User identification method and device, equipment and storage medium

Abstract

The invention discloses a user identification method and device, equipment and a storage medium, and is applied to the technical field of artificial intelligence. The method comprises the following steps: acquiring user contribution feature data of a user to be identified; inputting the user contribution feature data into a pre-trained user identification model to perform high-quality user identification to obtain a user identification result; wherein the user identification model is obtained by pre-training a preset identification network model based on the contribution feature sample data and the contribution feature weight of the sample training object. According to the technical scheme of the embodiment, accurate identification of high-quality users in the financial field is realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a user identification method, device, equipment and storage medium. Background Art

[0002] At present, the customer contribution evaluation system built based on accounting standards to measure the actual benefits provided by customers can achieve accurate evaluation of customer contributions, thereby facilitating the identification of high-quality customers and providing data support for various fields within the financial industry, such as tiered services, precise services and product pricing.

[0003] However, the factors that need to be considered in the process of high-quality customer identification in the financial professional field are different from those in other fields. The financial professional field has its own considerations and indicators. The existing high-quality user identification methods are not fully applicable to the financial professional field. Therefore, how to accurately identify high-quality users in the financial field has become a problem that needs to be solved urgently. Summary of the invention

[0004] The present invention provides a user identification method, device, equipment and storage medium to achieve accurate identification of high-quality users in the financial field.

[0005] According to one aspect of the present invention, a user identification method is provided, the method comprising:

[0006] Obtain user contribution feature data of the user to be identified;

[0007] Inputting the user contribution feature data into a pre-trained user identification model to perform high-quality user identification to obtain a user identification result;

[0008] The user recognition model is obtained by pre-training a preset recognition network model based on contribution feature sample data and contribution feature weights of a sample training object.

[0009] According to another aspect of the present invention, there is provided a user identification device, the device comprising:

[0010] A feature data acquisition module, used to acquire user contribution feature data of the user to be identified;

[0011] A user identification module, used to input the user contribution feature data into a pre-trained user identification model to perform high-quality user identification and obtain a user identification result;

[0012] The user recognition model is obtained by pre-training a preset recognition network model based on contribution feature sample data and contribution feature weights of a sample training object.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the user identification method described in any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the user identification method described in any embodiment of the present invention when executed.

[0018] According to another aspect of the present invention, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, the user identification method according to any embodiment of the present invention is implemented.

[0019] The technical solution of this application adopts a pre-trained user identification model to predict whether the user to be identified is a high-quality user based on the user contribution feature data. In the prediction process, the user contribution features of different dimensions of the user to be identified are comprehensively considered, thereby achieving accurate identification of user quality in the financial field. In addition, in the process of training the user identification model, contribution features of different dimensions and their corresponding contribution feature weights are introduced, so that the differences between contribution features can be effectively learned during the model training process, which improves the accuracy of model training, thereby further improving the accuracy of identifying user quality in the financial field.

[0020] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 is a flow chart of a user identification method provided according to Embodiment 1 of the present invention;

[0023] Figure 2 is a flow chart of a user identification method provided according to Embodiment 2 of the present invention;

[0024] Figure 3 is a schematic diagram of the structure of a user identification device provided according to Embodiment 3 of the present invention;

[0025] Figure 4 It is a schematic diagram of the structure of an electronic device for implementing the user identification method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention 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 interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] Embodiment 1

[0029] Figure 1 This is a flowchart of a user identification method provided in the first embodiment of the present invention. This embodiment can be applied to the situation of accurately identifying high-quality users in the financial field. The method can be executed by a user identification device, which can be implemented in the form of hardware and / or software, and the user identification device can be configured in an electronic device. Figure 1 As shown, the method includes:

[0030] S110: Obtain user contribution feature data of the user to be identified.

[0031] S120, inputting the user contribution feature data into a pre-trained user identification model to perform high-quality user identification and obtain a user identification result.

[0032] The user recognition model is obtained by pre-training a preset recognition network model based on contribution feature sample data and contribution feature weights of a sample training object.

[0033] Among them, the user to be identified may be a user to be identified for customer quality; the user contribution feature data may include contribution features under at least one feature category or feature dimension and their corresponding feature parameter values. Feature categories may include product categories, income categories, and cost categories; for example, contribution features under product categories may include deposits, loans, and intermediary business, etc.; contribution features under income categories may include capital income, interest income, and fee income, etc.; contribution features under cost categories may include capital cost, interest cost, risk cost, and tax cost, etc.

[0034] Exemplarily, user contribution feature data of the user to be identified is obtained, including user deposits, user loans, intermediary businesses handled by the user, user capital income, user interest income, user fee income, user capital cost, user interest cost, user risk cost and user tax cost. The user contribution feature data is input into the pre-trained user identification model to perform high-quality user identification and obtain a user identification result. The user identification result may include that the user to be identified is a high-quality user or that the user to be identified is not a high-quality user.

[0035] It should be noted that the information collected in this embodiment, such as the characteristic parameter values ​​of the user contribution features, is 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 the relevant data comply with the relevant laws, regulations and standards of the relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0036] The user identification model is used to predict user quality and identify whether the user is a high-quality user. The user identification model can be pre-trained by relevant technical personnel, and can be pre-trained based on a large number of sample training objects to obtain a preset identification network model. The preset identification network model can be a Maximum Entropy Model.

[0037] Specifically, the contribution feature sample data of the sample training object is obtained. The sample training object may be a historical user in a historical time period. The contribution feature sample data is a feature parameter value of the historical user under at least one contribution feature.

[0038] It should be noted that due to the differences between contribution features, different contribution features have different effects on the evaluation of user quality. Therefore, in order to improve the model accuracy of the user identification model, the difference of contribution features is reflected by introducing contribution feature weights during the model training process. Among them, the contribution feature weights of the contribution features can be set by relevant technical personnel based on actual experience values, experimental values ​​or based on expert experience.

[0039] The technical solution of this application adopts a pre-trained user identification model to predict whether the user to be identified is a high-quality user based on the user contribution feature data. In the prediction process, the user contribution features of different dimensions of the user to be identified are comprehensively considered, thereby achieving accurate identification of user quality in the financial field. In addition, in the process of training the user identification model, contribution features of different dimensions and their corresponding contribution feature weights are introduced, so that the differences between contribution features can be effectively learned during the model training process, which improves the accuracy of model training, thereby further improving the accuracy of identifying user quality in the financial field.

[0040] The present application also provides a model training method for a user identification model. In an optional embodiment, the model training method for the user identification model is as follows:

[0041] Step a1, obtaining contribution feature sample data of a sample training object; the contribution feature sample data includes at least one contribution feature and its corresponding feature parameter value; the sample training object is pre-labeled with a standard category label.

[0042] The labels of the standard categories of the sample training objects can be pre-labeled by relevant technicians, and can be labeled using existing sample labeling tools, which is not limited in this embodiment. For example, the expression form of the contribution sample data of the sample training objects can be shown in Table 1.

[0043] Table 1

[0044] Sample training objects Contribution feature a ...... Contribution feature b Contribution feature m Standard Category Historical User 1 XA1 ...... Yb1 Zm1 1 Historical User 2 XA ...... Yb2 Zm2 0 …… …… ...... …… …… 0 Historical User Xan ...... Yb Zn 1

[0045] Among them, the standard category "1" may be a category identifier of a high-quality user; and the standard category "0" may be a category identifier of a non-high-quality user.

[0046] Step a2: Determine the contribution feature weights corresponding to each contribution feature.

[0047] Exemplarily, the contribution feature weight of each contribution feature may be manually set by relevant technical personnel based on actual experience values ​​or experimental values, or may be randomly generated within a preset range.

[0048] To further improve the accuracy of determining the contribution feature weights, in an optional embodiment, the contribution feature weights corresponding to each contribution feature are determined, including: determining the feature category of each contribution feature; for any same feature category, determining the contribution feature weights of each contribution feature under the same feature dimension based on the feature association relationship between the contribution features under the same feature category.

[0049] Among them, the feature categories may include product categories, cost categories and income categories; each feature category includes contribution features under its category. For example, the income category may include capital income, interest income and fee income.

[0050] For any of the same feature categories, there is a certain feature association relationship between the contribution features under the same feature category. For example, for the product category, the feature category includes contribution features such as deposits, loans, and intermediary business. In the process of high-quality user evaluation, the importance of intermediary business is less than that of deposits, and the importance of deposits is less than that of loans. Therefore, based on this feature association relationship, a weight generation rule can be set for the product category, which must ensure that the contribution feature weight of the intermediary business is less than the contribution feature weight of the deposit, and the contribution feature weight of the deposit is less than the contribution feature weight of the loan.

[0051] It is understandable that in the process of generating the contribution feature weights of each contribution feature, in addition to the weight generation rules corresponding to the above feature categories, the weight generation range of each contribution feature weight can also be combined. For example, the weight generation range of the intermediate business can be [0.1, 0.2]. The weight generation range of each contribution feature can be pre-selected and set by relevant technical personnel based on actual experience values ​​or experimental values.

[0052] It can be seen that in this optional scheme, by determining the feature category of each contribution feature and determining the contribution feature weight of each contribution feature under the same feature dimension based on the feature correlation relationship between the contribution features under the same feature category, the feature correlation relationship between the contribution features is combined in the process of generating the contribution feature weight, thereby improving the generation accuracy of the contribution feature weight.

[0053] Step a3: input the contribution feature sample data and contribution feature weights into the preset recognition network model to obtain the predicted category of the model output.

[0054] The recognition network model may be a maximum entropy model. The calculation formula of the maximum entropy model may be as follows:

[0055]

[0056] Among them, P w (y|x) represents the conditional probability of output y given input x; w irepresents the i-th contributing feature; Z w (x) represents the normalization factor; f i (x, y) represents the calculation result of the characteristic function of the i-th contributing feature; f(x, y) represents the characteristic function, which is used to characterize the correlation between the input x and the output y; x0 represents the set contributing feature; y0 represents the set category value.

[0057] Step a4: Perform model training on the recognition model according to the predicted category and the standard category until the preset model training end condition is met to obtain a user recognition model.

[0058] Specifically, according to the prediction category and the standard category, based on the preset loss function, the loss value is determined; according to the loss value, the recognition model is trained until the preset model training end condition is reached to obtain the user recognition model. The model training end condition may be that the loss value tends to be stable or the loss value reaches a set threshold. The loss function may be pre-set by relevant technical personnel according to actual needs. For example, the loss function may be a regularized loss (RegularizedLoss Function) or a negative log-likelihood loss function (Negative Log-Likelihood Loss), etc.

[0059] The technical solution of the above-mentioned embodiment trains the user identification model by adopting the contribution feature sample data of the sample training object and the contribution feature weights of each contribution feature. During the model training process, the contribution features in different dimensions are considered, and the contribution feature weights of the contribution features in different dimensions are introduced. This enables the model training process to learn the influence of contribution features of different degrees of influence on the high-quality user identification process, thereby improving the accuracy of user identification model training, and further improving the accuracy of user quality identification of the users to be identified.

[0060] In order to meet the needs of personalized identification of high-quality users in the financial field and improve the interpretability of user identification models, thereby meeting the requirements of high-quality user identification in actual financial scenarios, this application also provides another method for determining the weights of contribution features.

[0061] In an optional embodiment, determining the contribution feature weights corresponding to the contribution features includes:

[0062] Step b1: Determine a product decision matrix based on the feature association relationship between the contribution features belonging to the product category.

[0063] Taking the contribution characteristics under the product category including deposits, loans and intermediary business as an example, there is a characteristic correlation relationship between deposits, loans and intermediary business. For example, the importance of intermediary business in the user quality assessment process is less than that of deposits and loans.

[0064] Since there is a specific feature correlation relationship between the contribution features of the product category, the field comparison boundary under the product category is introduced, that is, the product feature boundary range, to ensure that the matrix values ​​in the generated product judgment matrix meet the judgment conditions of the product feature boundary range. Among them, the product feature boundary range is shown in Table 2.

[0065] Table 2

[0066] Contribution characteristics deposit loan Intermediary business deposit 1 [0.5,1] [2,4] loan [1,2] 1 [4,8] Intermediary business [0.25,0.25] [0.125,0.125] 1

[0067] Based on the given product feature boundary range, the feature correlation relationship between the feature contributions under the product category can be expressed as shown in Table 3. Among them, the product judgment matrix combines the feature correlation relationship and the product feature boundary range in the generation process to ensure the rationality of the product judgment matrix.

[0068] Table 3

[0069] Contribution characteristics deposit loan Intermediary business deposit 1 0.5 2 loan 2 1 4 Intermediary business 0.5 0.25 1

[0070] The product decision matrix P is as follows:

[0071]

[0072] Step b2: Determine the income determination matrix based on the feature association relationship between the contribution features belonging to the income category.

[0073] Taking the contribution characteristics under the income category including capital income, interest income and fee income as an example, there is a characteristic correlation relationship between capital income, interest income and fee income. For example, the importance of capital income in the user quality evaluation process is less than that of interest income and fee income, and the importance of fee income in the user quality evaluation process is less than that of interest income, etc.

[0074] Since there is a specific characteristic correlation between the contribution characteristics of the income category, the field comparison boundary under the income category, that is, the income characteristic boundary range, is introduced to ensure that the matrix values ​​in the generated income judgment matrix meet the judgment conditions of the income characteristic boundary range. Among them, the income characteristic boundary range is shown in Table 4.

[0075] Table 4

[0076] Contribution characteristics Fund income Interest income Fee income Fund income 1 [0.5,1] 1 Interest income [1,2] 1 [1,2] Fee income 1 [0.5,1] 1

[0077] Based on the given income characteristic boundary range, the characteristic correlation relationship between the characteristic contributions under the income category can be expressed as shown in Table 5. Among them, the income judgment matrix combines the characteristic correlation relationship and the income characteristic boundary range in the generation process to ensure the rationality of the income judgment matrix.

[0078] Table 5

[0079] Contribution characteristics Fund income Interest income Fee income Fund income 1 0.5 1 Interest income 2 1 1 Fee income 1 1 1

[0080] The income determination matrix I is as follows:

[0081]

[0082] Step b3: Determine a cost determination matrix based on the feature association relationship between the contribution features belonging to the cost category.

[0083] Taking the contribution characteristics under the cost category including capital cost, interest cost, risk cost and tax cost as an example, there is a characteristic correlation relationship among capital cost, interest cost, risk cost and tax cost. For example, the importance of tax cost in the user quality assessment process is less than that of capital cost, interest cost and risk cost.

[0084] Since there is a specific characteristic correlation between the contribution characteristics of the cost category, the field comparison boundary under the cost category, that is, the cost characteristic boundary range, is introduced to ensure that the matrix values ​​in the generated cost judgment matrix meet the judgment conditions of the income characteristic boundary range. Among them, the cost characteristic boundary range is shown in Table 6.

[0085] Table 6

[0086] Contribution characteristics Capital Cost Interest costs Risk Cost Tax costs Fund income 1 [0.5,1] [0.25,0.5] 1 Interest income [1,2] 1 [0.5,1] [1,2] Fee income [2,4] [1,2] 1 [2,4] Tax costs 1 [0.25,0.5] [0.25,0.5] 1

[0087] Based on the given cost feature boundary range, the feature association relationship between the feature contributions under the cost category can be expressed as shown in Table 7. In the generation process of the cost judgment matrix, both the feature association relationship and the cost feature boundary range are considered to ensure the rationality of the cost judgment matrix.

[0088] Table 7

[0089] Contribution characteristics Capital Cost Interest costs Risk Cost Tax costs Fund income 1 0.5 0.25 1 Interest income 2 1 0.5 1 Fee income 4 2 1 2 Tax costs 1 1 0.5 1

[0090] The cost decision matrix C is as follows:

[0091]

[0092] The execution order of the above steps b1 to b3 is not limited in this embodiment. For example, steps b1 to b3 may be executed sequentially or in parallel.

[0093] Step b4: Determine the contribution feature weight of each contribution feature according to the product determination matrix, the income determination matrix and the cost determination matrix.

[0094] According to the product judgment matrix, the income judgment matrix and the cost judgment matrix, based on the AHP (Analytic Hierarchy Process) model, the contribution feature weights corresponding to each contribution feature are determined.

[0095] Specifically, the product categories, income categories and cost categories are hierarchically divided into product layer, income layer and cost layer respectively. For the criteria in each layer, the importance of each contribution feature is determined through the feature association relationship between the contribution features. Among them, the product judgment matrix, the income judgment matrix and the cost judgment matrix can be used to characterize the degree of association or importance between contribution features. The product judgment matrix, the income judgment matrix and the cost judgment matrix are respectively subjected to matrix standardization processing to obtain standardized matrices; according to each standardized matrix, the contribution feature weight of each contribution feature is determined based on the weight calculation method of AHP.

[0096] It should be noted that after obtaining the contribution feature weights, a consistency check is required to ensure that the judgment of the matrix is ​​consistent; if the consistency is poor, it is necessary to re-determine the judgment matrix of each type. In the process of consistency check, the check can be performed based on the random consistency reference indicator.

[0097] It can be seen that in the technical solution of this embodiment, by introducing the category judgment matrix of contribution features under different categories and performing the contribution feature weights corresponding to each contribution feature based on the judgment matrix under each category, the personalization of factors considered in the user quality assessment process in the financial field is met, so that the interpretability of the model is improved during the subsequent user identification model training, the contribution indicator weight determination requirements of the financial scenario are met, and the rationality of the contribution feature weights is improved.

[0098] In order to improve the prediction accuracy of the user recognition model, the trained user recognition model can also be model verified and tested. In an optional embodiment, after the recognition model is trained according to the prediction category and the standard category until the preset model training end condition is met and the user recognition model is obtained, it also includes:

[0099] Step c1: Obtain a verification data set of a sample verification object.

[0100] The verification data set of the sample verification object includes the contribution feature sample data of the sample verification object and its corresponding feature parameter values; the sample verification object is pre-labeled with a standard category label.

[0101] The sample quantity ratio among the sample training objects, the sample test objects and the sample verification objects can be preset by relevant technical personnel according to actual needs. For example, the sample quantity ratio among the sample training objects, the sample test objects and the sample verification objects can be 5:3:2. The sample verification objects can also be part of the sample objects divided from the sample training objects.

[0102] Step c2: Use the verification data set to perform model verification on the user identification model to obtain a model verification result.

[0103] The user identification model is validated or cross-validated using a validation data set to obtain a model validation result. For example, the model validation result may include accuracy and precision.

[0104] Step c3: update the contribution feature weights according to the model verification results to obtain updated contribution feature weights.

[0105] If the model verification result does not meet the preset model verification conditions, the contribution feature weights are updated to obtain updated contribution feature weights; if the model verification result meets the preset model verification conditions, the iterative update process of the model is terminated.

[0106] If the model validation result does not meet the preset model validation conditions, the contribution feature weights may be updated by randomly adjusting the values ​​of the contribution feature weights of each contribution feature to obtain updated contribution feature weights.

[0107] Optionally, in order to further improve the updating accuracy of the contribution feature weights, thereby reducing the number of model iteration updates and improving the model training efficiency, in an optional embodiment, the contribution feature weights are updated according to the model verification results to obtain the updated contribution feature weights, including:

[0108] Step c31: If the model verification result does not meet the preset model verification judgment conditions, the product judgment matrix is ​​updated according to the product feature boundary range, the income judgment matrix is ​​updated according to the income feature boundary range, and the cost judgment matrix is ​​updated according to the cost feature boundary range to obtain the updated product judgment matrix, income judgment matrix and cost judgment matrix.

[0109] Specifically, within the product feature boundary range, the matrix values ​​of the product judgment matrix may be randomly generated to update the product judgment matrix. Within the cost feature boundary range, the matrix values ​​of the cost judgment matrix may be randomly generated to update the cost judgment matrix. Within the revenue judgment matrix range, the matrix values ​​of the revenue judgment matrix may be randomly generated to update the revenue judgment matrix. It should be noted that in the process of randomly generating the matrix values ​​of various judgment matrices, under the condition of ensuring that the matrix values ​​are within the product feature boundary range, feature association relationships may also be introduced.

[0110] Step c32: update the contribution feature weights according to the updated product determination matrix, income determination matrix and cost determination matrix to obtain updated contribution feature weights.

[0111] According to the updated product judgment matrix, revenue judgment matrix and cost judgment matrix, based on the AHP model, the updated contribution feature weights are obtained.

[0112] It can be seen that the above technical solution takes into account the boundary range values ​​of various features in the process of updating the contribution feature weights, and generates a judgment matrix that satisfies the feature boundary range, so that the updated contribution feature weights are more reasonable, and the accuracy of the contribution feature weight update is improved, thereby reducing the number of model iteration updates, and further improving the efficiency of post-model training.

[0113] Step c4: perform iterative model update on the user identification model based on the updated contribution feature weights until a preset model iteration update end condition is met to obtain a target user identification model.

[0114] Among them, the model iteration update end condition can be that the user identification model is iteratively updated based on the updated contribution feature weight to meet the model verification judgment condition, and the user identification model that meets the model verification judgment condition is determined as the target user identification model.

[0115] For example, if the user identification model meets the model verification condition after five rounds of iterative updates, the user identification model after the fifth round of iterative updates is determined as the target user identification model.

[0116] In an optional embodiment, the user identification model is iteratively updated based on the updated contribution feature weights until a preset model iteration update end condition is met to obtain a target user identification model, including:

[0117] Step c41: performing iterative model update on the user identification model based on the updated contribution feature weights, and determining the user identification model that meets the preset model iterative update conditions as the target user identification model.

[0118] Step c42: If the user recognition model does not meet the preset model iteration update termination condition within the preset iteration update times, multiple reference recognition models under each iteration update times are obtained.

[0119] The number of iterative updates may be pre-set by relevant technical personnel based on actual experience or experimental values. For example, the number of iterative updates may be 50 times.

[0120] If the user recognition model still does not meet the preset model iteration update end condition when the number of iteration updates is reached, the reference recognition model at each iteration update number is obtained. For example, if the number of iteration updates is 50 times, and the model iteration update end condition is still not met after the 50th iteration update, the model obtained in each iteration of the 50 iterations is determined as the reference recognition model, and 50 reference recognition models can be obtained.

[0121] Step c43: Obtain a test data set of a sample test object.

[0122] The test data set is used to test the model performance, which may include accuracy test, recall test, and F1 score test.

[0123] Step c44: Use the test data set to perform model testing on each reference recognition model to obtain the model testing results of each reference recognition model.

[0124] Specifically, the test data sets are used to perform model tests on the reference recognition models in turn to obtain model test results corresponding to each reference recognition model.

[0125] Specifically, the test data set is input into the reference recognition model to obtain the prediction result of the model for the test data set. According to the standard category and the prediction result of the test data set, the evaluation result of at least one evaluation indicator is determined. Among them, the evaluation indicator may include accuracy, precision, recall rate and F1 score, etc. A model verification result including the evaluation results of each evaluation indicator is generated.

[0126] Optionally, during the model testing process, the model test results of the user identification model may also include overfitting test results and underfitting test results. Overfitting is specifically manifested as the model performing well on the training set but performing poorly on the test set, indicating that the model is over-trained on the training data and fails to generalize to new data. It is usually manifested as low training error and high test error. Underfitting is specifically manifested as the model performing poorly on both the training set and the test set, indicating that the model is too simple and cannot capture the regularity of the data.

[0127] Step c45: Determine the target user recognition model according to the model test results of each reference recognition model.

[0128] Combined with the model test results of each reference recognition model, the reference recognition model with the best model test result is selected from each reference recognition model as the target user recognition model.

[0129] It can be seen that the above technical solution sets the end condition of the number of iteration updates during the model iteration update process, thereby preventing the model iteration from falling into an infinite loop, which leads to a longer model training cycle and low efficiency. On the basis of the given model iteration update end condition, by performing model testing on the reference recognition model obtained by each iteration update, the optimal reference recognition model can be obtained while ensuring the model training efficiency and preventing the model from falling into an infinite loop. It is used as the final target user recognition model for subsequent high-quality user recognition, thereby improving the recognition accuracy of the model.

[0130] In summary, the model is verified by obtaining the verification data set of the sample verification object, and the contribution feature weights are updated according to the model verification results, so that the user recognition model is iteratively updated based on the updated contribution feature weights to obtain the target recognition model with the best effect, which further ensures the model quality of the trained target user recognition model, thereby improving the model's recognition efficiency and accuracy for high-quality users.

[0131] Embodiment 2

[0132] Figure 2 This is a schematic diagram of the interaction process of a user identification method provided in Embodiment 2 of the present invention. This embodiment provides a preferred example based on the above embodiments.

[0133] like Figure 2 As shown, the method execution steps of this embodiment are as follows:

[0134] S21. Use the contribution features of the sample training objects and their corresponding special parameter values ​​as well as the contribution feature weights of the contribution features to perform model training on the pre-built recognition model until the preset model training end conditions are met to obtain a user recognition model.

[0135] The contribution feature weight of the contribution feature can be obtained in two ways: the first way is a weight determination method based on feature categories; the second way is a weight determination method based on a judgment matrix.

[0136] The weight determination process for the first method is as follows:

[0137] S211a. Determine the feature category of each contributing feature.

[0138] S211b. For any same feature category, determine the contribution feature weights of the contribution features under the same feature dimension according to the feature association relationship between the contribution features under the same feature category.

[0139] The weight determination process for the second method is as follows:

[0140] S212a. Determine a product determination matrix according to the feature association relationship between the contribution features belonging to the product category.

[0141] S212b. Determine an income determination matrix based on the feature association relationship between the contribution features belonging to the income category.

[0142] S212c: Determine a cost determination matrix according to the feature association relationship between the contribution features belonging to the cost category.

[0143] S212d. Determine the contribution feature weight of each contribution feature according to the product determination matrix, the income determination matrix and the cost determination matrix.

[0144] S22. Use the verification data set of the sample verification object to perform model verification on the user identification model to obtain a model verification result.

[0145] S23a. If the model verification result does not meet the preset model verification judgment condition, the contribution feature weight is updated to obtain an updated contribution feature weight.

[0146] Specifically, the product judgment matrix is ​​updated according to the boundary range of product characteristics, the income judgment matrix is ​​updated according to the boundary range of income characteristics, and the cost judgment matrix is ​​updated according to the boundary range of cost characteristics to obtain updated product judgment matrix, income judgment matrix and cost judgment matrix; according to the updated product judgment matrix, income judgment matrix and cost judgment matrix, the contribution feature weights are updated to obtain updated contribution feature weights.

[0147] S23b: If the model verification result meets the preset model verification judgment condition, the current user identification model is determined as the target user identification model.

[0148] S24. If the user recognition model does not meet the preset model iteration update termination condition within the preset number of iteration updates, multiple reference recognition models at each number of iteration updates are obtained.

[0149] S25. Obtain a test data set of the sample test object.

[0150] S26. Perform model testing on each reference recognition model using a test data set to obtain model testing results for each reference recognition model.

[0151] S27. Determine the target user recognition model according to the model test results of each reference recognition model.

[0152] S28. Obtain user contribution feature data of the user to be identified.

[0153] S29. Input the user contribution feature data into the target user identification model to perform high-quality user identification and obtain a user identification result.

[0154] It should be noted that the information collected in this embodiment, such as the contribution feature sample data of the samples of the sample training object, the verification data set of the sample verification object, and the test data set of the sample test object, is 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 the relevant data comply with the relevant laws, regulations and standards of the relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0155] Embodiment 3

[0156] Figure 3 A schematic diagram of the structure of a user identification device provided in the third embodiment of the present invention. A user identification device provided in the embodiment of the present invention can be used to accurately identify high-quality users in the financial field. The user identification device can be implemented in the form of hardware and / or software, such as Figure 3 As shown, the device specifically includes: a feature data acquisition module 301 and a user identification module 302. Among them,

[0157] The feature data acquisition module 301 is used to acquire user contribution feature data of the user to be identified;

[0158] A user identification module 302 is used to input the user contribution feature data into a pre-trained user identification model to perform high-quality user identification and obtain a user identification result;

[0159] The user recognition model is obtained by pre-training a preset recognition network model based on contribution feature sample data and contribution feature weights of a sample training object.

[0160] The technical solution of this application adopts a pre-trained user identification model to predict whether the user to be identified is a high-quality user based on the user contribution feature data. In the prediction process, the user contribution features of different dimensions of the user to be identified are comprehensively considered, thereby achieving accurate identification of user quality in the financial field. In addition, in the process of training the user identification model, contribution features of different dimensions and their corresponding contribution feature weights are introduced, so that the differences between contribution features can be effectively learned during the model training process, which improves the accuracy of model training, thereby further improving the accuracy of identifying user quality in the financial field.

[0161] Optionally, the device further includes: a model training module; the model training module is used to train the user recognition model; the model training module includes:

[0162] A sample data acquisition unit, used to acquire contribution feature sample data of a sample training object; the contribution feature sample data includes at least one contribution feature and its corresponding feature parameter value; the sample training object is pre-labeled with a standard category label;

[0163] A feature weight determination unit, used to determine the contribution feature weights corresponding to each of the contribution features;

[0164] A category prediction unit, used to input the contribution feature sample data and the contribution feature weight into a preset recognition network model to obtain a predicted category output by the model;

[0165] The model training unit is used to perform model training on the recognition model according to the predicted category and the standard category until a preset model training end condition is met to obtain a user recognition model.

[0166] Optionally, the feature weight determination unit includes:

[0167] A feature category determination subunit, used to determine the feature category of each of the contributing features;

[0168] The first weight determination subunit is used to determine, for any same feature category, the contribution feature weights of the contribution features under the same feature dimension according to the feature association relationship between the contribution features under the same feature category.

[0169] Optionally, the feature weight determination unit includes:

[0170] A product matrix determination subunit is used to determine a product determination matrix according to a feature association relationship between contribution features belonging to a product category; and

[0171] An income matrix determination subunit, used to determine an income determination matrix according to a characteristic association relationship between various contribution characteristics belonging to an income category;

[0172] The cost matrix determination subunit is used to determine the cost determination matrix according to the feature association relationship between the contribution features belonging to the cost category;

[0173] The second weight determination subunit is used to determine the contribution feature weight of each contribution feature according to the product determination matrix, the income determination matrix and the cost determination matrix.

[0174] Optionally, the model training module further includes:

[0175] A verification data set acquisition unit, used for performing model training on the recognition model according to the prediction category and the standard category until a preset model training end condition is met and a user recognition model is obtained, and then acquiring a verification data set of a sample verification object;

[0176] A model verification unit, used to perform model verification on the user identification model using the verification data set to obtain a model verification result;

[0177] A feature weight updating unit, used to update the contribution feature weight according to the model verification result to obtain an updated contribution feature weight;

[0178] The model iteration updating unit is used to perform model iteration updating on the user identification model based on the updated contribution feature weights until a preset model iteration updating end condition is met to obtain a target user identification model.

[0179] Optionally, a feature weight updating unit is used to:

[0180] If the model verification result does not meet the preset model verification judgment condition, the product determination matrix is ​​updated according to the product feature boundary range, the income determination matrix is ​​updated according to the income feature boundary range, and the cost determination matrix is ​​updated according to the cost feature boundary range to obtain updated product determination matrix, income determination matrix and cost determination matrix;

[0181] The contribution feature weights are updated according to the updated product determination matrix, the income determination matrix, and the cost determination matrix to obtain updated contribution feature weights.

[0182] Optional, model iteration update unit, specifically used for:

[0183] Iteratively updating the user identification model based on the updated contribution feature weights, and determining the user identification model that meets the preset model iterative update conditions as the target user identification model;

[0184] If the user recognition model does not meet the preset model iteration update end condition within the preset number of iteration updates, multiple reference recognition models under each number of iteration updates are obtained;

[0185] Obtain a test data set of sample test objects;

[0186] Using the test data set to perform model testing on each of the reference recognition models to obtain model testing results of each of the reference recognition models;

[0187] A target user recognition model is determined according to the model test results of each of the reference recognition models.

[0188] The user identification device provided in the embodiment of the present invention can execute the user identification method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0189] Embodiment 4

[0190] Figure 4 A schematic diagram of the structure of an electronic device 40 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0191] like Figure 4 As shown, the electronic device 40 includes at least one processor 41, and a memory connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 41 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 to the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other through a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0192] A number of components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a disk, an optical disk, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0193] The processor 41 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 41 performs the various methods and processes described above, such as a user identification method.

[0194] In some embodiments, the user identification method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 48. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the user identification method described above may be performed. Alternatively, in other embodiments, the processor 41 may be configured to perform the user identification method in any other appropriate manner (e.g., by means of firmware).

[0195] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0196] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0197] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer 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 foregoing.

[0198] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0199] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0200] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0201] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0202] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A user identification method, characterized in that: include: Obtain user contribution feature data of the user to be identified; Inputting the user contribution feature data into a pre-trained user identification model to perform high-quality user identification to obtain a user identification result; The user recognition model is obtained by pre-training a preset recognition network model based on contribution feature sample data and contribution feature weights of a sample training object.

2. The method according to claim 1, characterized in that The model training method of the user identification model is as follows: Acquire contribution feature sample data of a sample training object; the contribution feature sample data includes at least one contribution feature and its corresponding feature parameter value; The sample training objects are pre-labeled with labels of standard categories; Determine contribution feature weights corresponding to each of the contribution features; Inputting the contribution feature sample data and the contribution feature weight into a preset recognition network model to obtain a predicted category output by the model; According to the predicted category and the standard category, the recognition model is trained until a preset model training end condition is met to obtain a user recognition model.

3. The method according to claim 2, characterized in that The determining of contribution feature weights corresponding to the contribution features respectively includes: Determining a feature category of each of the contributing features; For any same feature category, the contribution feature weights of the contribution features under the same feature dimension are determined according to the feature association relationship between the contribution features under the same feature category.

4. The method according to claim 2, characterized in that: The determining of contribution feature weights corresponding to the contribution features respectively includes: Determine a product decision matrix based on the feature association relationship between the contributing features belonging to the product category; and, Determine the income determination matrix based on the characteristic correlation between the contribution characteristics under the income category; Determine the cost determination matrix based on the feature association relationship between the contribution features under the cost category; The contribution feature weights of the contribution features are determined according to the product determination matrix, the revenue determination matrix and the cost determination matrix.

5. The method according to claim 4, characterized in that After the recognition model is trained according to the predicted category and the standard category until a preset model training end condition is met and a user recognition model is obtained, the method further includes: Get the validation data set of the sample validation object; Using the verification data set to perform model verification on the user identification model to obtain a model verification result; According to the model verification result, the contribution feature weight is updated to obtain an updated contribution feature weight; The user identification model is iteratively updated based on the updated contribution feature weights until a preset model iterative update end condition is met to obtain a target user identification model.

6. The method according to claim 5, characterized in that The step of updating the contribution feature weight according to the model verification result to obtain the updated contribution feature weight includes: If the model verification result does not meet the preset model verification judgment condition, the product determination matrix is ​​updated according to the product feature boundary range, the income determination matrix is ​​updated according to the income feature boundary range, and the cost determination matrix is ​​updated according to the cost feature boundary range to obtain updated product determination matrix, income determination matrix and cost determination matrix; The contribution feature weights are updated according to the updated product determination matrix, the income determination matrix, and the cost determination matrix to obtain updated contribution feature weights.

7. The method according to claim 5, characterized in that The user identification model is iteratively updated based on the updated contribution feature weights until a preset model iterative update end condition is met to obtain a target user identification model, including: Iteratively updating the user identification model based on the updated contribution feature weights, and determining the user identification model that meets the preset model iterative update conditions as the target user identification model; If the user recognition model does not meet the preset model iteration update end condition within the preset number of iteration updates, multiple reference recognition models under each number of iteration updates are obtained; Obtain a test data set of sample test objects; Using the test data set to perform model testing on each of the reference recognition models to obtain model testing results of each of the reference recognition models; A target user recognition model is determined according to the model test results of each of the reference recognition models.

8. A user identification device, characterized in that: include: A feature data acquisition module, used to acquire user contribution feature data of the user to be identified; A user identification module, used to input the user contribution feature data into a pre-trained user identification model to perform high-quality user identification and obtain a user identification result; The user recognition model is obtained by pre-training a preset recognition network model based on contribution feature sample data and contribution feature weights of a sample training object.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the user identification method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the user identification method according to any one of claims 1 to 7 when executed.

Images