creation device 200, the weight determination unit 230 is used to: determine the initial weights of the basic model and the scene sub-model; construct an effect evaluation objective function of the combined model composed of the basic model and the scene sub-model; and obtain the weights of the basic model and the scene sub-model by finding the optimal solution to the effect evaluation objective function.

[0089] In one example, in the above-mentioned user rating model creation device 200, the effect evaluation objective function is a KS value.

[0090] In one example, the user rating model creation device 200 further includes: a data acquisition unit for acquiring updated user data; a rating update unit for obtaining an updated user rating from the user rating model based on the updated user data; a difference calculation unit for calculating the difference between the updated user rating and the actual user rating based on the updated user data; and a model update unit for updating the user rating model in response to the difference being greater than a predetermined threshold.

[0091] In one example, in the above-mentioned user rating model creation device 200, the model updating unit is used to: perform deviation correction on the output score of the user rating model based on the actual user rating.

[0092] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the above-mentioned user rating model creation device 200 have been described in detail above. Figures 1 to 5 It has been described in detail in the description of the user rating model creation method, and therefore, its repeated description will be omitted.

[0093] As described above, the user rating model creation device 200 according to the embodiment of the present application can be implemented in various terminal devices, such as a server for loan risk control of a financial institution. In one example, the user rating model creation device 200 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the user rating model creation device 200 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the user rating model creation device 200 can also be one of the many hardware modules of the terminal device.

[0094] Alternatively, in another example, the user rating model creation apparatus 200 and the terminal device may also be separate devices, and the user rating model creation apparatus 200 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0095] Exemplary Electronic Devices

[0096] Below, reference Figure 7 To describe an electronic device according to an embodiment of the present application.

[0097] Figure 7 A block diagram of an electronic device according to an embodiment of the present application is illustrated.

[0098] like Figure 7 As shown, the electronic device 10 includes one or more processors 11 and a memory 12 .

[0099] The processor 11 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0100] The memory 12 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may run the program instructions to implement the user rating model creation method of each embodiment of the present application described above and / or other desired functions. Various contents such as basic models, scene sub-models, model weights, etc. may also be stored in the computer-readable storage medium.

[0101] In one example, the electronic device 10 may further include: an input device 13 and an output device 14, and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0102] The input device 13 may include, for example, a keyboard, a mouse, etc.

[0103] The output device 14 can output various information to the outside, including the generated user rating model, etc. The output device 14 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.

[0104] Of course, to simplify, Figure 7 Only some of the components related to the present application in the electronic device 10 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application situations, the electronic device 10 may also include any other appropriate components.

[0105] Exemplary computer program products and computer-readable storage media

[0106] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the user rating model creation method according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0107] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0108] In addition, an embodiment of the present application may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, causes the processor to execute the steps of the method for creating a user rating model according to various embodiments of the present application described in the above “Exemplary Method” section of this specification.

[0109] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0110] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0111] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0112] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0113] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0114] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A method for creating a user rating model, comprising: Design base models and scenario sub-models; Creating a base model and a scene sub-model that conform to the design based on the model data; Determining weights of the base model and the scene sub-model; as well as combining the base model and the scenario sub-model based on the weights to construct a user scoring model, the user scoring model being used to predict a user score from user data; The basic model uses the idea of ​​a general model, including one or more of a population information model, a credit investigation model, and a tripartite information model; and The scenario sub-model considers the characteristics of customer groups in different scenarios and scenario-specific data, including one or more of a provident fund sub-model, a social security sub-model, a personal tax sub-model, and other channel sub-models; Wherein, creating a basic model and a scene sub-model that conform to the design based on the model data includes: Divide the model data into training set and test set according to a predetermined ratio; Creating a basic model and scenario sub-model that conform to the design by means of WOE transformation and logistic regression, where WOE is used to represent the difference between the proportion of good customers in the current group to all good customers and the proportion of bad customers in the current group to all bad customers; and The created base model and scene sub-model are scored and calibrated.

2. The method for creating a user rating model according to claim 1, wherein: Scoring and calibrating the created base model and scene sub-model includes: The created base model and scenario sub-model are scored and calibrated using three parameters: standard score, standard likelihood, and the score increase required to double the likelihood.

3. The method for creating a user rating model according to claim 1, wherein: Determining the weights of the base model and the scene sub-model includes: Determining initial weights of the base model and the scene sub-model; Constructing an effect evaluation objective function of a combined model composed of the basic model and the scene sub-model; and The weights of the basic model and the scene sub-model are obtained by finding the optimal solution for the effect evaluation objective function.

4. The method for creating a user rating model according to claim 3, wherein: The effect evaluation objective function is the KS value.

5. The method for creating a user rating model according to claim 1, further comprising: Get updated user data; obtaining an updated user rating from the user rating model based on the updated user data; calculating a difference between the updated user score and an actual user score based on the updated user data; as well as In response to the difference being greater than a predetermined threshold, the user rating model is updated.

6. The method for creating a user rating model according to claim 5, wherein: Updating the user rating model includes: The output score of the user rating model is bias-corrected based on the actual user rating.

7. A user rating model creation device, comprising: Model design unit, used to design basic models and scene sub-models; A model creation unit, used to create a basic model and a scene sub-model that conform to the design based on the model data; A weight determination unit, used to determine the weights of the base model and the scene sub-model; as well as a model combination unit, configured to combine the basic model and the scenario sub-model based on the weight to construct a user scoring model, wherein the user scoring model is used to predict a user score from user data; The basic model uses the idea of ​​a general model, including one or more of a population information model, a credit investigation model, and a tripartite information model; and The scenario sub-model considers the characteristics of customer groups in different scenarios and scenario-specific data, including one or more of a provident fund sub-model, a social security sub-model, a personal tax sub-model, and other channel sub-models; wherein, creating a basic model and a scenario sub-model that conform to the design based on the model data includes: Divide the model data into training set and test set according to a predetermined ratio; Creating a basic model and scenario sub-model that conform to the design by means of WOE transformation and logistic regression, where WOE is used to represent the difference between the proportion of good customers in the current group to all good customers and the proportion of bad customers in the current group to all bad customers; and The created base model and scene sub-model are scored and calibrated.

8. An electronic device, comprising: processor; as well as A memory, wherein computer program instructions are stored in the memory, and when the computer program instructions are executed by the processor, the processor executes the user rating model creation method according to any one of claims 1 to 6.

Images