User classification method, device, equipment and storage medium

Abstract

The present application relates to the field of artificial intelligence technology and discloses a method, apparatus, device and storage medium for user classification. The method fills missing data items in acquired user data items, and inputs the filled special data items and regular data items into a trained primary classification model for classification processing to obtain a primary classification result. Since the inclusion of special data items will lead to deviations in the final classification result, the primary classification result is weighted accordingly and weighted summed to obtain the user classification result, thereby improving the stability of user classification and ensuring the accuracy of the classification result.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer equipment, and storage medium for user classification. Background Art

[0002] By rationally collecting user data and categorizing and scoring it, we can improve the user credit system and understand user needs and the corresponding risks of handling business. When obtaining large amounts of user data as variables for classification, the problem of the acquired user data types being solidified can easily arise. When adding multiple new user data types, the information value of the newly added user data types varies, resulting in reduced classification stability and affecting the accuracy of the scoring. Summary of the Invention

[0003] The present application provides a method, apparatus, computer device and storage medium for user classification, which solves the problem in the prior art of reduced stability of user classification when adding new user data types to improve user classification.

[0004] In a first aspect, an embodiment of the present application provides a method for user classification, including:

[0005] Acquire a user data item set, identify missing data items and regular data items in the user data item set, and obtain a plurality of missing data items and a plurality of regular data items;

[0006] Filling the missing data items into a plurality of special data items by a special value filling method;

[0007] Inputting the plurality of special data items and the plurality of regular data items into the trained primary classification model for primary classification processing to obtain a primary classification result set, wherein the primary classification result set includes a plurality of primary classification results;

[0008] After weight adjustment is performed on each of the initial classification results in the initial classification result set using the trained user classification model, the weighted summation algorithm is used for processing to obtain a user classification result.

[0009] In a second aspect, an embodiment of the present application further provides a user classification device, including:

[0010] an acquisition module, which acquires a user data item set, identifies missing data items and regular data items in the user data item set, and obtains a plurality of missing data items and a plurality of regular data items;

[0011] a missing data filling module, which fills the missing data items into a plurality of special data items by a special value filling method;

[0012] a primary classification module, inputting the plurality of special data items and the plurality of regular data items into the trained primary classification model for primary classification processing to obtain a primary classification result set, wherein the primary classification result set includes a plurality of primary classification results;

[0013] The user classification module adjusts the weight of each of the initial classification results in the initial classification result set using the trained user classification model, and processes the results using a weighted summation algorithm to obtain a user classification result.

[0014] In a third aspect, an embodiment of the present application further provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned user classification method when executing the computer program.

[0015] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned user classification method are implemented.

[0016] The user classification method, device, computer equipment and storage medium provided by the present application, among which, a user classification method is provided, wherein the missing data items in the acquired user data items are filled, and the filled special data items and regular data items are input into the trained initial classification model for classification processing to obtain the initial classification result; since the inclusion of special data items will lead to deviation in the final classification result, the initial classification result is weighted accordingly and the weighted sum is taken to obtain the user classification result, thereby improving the stability of the user classification and ensuring the accuracy of the classification result. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 This is a schematic diagram of an application environment of a user classification method provided by an embodiment of the present application;

[0019] Figure 2 This is a flowchart of a method for implementing user classification provided by an embodiment of the present application;

[0020] Figure 3This is a flowchart of step S30 in a user classification method provided in one embodiment of the present application;

[0021] Figure 4 This is a flowchart of step S301 in a user classification method provided in one embodiment of the present application;

[0022] Figure 5 This is a schematic diagram of the structure of a device for user classification provided in one embodiment of the present application;

[0023] Figure 6 Schematic diagram of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0025] The user classification method provided in the embodiment of the present application can be applied in the following situations: Figure 1 In the application environment shown. Figure 1 As shown, a client (computer device) communicates with a server via a network. The client (computer device) includes, but is not limited to, various personal computers, laptops, smartphones, tablets, cameras, and portable wearable devices. The server can be a standalone server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0026] The user classification method provided in this embodiment can be executed by the server. For example, the client sends a set of user data items to the server. The server executes the user classification method provided in this embodiment based on the set of user data items, and then obtains the user classification result after user classification processing, and finally sends the user classification result to the client.

[0027] exist Figure 1 In some other scenarios, the user classification method can also be executed by the client, directly based on the user data item set of the client, by executing the user classification method provided in this embodiment to obtain the user classification result after user classification processing, and then send the user classification result to the server for storage.

[0028] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0029] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0030] Figure 2 FIG1 shows a flow chart of the user classification method provided by an embodiment of the present application. Figure 2 As shown, a method for user classification is provided, and its technical solution mainly includes the following steps S10-S40:

[0031] Step S10: Acquire a user data item set, identify missing data items and regular data items in the user data item set, and obtain a plurality of missing data items and a plurality of regular data items.

[0032] In step S10, in order to obtain more accurate classification results, it is generally necessary to obtain more comprehensive user data to improve the accuracy of the user classification results. However, when the scope of user data collection is expanded, some items will often be missing data. Therefore, the acquired data is divided into missing data items and regular data items that have been successfully collected.

[0033] In one embodiment, user data includes, but is not limited to, social network data, behavioral preferences, credit history, and relationships. Missing user data often results from information being temporarily unavailable, such as the lagged effect of a product's revenue, or the unavailability of one or more attributes of certain objects, such as the spouse's name of an unmarried person or the fixed income status of a child. Missing data can exacerbate the uncertainty exhibited by the system, making it more difficult to grasp the inherent deterministic elements of the system. Data containing null values ​​can disrupt the data mining process, leading to unreliable output.

[0034] In one embodiment, obtaining user original data specifically includes but is not limited to some basic data and business data of the user's use of the system; collecting user behavior data through methods such as embedding points and probes, and obtaining user data in third-party systems through cooperation with third-party merchants.

[0035] Step S20: Fill the missing data items into a plurality of special data items using a special value filling method.

[0036] In step S20, some project data has a certain impact on the classification results, which will make the uncertainty shown in the system more significant and the deterministic components contained in the system more difficult to grasp. When this part of the project data is missing, it will be improved by filling in the missing data items, filling in the missing data items with special data items, and then input into the trained primary classification model for initial classification processing.

[0037] In one embodiment, missing data items can be deleted or filled based on the risk profile of the missing data. Filling methods include special value filling, average value filling, hot-card filling, and multiple imputation. This embodiment utilizes special value filling, which assumes that null values ​​in data also contain information. They are null because they are different from any other data. Therefore, null values ​​are treated as special attribute values, different from any other attribute values. For example, all null values ​​are filled with "unknown."

[0038] In another embodiment, missing data can also be filled by machine learning, such as random forest filling of missing values ​​and Bayesian filling of missing values. The general idea is to use the correlation between different project data to fill in the missing data. For example, we can use "region", "transportation", and "number of school districts" to predict "housing prices", and conversely, we can also use "housing prices", "transportation", and "number of school districts" to predict "region".

[0039] Step S30: Input the plurality of special data items and the plurality of regular data items into the trained primary classification model for primary classification processing to obtain a primary classification result set, wherein the primary classification result set includes multiple primary classification results.

[0040] In step S30, the trained primary classification model is used to perform an initial processing on the regular data items and the missing data items, and the corresponding initial classification results are obtained. The trained primary classification model is more stable and the classification results are more accurate. Different methods are used to process the regular data items and the missing data items, thereby improving the accuracy of the classification results.

[0041] like Figure 3 As shown, in step S30 of inputting the plurality of special data items and the plurality of regular data items into the trained primary classification model for primary classification processing to obtain a primary classification result set, wherein the primary classification result set includes the plurality of primary classification results, the following steps are included:

[0042] Step S301, performing a primary classification process according to the IV value of each of the regular data items using the trained primary classification model to obtain a regular data primary classification result corresponding to the regular data item;

[0043] Step S302, performing a primary classification process on each of the special data items using the WOE encoding unit in the trained primary classification model to obtain a primary classification result of the special data corresponding to the special data item;

[0044] Step S303: Integrate the initial classification results of each of the special data and the initial classification results of each of the regular data to obtain a set of initial classification results.

[0045] In step S301 , in processing regular data items, since different data items have different IV values, feature enhancement can be performed on data items with weak IV values, thereby improving the utilization of data items with weak IV values.

[0046] like Figure 4 As shown, in step S301 of performing a primary classification process according to the IV value of each of the regular data items using the trained primary classification model to obtain the regular data primary classification result corresponding to the regular data item, the following steps are included:

[0047] Step S3011: Classify the regular data items according to their IV values ​​to obtain strong IV value data items and weak IV value data items;

[0048] Step S3012: using the machine learning unit in the trained primary classification model to perform feature enhancement processing on each of the weak IV value data items;

[0049] Step S3013: Use the scoring card unit in the trained primary classification model to perform initial classification processing on the weak IV value data items and strong IV value data items that have undergone feature enhancement processing to obtain the initial classification results of the regular data.

[0050] In one embodiment, the primary classification model includes a scoring card unit and a machine learning unit. Since the IV value (Information value) of each user data is different, the scoring card unit can be directly used to process the strong IV value data items with stronger IV values, thereby obtaining the initial classification results corresponding to the strong IV value data items. In this embodiment, the weak IV value data items with weaker IV values ​​are enhanced by using the machine learning unit to perform feature enhancement processing, thereby improving the IV value of the weak IV value data items, and the weak IV value after feature enhancement processing is used for initial classification processing using the scoring card unit.

[0051] In another embodiment, the machine learning unit can utilize the extreme gradient boosting algorithm (eXtreme GradientBoosting, XGB), random forest algorithm (Random forest, RF), decision tree algorithm (Decision Tree) and k-nearest neighbor algorithm (KNN). In one embodiment, the XGB algorithm is used to construct a machine learning unit as follows: multiple tree models are constructed using multiple weak IV value data items. The gradient boosting tree initially uses all the data to construct a regression tree model. Since the tree model will eventually output the mean in each region, the difference between the original data in each region and the mean, i.e., the residual, is calculated, and a new regression tree model is continued to be established using these residuals. The new model tree will also generate residuals, and the previous steps are repeated until the residual is less than a given threshold or the number of iterations reaches a specified number of times. After multiple iterations, the obtained model contains multiple regression tree models. New data is input into the model, and the output result is the sum of the results of each regression tree.

[0052] Understandably, the calculation of IV (Information Value) is based on WOE. The WOE values ​​of the independent variables are weighted and summed to measure the value of information. WOE (Weight of Evidence) is a form of encoding of the original features. To encode a feature using WOE, the continuous variable needs to be discretized, an operation known as grouping or binning. The WOE value is then calculated by taking the difference between the proportion of positive samples in the current group to all positive samples and the proportion of negative samples in the current group to all negative samples.

[0053] A scorecard is a classification model that categorizes multiple user data items and represents them as a score. Scorecards can be used to score credit customers, and can be used for pre-loan, mid-loan, post-loan, and anti-fraud purposes, depending on the business scenario. They are generally referred to as ABCF cards. The model generates scores, and evaluation results are generated by setting cutoff thresholds. These results can be used for policy application. Unlike machine learning models, scorecards use logistic regression, which is a generalized linear regression and offers strong interpretability of features.

[0054] Understandably, XGBoost (eXtreme Gradient Boosting) is an algorithm based on GBDT. XGBoost makes some optimizations to GBDT, such as using second-order derivatives to make the loss function more accurate, regularization terms to prevent tree overfitting, and block storage to enable parallel computing. XGBoost is efficient, flexible, and lightweight, and is widely used in fields such as data mining and recommendation systems.

[0055] In step S302, the special data items corresponding to the missing data items are processed using WOE coding, and the special data items are individually WOE-coded as a feature, thereby reducing the uncertainty caused by the missing items after filling, and obtaining the initial classification results of the special data corresponding to the special data items.

[0056] Understandably, WOE (Weight of Evidence) is a coding method for scorecard units. The larger the absolute value of the WOE, the more significant the difference, while the smaller the absolute value, the less significant the difference. If the WOE is 0, the ratio of good to bad customers in the bin is equal to the random ratio of bad customers to good customers, and the bin has no predictive power.

[0057] In step S303, the initial classification results of each special data and the initial classification results of each regular data are integrated to obtain a set of initial classification results, which is input into the user classification model and used for user classification processing.

[0058] Step S40: After weighting each of the initial classification results in the initial classification result set is adjusted using the trained user classification model, a weighted summation algorithm is used to process the results to obtain a user classification result.

[0059] In step S40, user classification is performed based on several preliminary classification results in the preliminary classification result set to obtain more accurate user classification results. The preliminary classification result set corresponding to missing data items in the user data item set is processed by the trained user classification model. Because the variance of the distribution of each preliminary classification result in the corresponding preliminary classification result set decreases when the acquired user data item set contains missing data items, resulting in a decrease in the stability of the user classification process, the user classification model is required to adjust the weights of each preliminary classification result during the user classification process to improve the stability of the user classification process.

[0060] In one embodiment, the normalized value of the normalized data is obtained by normalizing the results of the initial classification of the normalized data. The normalized value of the normalized data is obtained by normalizing the results of the initial classification of the normalized data using the following formula: τ = ∑i ω i σ i , where τ is the normalized value of the conventional data, ω i is the weight of the i-th initial classification result, σ i is the standard deviation of the initial classification results of regular data, and i is the number of initial classification results of regular data.

[0061] Normalization is performed on multiple initial classification results to obtain a normalized value of all data, wherein the initial classification results are normalized by the following formula to obtain the normalized value of all data, the formula is τ′=∑ j ω j σ j , where τ′ is the normalized value of all the data, ω j is the weight of the jth initial classification result, σ j is the standard deviation of the initial classification results, and j is the number of initial classification results.

[0062] The weight of the initial classification result is adjusted using the ratio of the normalized value of the conventional data to the normalized value of all data, and the user classification result is calculated using a weighted summation algorithm. The mathematical expression of the user classification processing in the user classification model is: Where Score is the user classification result, τ is the normalized value of the regular data, τ′ is the normalized value of all data, S j is the jth initial classification result, σ j is the standard deviation of the initial classification results, ω j is the weight of the jth initial classification result, and j is the number of initial classification results.

[0063] In another embodiment, if the acquired user data item set does not include missing data items, the user classification model can be fitted using a linear model, such as a Logistics regression algorithm, which has a stronger classification interpretability and is convenient for business applications. In this embodiment, the primary classification result set obtained after the user data item set is subjected to the initial classification processing is weighted summed to obtain the user classification result. Among them, the user classification model can be fitted using a linear model, such as a Logistics regression algorithm, which has a stronger classification interpretability and is convenient for business applications. The mathematical expression of the user classification model is Score = ∑ j ω j S j , where Score is the user classification result, S j is the jth initial classification result, ω j is the weight of the jth initial classification result, and j is the number of initial classification results.

[0064] As you can understand, logistic regression is a classification model that is similar to multiple linear regression and is a generalized linear model. The difference lies in the dependent variable. If it is continuous, it is multiple linear regression. If it is binomial distributed, it is logistic regression. If it is Poisson distributed, it is Poisson regression. If it is negative binomial distributed, it is negative binomial regression. The dependent variable of logistic regression can be a binary nonlinear difference equation or a multi-classification. The logistic regression algorithm can be divided into binary logistic regression and multi-classification logistic regression according to whether the dependent variable is binary or multi-classification. When the dependent variable is a binary variable, the analysis can be performed using unconditional logistic regression or conditional logistic regression models, depending on the design type.

[0065] In one embodiment, the final rating result may be user rating, user classification, data classification, etc.

[0066] Step S51: Input the user data item sample set into the primary classification model to obtain a corresponding initial classification result sample set; input the initial classification result sample set into the user classification model, and based on the sample user classification results corresponding to the initial classification result sample set, test the correlation between each of the initial classification result samples and the weight of each of the initial classification result samples in the user classification process.

[0067] In step S51, the user classification model utilizes a weighted summation in the final user classification process. This weighted summation requires the weight of each initial classification result corresponding to each initial classification result. The weight of each initial classification result is calculated using historical samples. The more samples used for training, the more stable the user classification process. The primary classification model is trained based on a set of user data item samples. The correlation between the initial classification result samples and the weight of each initial classification result sample in the user classification process are tested.

[0068] In one embodiment, the user classification model determines the weight of each initial classification result sample in the user classification process through a decision tree algorithm, and inputs each initial classification result sample set into the decision tree algorithm for training. The decision tree algorithm continuously selects entropy reduction according to the classification results in each initial classification result sample, and after training each initial classification result sample, the test data can be retrieved according to the judgment conditions of the classification result until the weight of the initial classification result sample corresponding to the classification result is finally determined.

[0069] In another embodiment, the method of calculating the weight of the user classification model further includes using a linear regression algorithm, a factor analysis method, and a principal component method.

[0070] Understandably, the decision tree algorithm trains by feeding the training data based on the selected features into the algorithm. The algorithm continuously selects the partitioning method that maximizes entropy reduction to distinguish samples, iterating until all sub-branch branches have the same set of sample labels. After training, test data can be retrieved according to the judgment criteria of the branch nodes until the final leaf node is reached. The label of this leaf is returned as the label threshold for the test data.

[0071] In one embodiment, a user classification device is provided, which corresponds to the user classification method in the above embodiment. Figure 5 As shown, the user classification device includes an acquisition module 11, a missing data filling module 12, a primary classification module 13 and a user classification module 14. The detailed description of each functional module is as follows:

[0072] An acquisition module 11 acquires a user data item set, identifies missing data items and regular data items in the user data item set, and obtains a plurality of missing data items and a plurality of regular data items;

[0073] A missing data filling module 12 is configured to fill the missing data items into a plurality of special data items using a special value filling method;

[0074] A primary classification module 13 inputs the plurality of special data items and the plurality of regular data items into a trained primary classification model for primary classification processing to obtain a primary classification result set, wherein the primary classification result set includes a plurality of primary classification results;

[0075] The user classification module 14 adjusts the weight of each of the initial classification results in the initial classification result set using the trained user classification model, and processes the results using a weighted summation algorithm to obtain a user classification result.

[0076] For the specific definition of the user classification device, please refer to the definition of the user classification method above and will not be repeated here. Each module in the above-mentioned user classification device can be implemented in whole or in part through software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each of the above modules.

[0077] In one embodiment, a computer device is provided. The computer device can be a client or a server. The internal structure diagram thereof can be as follows: Figure 6As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium and an internal memory. The readable storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the readable storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When executed by the processor, the computer program implements a method for user classification.

[0078] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the user classification method in the above embodiment is implemented.

[0079] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for user classification in the above embodiment is implemented.

[0080] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0081] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0082] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A method for user classification, characterized in that: include: Acquire a user data item set, identify missing data items and regular data items in the user data item set, and obtain a plurality of missing data items and a plurality of regular data items; Filling the missing data items into a plurality of special data items by a special value filling method; Inputting the plurality of special data items and the plurality of regular data items into the trained primary classification model for primary classification processing to obtain a primary classification result set, wherein the primary classification result set includes a plurality of primary classification results; After weighting each of the initial classification results in the initial classification result set is adjusted by the trained user classification model, the weighted summation algorithm is used to process the results to obtain a user classification result; The user classification model after training adjusts the weight of each of the initial classification results in the initial classification result set, and then processes the results using a weighted summation algorithm to obtain a user classification result, including: Normalizing the initial classification results of multiple conventional data to obtain normalized values ​​of conventional data; Normalize multiple initial classification results to obtain normalized values ​​of all data; Adjusting the weight of the initial classification result using the ratio of the normalized value of the regular data to the normalized value of all data, and calculating the user classification result using a weighted sum algorithm; The method of adjusting the weight of the initial classification result by using the ratio of the normalized value of the regular data to the normalized value of all data, and calculating the user classification result by using a weighted summation algorithm, includes: The normalized value of the conventional data is obtained by normalizing the initial classification result of the conventional data using the following formula: = ,in is the normalized value of the conventional data, is the weight of the i-th initial classification result, is the standard deviation of the initial classification results of conventional data, and i is the number of initial classification results of conventional data; The initial classification results are normalized using the following formula to obtain the normalized value of all data: = ,in is the normalized value of all the data, is the weight of the jth initial classification result, is the standard deviation of the initial classification results, j is the number of initial classification results; The mathematical expression of user classification processing in the user classification model is: = ,in Classification results for the user, is the normalized value of the conventional data, is the normalized value of all the data, is the jth initial classification result, is the standard deviation of the initial classification results, is the weight of the jth initial classification result, and j is the number of initial classification results.

2. The user classification method according to claim 1, wherein: The inputting of the plurality of special data items and the plurality of regular data items into the trained primary classification model for primary classification processing to obtain a primary classification result set including a plurality of primary classification results, including: Performing a primary classification process according to the IV value of each of the conventional data items using the trained primary classification model to obtain a primary classification result of conventional data corresponding to the conventional data item; Performing a primary classification process on each of the special data items using the WOE encoding unit in the trained primary classification model to obtain a primary classification result of the special data corresponding to the special data item; The initial classification results of each of the special data and the initial classification results of each of the regular data are integrated to obtain a set of initial classification results.

3. The user classification method according to claim 2, wherein: The trained primary classification model performs a primary classification process according to the IV value of each of the regular data items to obtain a regular data primary classification result corresponding to the regular data item, including: Classify the regular data items according to the IV value of each regular data item to obtain strong IV value data items and weak IV value data items; Performing feature enhancement processing on each of the weak IV value data items using a machine learning unit in the trained primary classification model; The weak IV value data items and strong IV value data items subjected to the feature enhancement processing are subjected to initial classification processing using the score card unit in the trained primary classification model to obtain the initial classification result of the conventional data.

4. The user classification method according to claim 1, wherein: After the steps of acquiring a user data item set, identifying missing data items and regular data items in the user data item set, and obtaining a plurality of missing data items and a plurality of regular data items, the method further includes: If the acquired user data item set does not include missing data items, a weighted sum is performed on a primary classification result set obtained after the user data item set is initially classified, thereby obtaining a user classification result.

5. The user classification method according to claim 1, wherein: Before the step of inputting the plurality of special data items and the plurality of regular data items into the trained primary classification model for primary classification processing to obtain a primary classification result set including a plurality of primary classification results, the method further includes: Input the user data item sample set into the primary classification model to obtain the corresponding initial classification result sample set; input the initial classification result sample set into the user classification model, and test the weight of each initial classification result sample in the user classification process based on the sample user classification results corresponding to the initial classification result sample set.

6. A user classification device, characterized in that: include: an acquisition module, which acquires a user data item set, identifies missing data items and regular data items in the user data item set, and obtains a plurality of missing data items and a plurality of regular data items; a missing data filling module, which fills the missing data items into a plurality of special data items by a special value filling method; a primary classification module, inputting the plurality of special data items and the plurality of regular data items into the trained primary classification model for primary classification processing to obtain a primary classification result set, wherein the primary classification result set includes a plurality of primary classification results; A user classification module, which adjusts the weight of each of the initial classification results in the initial classification result set using a trained user classification model, and processes the results using a weighted summation algorithm to obtain a user classification result; The user classification model after training adjusts the weight of each of the initial classification results in the initial classification result set, and then processes the results using a weighted summation algorithm to obtain a user classification result, including: Normalizing the initial classification results of multiple conventional data to obtain normalized values ​​of conventional data; Normalize multiple initial classification results to obtain normalized values ​​of all data; Adjusting the weight of the initial classification result using the ratio of the normalized value of the regular data to the normalized value of all data, and calculating the user classification result using a weighted sum algorithm; The method of adjusting the weight of the initial classification result by using the ratio of the normalized value of the regular data to the normalized value of all data, and calculating the user classification result by using a weighted summation algorithm, includes: The normalized value of the conventional data is obtained by normalizing the initial classification result of the conventional data using the following formula: = ,in is the normalized value of the conventional data, is the weight of the i-th initial classification result, is the standard deviation of the initial classification results of conventional data, and i is the number of initial classification results of conventional data; The initial classification results are normalized using the following formula to obtain the normalized value of all data: = ,in is the normalized value of all the data, is the weight of the jth initial classification result, is the standard deviation of the initial classification results, j is the number of initial classification results; The mathematical expression of user classification processing in the user classification model is: = ,in Classification results for the user, is the normalized value of the conventional data, is the normalized value of all the data, is the jth initial classification result, is the standard deviation of the initial classification results, is the weight of the jth initial classification result, and j is the number of initial classification results.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the user classification method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for user classification according to any one of claims 1 to 5 is implemented.

Images