Target object determination method and device, storage medium and electronic equipment

Abstract

The invention discloses a target object determination method and device, a storage medium and electronic equipment, and relates to the field of financial science and technology. The method comprises the following steps: acquiring feature data of a plurality of to-be-predicted objects of at least one to-be-predicted credit card product; the feature data of the multiple to-be-predicted objects of each credit card product are input into a regression model corresponding to each credit card product for prediction, prediction mark distribution corresponding to each credit card product is obtained, and the prediction mark distribution is probability distribution representing the active state of the to-be-predicted objects to the credit card products; and determining a target object of each credit card product from the plurality of to-be-predicted objects of each credit card product according to the prediction mark distribution. According to the method and the device, the technical problem that the accuracy is relatively low when inactive customers of credit card products are determined in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and specifically, to a method, device, storage medium and electronic device for determining a target object. Background Art

[0002] Many credit card customers rarely use their cards for a long time after opening them. They rarely make credit card transactions or update their information within a certain period of time. In the long run, banks will convert these credit card customers into inactive accounts, thereby losing a group of customers who may have a tendency to use their cards. Therefore, it is necessary to predict inactive accounts. At present, the accuracy of the existing technology in determining inactive customers of credit card products is relatively low.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The main purpose of the present application is to provide a method, device, storage medium and electronic device for determining a target object, so as to at least solve the technical problem of low accuracy in determining inactive customers of credit card products in the prior art.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for determining a target object is provided, including: obtaining feature data of multiple objects to be predicted of at least one credit card product to be predicted; inputting the feature data of multiple objects to be predicted of each credit card product into a regression model corresponding to each credit card product for prediction, and obtaining a prediction mark distribution corresponding to each credit card product, wherein the prediction mark distribution is a probability distribution representing the active state of the object to be predicted for the credit card product; and determining the target object of each credit card product from the multiple objects to be predicted of each credit card product based on the prediction mark distribution.

[0006] Furthermore, based on the prediction mark distribution, the target object of each credit card product is determined from the multiple objects to be predicted of each credit card product, including: comparing the probability value corresponding to each object to be predicted in the prediction mark distribution with a preset threshold; and taking the object to be predicted whose probability value is greater than the preset threshold among the multiple objects to be predicted as the target object.

[0007] Furthermore, the regression model corresponding to each credit card product is obtained through the following steps: the historical card usage data corresponding to each credit card product is obtained, and the historical card usage data is vectorized to obtain the card usage feature vector corresponding to each credit card product; a discrete Gaussian distribution is constructed based on the card usage feature vector corresponding to each credit card product to obtain the label distribution corresponding to each credit card product; based on the label distribution corresponding to each credit card product, an independent variable matrix and a dependent variable matrix are constructed, and based on the independent variable matrix and the dependent variable matrix, an initial regression model is established using partial least squares method, wherein the independent variable matrix is ​​used to represent the feature information of all credit card products, and the dependent variable matrix is ​​used to represent the activity information of the sample object for each credit card product; based on the initial regression model, an adaptive standard deviation adjustment is performed on each credit card product to obtain a regression model corresponding to each credit card product.

[0008] Furthermore, based on the independent variable matrix and the dependent variable matrix, the partial least squares method is used to establish an initial regression model, including: constructing a linear dependency relationship between the independent variable matrix and the dependent variable matrix; based on the linear dependency relationship, extracting multiple pairs of latent variables associated with the independent variables and the dependent variables, and maximizing the covariance between the latent variables to obtain the initial regression model.

[0009] Furthermore, based on the initial regression model, an adaptive standard deviation adjustment is performed on each credit card product to obtain a regression model corresponding to each credit card product, including: configuring a first standard deviation for each credit card product, and determining an initial tag distribution corresponding to each credit card product based on the first standard deviation and the tag distribution corresponding to each credit card product; determining a predicted tag distribution corresponding to each credit card product based on the initial regression model, and determining a target standard deviation corresponding to each credit card product based on the initial tag distribution corresponding to each credit card product and the predicted tag distribution corresponding to each credit card product; updating the initial regression model based on the target standard deviation corresponding to each credit card product to obtain a regression model corresponding to each credit card product.

[0010] Furthermore, based on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product, the target standard deviation corresponding to each credit card product is determined, including: performing similarity calculation on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product to obtain a calculation result; optimizing the first standard deviation with the goal of minimizing the calculation result to obtain the target standard deviation corresponding to each credit card product.

[0011] Furthermore, after determining the target object of each credit card product from the multiple objects to be predicted of each credit card product, the method further includes: sending prompt information to the target object of each credit card product respectively, wherein the prompt information is used to remind the target object to use the card.

[0012] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a device for determining a target object is also provided, including: a first acquisition unit, used to acquire feature data of multiple objects to be predicted of at least one credit card product to be predicted; a first processing unit, used to input the feature data of multiple objects to be predicted of each credit card product into a regression model corresponding to each credit card product for prediction, and obtain a prediction mark distribution corresponding to each credit card product, wherein the prediction mark distribution is a probability distribution representing the active state of the object to be predicted for the credit card product; a determination unit, used to determine the target object of each credit card product from the multiple objects to be predicted of each credit card product based on the prediction mark distribution.

[0013] Furthermore, the determination unit includes: a comparison subunit, used to compare the probability value corresponding to each object to be predicted in the prediction mark distribution with a preset threshold; a determination subunit, used to select an object to be predicted whose probability value is greater than the preset threshold among multiple objects to be predicted as a target object.

[0014] Furthermore, the device also includes the following units, which are used to obtain the regression model corresponding to each credit card product through the following steps: a second acquisition unit, which is used to obtain the historical card usage data corresponding to each credit card product, and vectorize the historical card usage data to obtain the card usage feature vector corresponding to each credit card product; a second processing unit, which is used to construct a discrete Gaussian distribution based on the card usage feature vector corresponding to each credit card product, and obtain the label distribution corresponding to each credit card product; a third processing unit, which is used to construct an independent variable matrix and a dependent variable matrix based on the label distribution corresponding to each credit card product, and establish an initial regression model based on the independent variable matrix and the dependent variable matrix using partial least squares method, wherein the independent variable matrix is ​​used to represent the feature information of all credit card products, and the dependent variable matrix is ​​used to represent the activity information of the sample object for each credit card product; a fourth processing unit, which is used to perform adaptive standard deviation adjustment on each credit card product based on the initial regression model to obtain the regression model corresponding to each credit card product.

[0015] Furthermore, the third processing unit includes: a first processing sub-unit, used to construct a linear dependency relationship between the independent variable matrix and the dependent variable matrix; a second processing sub-unit, used to extract multiple pairs of latent variables associated with the independent variables and the dependent variables based on the linear dependency relationship, and maximize the covariance between the latent variables to obtain an initial regression model.

[0016] Furthermore, the fourth processing unit includes: a third processing sub-unit, used to configure a first standard deviation for each credit card product, and determine an initial mark distribution corresponding to each credit card product based on the first standard deviation and the mark distribution corresponding to each credit card product; a fourth processing sub-unit, used to determine a predicted mark distribution corresponding to each credit card product based on an initial regression model, and determine a target standard deviation corresponding to each credit card product based on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product; a fifth processing sub-unit, used to update the initial regression model based on the target standard deviation corresponding to each credit card product to obtain a regression model corresponding to each credit card product.

[0017] Furthermore, the fourth processing sub-unit includes: a calculation module, which is used to perform similarity calculation on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product to obtain a calculation result; a processing module, which is used to optimize the first standard deviation with the goal of minimizing the calculation result to obtain a target standard deviation corresponding to each credit card product.

[0018] Furthermore, the device also includes: a sending unit, which is used to send prompt information to the target object of each credit card product after determining the target object of each credit card product from multiple objects to be predicted of each credit card product, wherein the prompt information is used to remind the target object to use the card.

[0019] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a computer program product is further provided, including a computer program, and the computer program implements the above-mentioned method for determining the target object when executed by a processor.

[0020] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a computer-readable storage medium is further provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned method for determining the target object when running.

[0021] In order to achieve the above-mentioned purpose, according to another aspect of the present application, an electronic device is also provided, which includes one or more processors; a memory for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors are implemented to run the programs, wherein the programs are configured to execute the above-mentioned method for determining the target object when running.

[0022] In the embodiment of the present application, a label distribution learning algorithm based on multiple multivariate regression is used for modeling. First, feature data of multiple objects to be predicted of at least one credit card product to be predicted is obtained. Then, the feature data of multiple objects to be predicted of each credit card product is input into the regression model corresponding to each credit card product for prediction, and the prediction label distribution corresponding to each credit card product is obtained, wherein the prediction label distribution is a probability distribution representing the active state of the object to be predicted for the credit card product. Then, based on the prediction label distribution, the target object of each credit card product is determined from the multiple objects to be predicted of each credit card product. By establishing an adaptive inactive customer prediction model for each credit card product, it is possible to screen out inactive customer groups for each credit card product, accurately locate customers who may become inactive credit card customers in the future, and achieve the purpose of accurately awakening inactive credit card customers, thereby achieving the technical effect of improving accuracy, and thus solving the technical problem of relatively low accuracy in determining inactive customers of credit card products in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0024] Figure 1 is a flowchart of an optional method for determining a target object according to an embodiment of the present application;

[0025] Figure 2 is a schematic diagram of an optional marker distribution according to an embodiment of the present application;

[0026] Figure 3 is a schematic diagram of an optional device for determining a target object according to an embodiment of the present application;

[0027] Figure 4 It is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] It should be noted that the collected information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data are in compliance with relevant laws, regulations and standards, necessary confidentiality measures are taken, and public order and good customs are not violated, and corresponding operation entrances are provided for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions to provide users with corresponding operation entrances for users to choose to agree or refuse the results of automated decision-making; if the user chooses to refuse, the expert decision-making process will be entered.

[0031] Example 1

[0032] According to an embodiment of the present application, an embodiment of a method for determining a target object is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0033] Figure 1 is a flowchart of an optional method for determining a target object according to an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0034] Step S101, acquiring feature data of multiple objects to be predicted of at least one credit card product to be predicted.

[0035] In the above steps, the application system, processor, electronic device and other devices can be used as the execution subject, and optionally, the target object determination system can be used as the execution subject. First, the feature data of multiple objects to be predicted of at least one credit card product to be predicted are obtained, wherein the feature data can be the card opening date, card level, card limit, cardholder credit level, credit card consumption amount, credit card consumption times, credit card installment period, credit card installment amount, etc. Optionally, the feature data is pre-processed by standardization or centralization to ensure that the data format is consistent with that during model training, so as to facilitate direct application of the model.

[0036] Step S102, input the feature data of multiple objects to be predicted of each credit card product into the regression model corresponding to each credit card product for prediction, and obtain the prediction mark distribution corresponding to each credit card product, wherein the prediction mark distribution is the probability distribution representing the active state of the object to be predicted for the credit card product.

[0037] Step S103, determining the target object of each credit card product from the multiple objects to be predicted of each credit card product according to the distribution of the prediction tags.

[0038] Optionally, the characteristic data of multiple objects to be predicted for each credit card product are input into the regression model corresponding to each credit card product for prediction. The model can calculate the customer's activity prediction for the credit card product based on the linear relationship between the regression coefficient and the independent variable and the dependent variable, that is, obtain the prediction mark distribution corresponding to each credit card product, and then determine the target object of each credit card product from the multiple objects to be predicted for each credit card product based on the prediction mark distribution. That is, based on this distribution, it can be identified which customer is most likely to become an inactive account for each credit card product, that is, find the mark with the lowest descriptiveness, so as to help banks accurately identify which customers have a higher risk of inactivity, and then adopt corresponding awakening strategies to improve the activity of these customers.

[0039] Based on the scheme defined in the above steps S101 to S103, it can be known that in the embodiment of the present application, the label distribution learning algorithm based on multiple multivariate regression is used for modeling. First, the feature data of multiple objects to be predicted of at least one credit card product to be predicted is obtained, and then the feature data of multiple objects to be predicted of each credit card product is input into the regression model corresponding to each credit card product for prediction, and the prediction label distribution corresponding to each credit card product is obtained, wherein the prediction label distribution is a probability distribution representing the active state of the object to be predicted for the credit card product, and then the target object of each credit card product is determined from the multiple objects to be predicted of each credit card product according to the prediction label distribution. By establishing an adaptive inactive customer prediction model for each credit card product, it is possible to screen out inactive customer groups for each credit card product, accurately locate customers who may become inactive credit card customers in the future, and achieve the purpose of accurately awakening inactive credit card customers, thereby achieving the technical effect of improving accuracy, and thus solving the technical problem of relatively low accuracy in determining inactive customers of credit card products in the prior art.

[0040] In an optional embodiment, based on the prediction mark distribution, the target object of each credit card product is determined from multiple objects to be predicted for each credit card product, including: comparing the probability value corresponding to each object to be predicted in the prediction mark distribution with a preset threshold; and taking the object to be predicted whose probability value is greater than the preset threshold among the multiple objects to be predicted as the target object.

[0041] Optionally, a threshold is set to determine whether a customer is marked as an inactive customer. For example, if the predicted probability is less than the threshold, the customer is marked as active; if it is greater than the threshold, the customer is marked as an inactive customer. Therefore, the probability value corresponding to each object to be predicted in the predicted label distribution is compared with the preset threshold, and the object to be predicted with a probability value greater than the preset threshold among multiple objects to be predicted is taken as the target object.

[0042] It should be noted that by comparing the probability value corresponding to each object to be predicted in the prediction tag distribution with the preset threshold, it is possible to screen out inactive customer groups for each credit card product and accurately locate customers who may become inactive credit card holders in the future.

[0043] In an optional embodiment, the regression model corresponding to each credit card product is obtained by the following steps: obtaining the historical card usage data corresponding to each credit card product, and vectorizing the historical card usage data to obtain the card usage feature vector corresponding to each credit card product; constructing a discrete Gaussian distribution based on the card usage feature vector corresponding to each credit card product to obtain the label distribution corresponding to each credit card product; constructing an independent variable matrix and a dependent variable matrix based on the label distribution corresponding to each credit card product, and using partial least squares method to establish an initial regression model based on the independent variable matrix and the dependent variable matrix, wherein the independent variable matrix is ​​used to represent the feature information of all credit card products, and the dependent variable matrix is ​​used to represent the activity information of the sample object for each credit card product; based on the initial regression model, adaptively adjusting the standard deviation of each credit card product to obtain the regression model corresponding to each credit card product.

[0044] Optionally, constructing a regression model corresponding to each credit card product mainly includes feature extraction, label distribution, multiple multivariate regression learning, and adaptive adjustment model. First, feature extraction is performed to obtain the historical card usage data corresponding to each credit card product, and the historical card usage data is vectorized to obtain the card usage feature vector corresponding to each credit card product. Specifically, the historical card usage data is obtained from the credit card products and cardholder card usage data in the data lake and vectorized, and the card usage feature values ​​of all cardholders under the same credit card product (for example, card opening date, card level, card limit, cardholder credit level, credit card consumption amount, credit card consumption times, credit card installment period, credit card installment amount, etc.) and the feature values ​​of all credit card products (for example, card issuance time, card issuance area, card fee, etc.) are uniformly extracted to form a one-to-many association relationship between credit card products and customers.

[0045] Furthermore, a discrete Gaussian distribution is constructed based on the card usage feature vector corresponding to each credit card product to obtain the label distribution corresponding to each credit card product. Specifically, the credit card product is regarded as an instance x, and the cardholder's card usage feature is regarded as a label w. Each possible label covered by instance x is represented as w. In the label distribution of instance x, a real number y is assigned to each label w. x,w , where y x,w ∈[0,1], which represents the descriptiveness of the label w for the instance x. Add the descriptiveness of all possible labels w for the instance x, assuming that the sum must be equal to 1, that is, ∑ w y x,w =1, which means that all tags in the tag set of instance x can fully describe instance x. Select discrete Gaussian distribution N(α, σ 2 ) is used as the labeled distribution of credit card product x, where the mean is the cardholder’s card usage feature α and the standard deviation is σ. The formula is as follows:

[0046]

[0047] Among them, the normalization factor Z is expressed as follows:

[0048]

[0049] Figure 2 is a schematic diagram of an optional marker distribution according to an embodiment of the present application, such as Figure 2 As shown, the discrete Gaussian distribution of the credit card product cardholder's card usage feature α is displayed.

[0050] Furthermore, multiple multivariate regression learning is performed, that is, based on the label distribution corresponding to each credit card product, an independent variable matrix and a dependent variable matrix are constructed, and based on the independent variable matrix and the dependent variable matrix, an initial regression model (i.e., a unified customer activity prediction model) is established using partial least squares method.

[0051] In an optional embodiment, an initial regression model is established based on the independent variable matrix and the dependent variable matrix using partial least squares method, including: constructing a linear dependency relationship between the independent variable matrix and the dependent variable matrix; based on the linear dependency relationship, extracting multiple pairs of latent variables associated with the independent variables and the dependent variables, and maximizing the covariance between the latent variables to obtain an initial regression model.

[0052] Optionally, first construct a linear dependency relationship between the independent variable matrix and the dependent variable matrix, then extract multiple pairs of latent variables based on the association between the independent variables and the dependent variables based on the linear dependency relationship, and maximize the covariance between the latent variables to obtain the initial regression model. Specifically, assume that the feature space of all credit card products is represented by X∈R n×d , the finite set of all credit card customers is represented by w = {w1,w2,…,w c}. Given a training set S = {(x1,α1),(x2,α2),…,(x n ,α n )}, where the feature vector of the i-th credit card product is represented by x i ∈X,x i The customer is denoted as α i ∈w.x i The corresponding generated label distribution y i =N(α i ,σ 2 ) is as follows:

[0053]

[0054] It is known that the training set S is composed of n groups of d-dimensional independent variables and n groups of c-dimensional dependent variables. The labeled distribution learning problem is transformed into a multiple multiple regression analysis problem, that is, a linear dependency relationship between multiple groups of independent variables and dependent variables is established, and then a multi-dependent variable partial least squares (PLS) method is used to establish a unified customer activity prediction model for all credit card products, that is, a credit card inactive customer prediction model. The independent variable matrix X and the dependent variable matrix Y are expressed as follows:

[0055]

[0056] When PLS is used to establish a multiple linear regression model, it is necessary to extract several pairs of latent variables (latent components) based on the association between the independent variable and the dependent variable. When establishing a multiple multiple regression model, PLS is required to maximize the covariance between latent variables. Assuming there are n groups of samples, taking the first pair of principal component axis vectors a1 and b1 of X and Y as an example, the optimization goal of PLS ​​can be formulated as:

[0057]

[0058] Among them, a1 and b1 are the symmetric matrices X T YY T X and Y T XX T The unit eigenvector corresponding to the maximum eigenvalue of Y can be solved by Lagrange multipliers respectively.

[0059] According to the linear transformation, the first set of latent variables u1 and v1 of X and Y are extracted:

[0060] u1=Xa1,v1=Yb1 (6)

[0061] According to u1 and v1, the regression equations of X and Y on the latent variables are established:

[0062]

[0063] Based on the correlation between u1 and v1, Y is transformed into a regression equation for the latent variable u1 of X:

[0064]

[0065] The load vector E1, F1 and G1 are the residual matrices of the corresponding regression equations.

[0066] Consider E1 as X and F1 as Y, extract the second pair of latent variables u2 and v2, and establish a new regression equation:

[0067]

[0068] Repeat the above steps until m pairs of latent variables are extracted:

[0069]

[0070] Expressing formula (10) in matrix form, the final regression equation of Y with respect to X is:

[0071] Y=UR T +F=XAR T +F=Xβ+F (11)

[0072] Among them, the regression coefficient β = AR T , the residual matrix is ​​F.

[0073] After PLS establishes the regression model and obtains the regression coefficient β, i Output a predicted label distribution The mark corresponding to the minimum description degree is x i Least active customers

[0074]

[0075] It should be noted that the PLS regression model can consider the potential connections between multiple dependent variables and overcome the multiple correlations of variables. When the sample data is less than the number of variables, PLS can better ensure the accuracy of the regression model.

[0076] Furthermore, an adaptive adjustment model is performed, that is, based on the initial regression model, an adaptive standard deviation adjustment is performed on each credit card product to obtain a regression model corresponding to each credit card product.

[0077] In an optional embodiment, based on the initial regression model, an adaptive standard deviation adjustment is performed on each credit card product to obtain a regression model corresponding to each credit card product, including: configuring a first standard deviation for each credit card product, and determining an initial tag distribution corresponding to each credit card product based on the first standard deviation and the tag distribution corresponding to each credit card product; determining a predicted tag distribution corresponding to each credit card product based on the initial regression model, and determining a target standard deviation corresponding to each credit card product based on the initial tag distribution corresponding to each credit card product and the predicted tag distribution corresponding to each credit card product; updating the initial regression model based on the target standard deviation corresponding to each credit card product to obtain a regression model corresponding to each credit card product.

[0078] The PLS regression model is a unified prediction model for credit card inactive customers based on all credit card products. Since different credit card products have different least active customers, an adaptive model adjustment is performed on each credit card product to find the least active customers for each credit card product, thereby achieving accurate awakening. The goal of the adaptive adjustment model is: the standard deviation σ of the discrete Gaussian distribution in formula (3) must be adaptive to the least active customers of each credit card product, that is, a different standard deviation is assigned to each credit card product.

[0079] In order to achieve this goal, the label distribution of all credit card products is first initialized to a discrete Gaussian distribution with an initial standard deviation, that is, first configure the first standard deviation (i.e., the initial standard deviation) for each credit card product, and determine the initial label distribution corresponding to each credit card product based on the first standard deviation and the label distribution corresponding to each credit card product. Then, a regression model is established using partial least squares to obtain the predicted label distribution of all credit card products, that is, the predicted label distribution corresponding to each credit card product is determined based on the initial regression model. Then, an "excellent" predicted label distribution is selected to adaptively learn the standard deviation of each credit card product, and then the label distribution of all credit card products is updated according to the standard deviation to establish a new regression model. Finally, the coefficients and standard deviations of the regression model are alternately iterated and updated mutually until the algorithm converges (for example, reaches a preset number of iterations), that is, the target standard deviation corresponding to each credit card product is determined based on the initial label distribution corresponding to each credit card product and the predicted label distribution corresponding to each credit card product, and the initial regression model is updated according to the target standard deviation corresponding to each credit card product to obtain the regression model corresponding to each credit card product.

[0080] In an optional embodiment, a target standard deviation corresponding to each credit card product is determined based on an initial mark distribution corresponding to each credit card product and a predicted mark distribution corresponding to each credit card product, including: performing similarity calculation on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product to obtain a calculation result; optimizing the first standard deviation with the goal of minimizing the calculation result to obtain a target standard deviation corresponding to each credit card product.

[0081] Optionally, similarity calculation is performed on the initial label distribution corresponding to each credit card product and the predicted label distribution corresponding to each credit card product to obtain a calculation result, and then the first standard deviation is optimized with the goal of minimizing the calculation result to obtain a target standard deviation corresponding to each credit card product.

[0082] Specifically, given a training set S, according to each least active customer α learned in the k-1th iteration i The adaptive standard deviation Formula (3) is used for each credit card product x i Generate marker distribution The algorithm first initializes the standard deviation of all credit card products to a fixed constant, and generates the initial label distribution of all credit card products using formula (3):

[0083] Then, the partial least squares method is used to establish the regression equation of the dependent variable matrix Y with respect to the independent variable matrix X, namely the PLS regression model. The regression coefficient β should be solved so that the predicted marker distribution generated by the regression model is consistent with the marker distribution generated by formula (3). Most similar. After training and generating the regression model Y = Xβ, for any credit card product x in the training set i , the corresponding predicted marker distribution is obtained through the regression model x i The model predicts that the customer is The absolute error of the model predicting the customer is The mean absolute error (MAE) on the entire training set is

[0084] Then update the standard deviation. Given the predicted label distribution of all credit card products, select the predicted label distribution with an absolute error lower than MAE as the standard deviation for learning each inactive customer. The predicted labeled distribution samples in this training set show excellent learning ability in customer recommendation and are more suitable for training and learning the standard deviation that is adaptive to each inactive customer. These "excellent" training samples can be divided into c groups according to the least active customers corresponding to their credit card products. At the same time, formula (3) generates a labeled distribution with the same standard deviation for all credit card products in the same group. Grouping all credit card products with the same least active customer into the same group is conducive to the adaptive learning of the standard deviation of the product. The subscript set of all credit card products in the group with the least active customer α is denoted as

[0085] From formula (3), we can get set I α All credit card products in have the same standard deviation σ α The discrete marker distribution of α Each credit card product in x g ,g∈I α With a marker distribution Related, that is:

[0086]

[0087] Among them, g∈Iα ,j∈{1,2,…,c}.

[0088] In the same group, the standard deviation σ α For credit card products α g The resulting discrete Gaussian distribution The distribution of predicted markers calculated by the regression model should be As similar as possible, optionally using KL divergence to measure the similarity between the labeled distribution and the predicted labeled distribution. At the kth iteration, the optimal standard deviation σ for credit card product α α It can be optimized as follows:

[0089]

[0090] Substituting formula (13) into formula (14) yields a nonlinear optimization problem, which is solved using the logarithmic barrier interior point method. This yields a prediction model for inactive customers that is adaptive to each credit card product.

[0091] It should be noted that by extracting multiple latent variables, the PLS method can handle multiple correlations between variables, and can ensure the accuracy and stability of the model even when the number of samples is less than the number of variables. By learning the relationship between independent variables and dependent variables, a data-driven method is provided for predicting inactive customers. An adaptive inactive customer prediction model can be established for each credit card product, inactive customer groups can be screened out for each credit card product, and customers who are likely to become inactive credit card customers in the future can be accurately located.

[0092] In an optional embodiment, after determining the target object of each credit card product from multiple objects to be predicted of each credit card product, prompt information is sent to the target object of each credit card product respectively, wherein the prompt information is used to remind the target object to use the card.

[0093] Optionally, after determining the inactive customer groups for each credit card product, reminder messages can be sent to these customers respectively, and corresponding awakening strategies can be adopted to increase the activity of these customers and achieve accurate awakening.

[0094] In the embodiment of the present application, a label distribution learning algorithm based on multiple multivariate regression is used for modeling. First, feature data of multiple objects to be predicted of at least one credit card product to be predicted is obtained. Then, the feature data of multiple objects to be predicted of each credit card product is input into the regression model corresponding to each credit card product for prediction, and the prediction label distribution corresponding to each credit card product is obtained, wherein the prediction label distribution is a probability distribution representing the active state of the object to be predicted for the credit card product. Then, based on the prediction label distribution, the target object of each credit card product is determined from the multiple objects to be predicted of each credit card product. By establishing an adaptive inactive customer prediction model for each credit card product, it is possible to screen out inactive customer groups for each credit card product, accurately locate customers who may become inactive credit card customers in the future, and achieve the purpose of accurately awakening inactive credit card customers, thereby achieving the technical effect of improving accuracy, and thus solving the technical problem of relatively low accuracy in determining inactive customers of credit card products in the prior art.

[0095] Example 2

[0096] According to an embodiment of the present application, an embodiment of a device for determining a target object is provided, wherein: Figure 3 is a schematic diagram of an optional target object determination device according to an embodiment of the present application, such as Figure 3 As shown, the device includes: a first acquisition unit 301, a first processing unit 302, and a determination unit 303.

[0097] The first acquisition unit 301 is used to acquire feature data of multiple objects to be predicted of at least one credit card product to be predicted;

[0098] The first processing unit 302 is used to input the feature data of multiple objects to be predicted of each credit card product into the regression model corresponding to each credit card product to perform prediction, and obtain the prediction mark distribution corresponding to each credit card product, wherein the prediction mark distribution is a probability distribution representing the active state of the object to be predicted for the credit card product;

[0099] The determination unit 303 is used to determine the target object of each credit card product from the multiple objects to be predicted of each credit card product according to the distribution of the prediction tags.

[0100] In the embodiment of the present application, the first acquisition unit 301 acquires the feature data of multiple objects to be predicted of at least one credit card product to be predicted; the first processing unit 302 inputs the feature data of multiple objects to be predicted of each credit card product into the regression model corresponding to each credit card product for prediction, and obtains the prediction mark distribution corresponding to each credit card product, wherein the prediction mark distribution is the probability distribution representing the active state of the object to be predicted for the credit card product; the determination unit 303 determines the target object of each credit card product from the multiple objects to be predicted of each credit card product according to the prediction mark distribution. By establishing an adaptive inactive customer prediction model for each credit card product, it is possible to screen out inactive customer groups for each credit card product, accurately locate customers who may become inactive credit card customers in the future, and achieve the purpose of accurately awakening inactive credit card customers, thereby achieving the technical effect of improving accuracy, and thus solving the technical problem of relatively low accuracy in determining inactive customers of credit card products in the prior art.

[0101] It should be noted that the above-mentioned first acquisition unit 301, first processing unit 302, and determination unit 303 correspond to steps S101 to S103 in the above-mentioned embodiment, and the examples and application scenarios implemented by the three units and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0102] Furthermore, the determination unit includes: a comparison subunit, used to compare the probability value corresponding to each object to be predicted in the prediction mark distribution with a preset threshold; a determination subunit, used to select an object to be predicted whose probability value is greater than the preset threshold among multiple objects to be predicted as a target object.

[0103] Furthermore, the device also includes the following units, which are used to obtain the regression model corresponding to each credit card product through the following steps: a second acquisition unit, which is used to obtain the historical card usage data corresponding to each credit card product, and vectorize the historical card usage data to obtain the card usage feature vector corresponding to each credit card product; a second processing unit, which is used to construct a discrete Gaussian distribution based on the card usage feature vector corresponding to each credit card product, and obtain the label distribution corresponding to each credit card product; a third processing unit, which is used to construct an independent variable matrix and a dependent variable matrix based on the label distribution corresponding to each credit card product, and establish an initial regression model based on the independent variable matrix and the dependent variable matrix using partial least squares method, wherein the independent variable matrix is ​​used to represent the feature information of all credit card products, and the dependent variable matrix is ​​used to represent the activity information of the sample object for each credit card product; a fourth processing unit, which is used to perform adaptive standard deviation adjustment on each credit card product based on the initial regression model to obtain the regression model corresponding to each credit card product.

[0104] Furthermore, the third processing unit includes: a first processing sub-unit, used to construct a linear dependency relationship between the independent variable matrix and the dependent variable matrix; a second processing sub-unit, used to extract multiple pairs of latent variables associated with the independent variables and the dependent variables based on the linear dependency relationship, and maximize the covariance between the latent variables to obtain an initial regression model.

[0105] Furthermore, the fourth processing unit includes: a third processing sub-unit, used to configure a first standard deviation for each credit card product, and determine an initial mark distribution corresponding to each credit card product based on the first standard deviation and the mark distribution corresponding to each credit card product; a fourth processing sub-unit, used to determine a predicted mark distribution corresponding to each credit card product based on an initial regression model, and determine a target standard deviation corresponding to each credit card product based on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product; a fifth processing sub-unit, used to update the initial regression model based on the target standard deviation corresponding to each credit card product to obtain a regression model corresponding to each credit card product.

[0106] Furthermore, the fourth processing sub-unit includes: a calculation module, which is used to perform similarity calculation on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product to obtain a calculation result; a processing module, which is used to optimize the first standard deviation with the goal of minimizing the calculation result to obtain a target standard deviation corresponding to each credit card product.

[0107] Furthermore, the device also includes: a sending unit, which is used to send prompt information to the target object of each credit card product after determining the target object of each credit card product from multiple objects to be predicted of each credit card product, wherein the prompt information is used to remind the target object to use the card.

[0108] Example 3

[0109] An embodiment of the present application may provide an electronic device, Figure 4 is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 4 As shown, the electronic device may include: one or more ( Figure 4 (only one is shown) processor 402, memory 404, storage controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0110] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0111] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: obtain feature data of multiple objects to be predicted of at least one credit card product to be predicted; input the feature data of multiple objects to be predicted of each credit card product into the regression model corresponding to each credit card product for prediction, and obtain the prediction mark distribution corresponding to each credit card product, wherein the prediction mark distribution is the probability distribution representing the active state of the object to be predicted for the credit card product; based on the prediction mark distribution, determine the target object of each credit card product from the multiple objects to be predicted of each credit card product.

[0112] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: compare the probability value corresponding to each object to be predicted in the prediction mark distribution with a preset threshold; and take the object to be predicted whose probability value is greater than the preset threshold among multiple objects to be predicted as the target object.

[0113] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: obtain the historical card usage data corresponding to each credit card product, and vectorize the historical card usage data to obtain the card usage feature vector corresponding to each credit card product; construct a discrete Gaussian distribution based on the card usage feature vector corresponding to each credit card product to obtain the label distribution corresponding to each credit card product; based on the label distribution corresponding to each credit card product, construct an independent variable matrix and a dependent variable matrix, and based on the independent variable matrix and the dependent variable matrix, use partial least squares method to establish an initial regression model, wherein the independent variable matrix is ​​used to represent the feature information of all credit card products, and the dependent variable matrix is ​​used to represent the activity information of the sample object for each credit card product; based on the initial regression model, adaptively adjust the standard deviation of each credit card product to obtain the regression model corresponding to each credit card product.

[0114] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: construct a linear dependency relationship between the independent variable matrix and the dependent variable matrix; based on the linear dependency relationship, extract multiple pairs of latent variables associated with the independent variables and the dependent variables, and maximize the covariance between the latent variables to obtain an initial regression model.

[0115] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: configure a first standard deviation for each credit card product, and determine the initial mark distribution corresponding to each credit card product based on the first standard deviation and the mark distribution corresponding to each credit card product; determine the predicted mark distribution corresponding to each credit card product based on the initial regression model, and determine the target standard deviation corresponding to each credit card product based on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product; update the initial regression model based on the target standard deviation corresponding to each credit card product to obtain a regression model corresponding to each credit card product.

[0116] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: perform similarity calculation on the initial mark distribution corresponding to each credit card product and the predicted mark distribution corresponding to each credit card product to obtain a calculation result; optimize the first standard deviation with the goal of minimizing the calculation result to obtain a target standard deviation corresponding to each credit card product.

[0117] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: after determining the target object of each credit card product from multiple objects to be predicted for each credit card product, send prompt information to the target object of each credit card product respectively, wherein the prompt information is used to remind the target object to use the card.

[0118] It can be understood by those skilled in the art that Figure 4 The structure shown is for illustration only, and the electronic device may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, and other terminal devices. Figure 4 The structure of the electronic device is not limited. Figure 4 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 4 Different configurations are shown.

[0119] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0120] Example 4

[0121] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the target object determination method provided in the first embodiment.

[0122] Optionally, in this embodiment, the above storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0123] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the program of the target object determination method steps.

[0124] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0125] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0126] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0127] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0128] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0129] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0130] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining a target object, characterized in that: include: Acquire feature data of multiple objects to be predicted of at least one credit card product to be predicted; Inputting feature data of multiple objects to be predicted for each credit card product into a regression model corresponding to each credit card product for prediction, and obtaining a prediction mark distribution corresponding to each credit card product, wherein the prediction mark distribution is a probability distribution representing the active state of the object to be predicted for the credit card product; According to the prediction tag distribution, the target object of each credit card product is determined from the multiple objects to be predicted of each credit card product.

2. The method according to claim 1, characterized in that Determining the target object of each credit card product from the multiple objects to be predicted of each credit card product according to the prediction tag distribution includes: Comparing the probability value corresponding to each object to be predicted in the prediction marker distribution with a preset threshold; Among the multiple objects to be predicted, the object to be predicted whose probability value is greater than the preset threshold is taken as the target object.

3. The method according to claim 1, characterized in that: The regression model corresponding to each credit card product is obtained by the following steps: Acquire historical card usage data corresponding to each credit card product, and vectorize the historical card usage data to obtain a card usage feature vector corresponding to each credit card product; Constructing a discrete Gaussian distribution according to the card usage feature vector corresponding to each credit card product to obtain a label distribution corresponding to each credit card product; Based on the label distribution corresponding to each credit card product, an independent variable matrix and a dependent variable matrix are constructed, and based on the independent variable matrix and the dependent variable matrix, an initial regression model is established using the partial least squares method, wherein the independent variable matrix is ​​used to represent the characteristic information of all credit card products, and the dependent variable matrix is ​​used to represent the activity information of the sample objects for each credit card product; Based on the initial regression model, an adaptive standard deviation adjustment is performed on each credit card product to obtain a regression model corresponding to each credit card product.

4. The method according to claim 3, characterized in that Based on the independent variable matrix and the dependent variable matrix, an initial regression model is established using the partial least squares method, including: Constructing a linear dependency relationship between the independent variable matrix and the dependent variable matrix; Based on the linear dependency, multiple pairs of latent variables associated with independent variables and dependent variables are extracted, and the covariance between the latent variables is maximized to obtain the initial regression model.

5. The method according to claim 3, characterized in that: Based on the initial regression model, adaptive standard deviation adjustment is performed on each credit card product to obtain a regression model corresponding to each credit card product, including: configuring a first standard deviation for each of the credit card products, and determining an initial label distribution corresponding to each of the credit card products based on the first standard deviation and the label distribution corresponding to each of the credit card products; Determine the predicted marker distribution corresponding to each credit card product based on the initial regression model, and determine the target standard deviation corresponding to each credit card product according to the initial marker distribution corresponding to each credit card product and the predicted marker distribution corresponding to each credit card product; The initial regression model is updated according to the target standard deviation corresponding to each credit card product to obtain a regression model corresponding to each credit card product.

6. The method according to claim 5, characterized in that Determining a target standard deviation corresponding to each credit card product according to the initial label distribution corresponding to each credit card product and the predicted label distribution corresponding to each credit card product includes: Performing similarity calculation on the initial label distribution corresponding to each credit card product and the predicted label distribution corresponding to each credit card product to obtain a calculation result; The first standard deviation is optimized with the goal of minimizing the calculated result to obtain a target standard deviation corresponding to each credit card product.

7. The method according to claim 1, characterized in that After determining the target object of each credit card product from the multiple objects to be predicted of each credit card product, the method further includes: Prompt information is sent to the target object of each credit card product respectively, wherein the prompt information is used to remind the target object to use the card.

8. A device for determining a target object, characterized in that: include: A first acquisition unit is used to acquire feature data of multiple objects to be predicted of at least one credit card product to be predicted; A first processing unit is used to input feature data of multiple objects to be predicted of each credit card product into a regression model corresponding to each credit card product to perform prediction, and obtain a prediction mark distribution corresponding to each credit card product, wherein the prediction mark distribution is a probability distribution representing the active state of the object to be predicted for the credit card product; A determination unit is used to determine the target object of each credit card product from the multiple objects to be predicted of each credit card product according to the prediction mark distribution.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the method for determining a target object as claimed in any one of claims 1 to 7 when running.

10. An electronic device, characterized in that: The electronic device includes one or more processors; A memory for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to run the programs, wherein the program is configured to execute the method for determining the target object described in any one of claims 1 to 7 when running.

Images