User classification method and system

Abstract

The present invention provides a user classification method and system, belonging to the field of big data technology. The user classification method includes: obtaining the number of users, the number of user categories and user characteristics; constructing a constraint matrix according to the user characteristics; determining a rank constraint condition according to the number of users and the number of user categories; judging whether the constraint matrix satisfies the rank constraint condition, and classifying the users according to the judgment result. The present invention can quickly and accurately classify users, thereby providing personalized services for various types of users, and improving user experience and user stickiness.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a user classification method and system. Background Art

[0002] The development of big data and artificial intelligence has brought a large number of users to the banking industry. However, this large number of users has brought challenges to the traditional strategies of the banking industry. Although the user base of banks is huge and the products launched are becoming more and more diversified, the number of loyal users has not seen a huge increase. If valuable users can be mined from a large number of users and targeted services can be provided to various types of users, it will be of great help to the banking industry in gaining market share and increasing the number of loyal users.

[0003] As a key technology in data analysis, graph clustering can classify unlabeled samples, greatly reducing the labeling cost. In addition, the model is simple and can classify a large number of users well and screen out valuable users. However, most of the current clustering methods ultimately rely on the traditional K-means method, because the Laplace matrix partitioning rule cannot guarantee the final result, and K-means is needed to complete the final classification. The K-means clustering method requires random selection of the initial cluster center, and continuous iteration around the selected cluster center to complete the clustering. It has great randomness and instability, and the method is particularly sensitive to outliers in the sample. In actual situations, outliers are ubiquitous, which will seriously affect the accuracy of clustering.

[0004] Although many methods have been proposed to optimize K-means, such as binary K-means, K-means++, etc., these methods have brought new parameters and calculation problems, resulting in little improvement in clustering accuracy and performance, and it is impossible to quickly and accurately classify users. Summary of the invention

[0005] The main purpose of the embodiments of the present invention is to provide a user classification method and system to quickly and accurately classify users, thereby providing personalized services for various types of users and improving user experience and user stickiness.

[0006] In order to achieve the above object, an embodiment of the present invention provides a user classification method, including:

[0007] Obtain the number of users, number of user categories, and user characteristics;

[0008] Construct a constraint matrix based on user characteristics;

[0009] Determine the rank constraint condition according to the number of users and the number of user categories;

[0010] Determine whether the constraint matrix satisfies the rank constraint condition and classify the users according to the determination result.

[0011] The embodiment of the present invention further provides a user classification system, including:

[0012] An acquisition unit, used to acquire the number of users, the number of user categories and user characteristics;

[0013] A constraint matrix construction unit, used to construct a constraint matrix according to user characteristics;

[0014] A rank constraint condition determination unit, used to determine the rank constraint condition according to the number of users and the number of user categories;

[0015] The user classification unit is used to determine whether the constraint matrix satisfies the rank constraint condition and classify the users according to the determination result.

[0016] An embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the user classification method when executing the computer program.

[0017] The embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the user classification method are implemented.

[0018] The user classification method and system of the embodiment of the present invention first constructs a constraint matrix according to user characteristics, then determines the rank constraint condition according to the number of users and the number of user categories, and then determines whether the constraint matrix satisfies the rank constraint condition, and classifies the users according to the judgment result. It can quickly and accurately classify users, thereby providing personalized services for various types of users and improving user experience and user stickiness. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the description of the embodiments will be briefly introduced below. 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 paying creative work.

[0020] Figure 1 is a flow chart of a user classification method in an embodiment of the present invention;

[0021] Figure 2 is a flow chart of S103;

[0022] Figure 3 is a flow chart of S201;

[0023] Figure 4is a flow chart of a user classification method in another embodiment of the present invention.

[0024] Figure 5 is a similarity graph corresponding to the similarity matrix in an embodiment of the present invention;

[0025] Figure 6 is a structural block diagram of a user classification system in an embodiment of the present invention;

[0026] Figure 7 It is a structural block diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.

[0028] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0029] In view of the fact that the prior art cannot quickly and accurately classify users, an embodiment of the present invention provides a user classification method, which belongs to a graph clustering method. By citing the Laplace rank constraint method, the division rule of the similar graph is improved, so that it can be more accurately divided, avoiding the subsequent K-means processing, and reducing the inaccuracy caused by the K-means method. The traditional similarity measurement generally adopts a unified rule and uses the inverse function of the Euclidean distance to calculate the similarity. The present invention adopts an adaptive neighbor method, that is, the principle of nearby distribution is adopted. The closer the distance, the greater the similarity of the distribution; the distance is arranged from large to small, and the first k data are taken as neighbors, so that the similarity between objects is more accurate, and the influence from the outside world is weakened. The accuracy and performance of clustering can be improved, and a large number of bank users can be classified more accurately and effectively, and then different marketing strategies are adopted for different categories of users to achieve precision marketing, improve user satisfaction, user experience and user stickiness, thereby improving the conversion rate of loyal users. The present invention is described in detail below in conjunction with the accompanying drawings.

[0030] Figure 1 4 is a flow chart of a user classification method in an embodiment of the present invention. Figure 4 FIG. 1 is a flow chart of a user classification method in another embodiment of the present invention. Figure 1 and Figure 4As shown, the user classification methods include:

[0031] S101: Obtain the number of users, the number of user categories, and user characteristics.

[0032] In specific implementation, age, region, occupation, income, education level, preferences, consumption, transaction frequency and transaction amount can be selected as user characteristics from the user's external attributes, internal attributes and consumption attributes.

[0033] S102: Construct a constraint matrix according to the user characteristics.

[0034] Figure 2 This is a flow chart of S102. Figure 2 As shown, S102 includes:

[0035] S201: Construct a classification input matrix according to user characteristics.

[0036] In the specific implementation, each feature of the user is quantized. Assuming there are n users in total, and each user selects m features, after quantization, an m×n dimensional matrix can be obtained as the classification input matrix X of clustering.

[0037] S202: Determine a similarity objective function according to the classification input matrix, the regularization parameter and the initial similarity matrix.

[0038] Figure 3 This is a flow chart of S202. Figure 3 As shown, S202 includes:

[0039] S301: Determine an initial noise matrix and an initial clean data matrix according to a classification input matrix.

[0040] S302: Determine a noise separation objective function according to an initial noise matrix, an initial clean data matrix and sparse coefficients.

[0041] When the quantization standard is inaccurate, it will affect the classification results. For this reason, we first reduce the dimension of the classification input matrix X to separate the noise. The specific formula involved is as follows:

[0042]

[0043] Among them, ||U|| * +α||E||1 is the noise separation objective function, U is the initial clean data matrix without noise of m×n dimensions, E is the initial noise matrix, and α is the sparse coefficient, which can make the noise matrix E more sparse.

[0044] S303: Determine the initial clean data matrix corresponding to the minimum value of the noise separation objective function as the clean data matrix.

[0045] In specific implementation, the noise matrix caused by quantization can be separated by solving the above formula, ||U|| * The U corresponding to the optimal value (minimum value) of +α||E||1 is the clean data matrix.

[0046] S304: Determine a similarity objective function according to the clean data matrix, the regularization parameter and the initial similarity matrix.

[0047] In specific implementation, an adaptive method can be used to construct a similarity matrix, and the specific formula involved in determining the similarity objective function is as follows:

[0048]

[0049] in, is the similarity objective function, u i is the i-th column of the clean data matrix U, u j is the jth column of the clean data matrix U, s ij is the data (element) of the i-th row and j-th column of the initial similarity matrix S, is the transpose of the i-th column of the initial similarity matrix S (the i-th row of the initial similarity matrix S), is the degree of the node, that is, the data of the i-th row of the initial similarity matrix S, s i is the i-th column of the initial similarity matrix S, and γ is the regularization parameter. Since the degree of a node is 1, when the similarity of a neighbor is 1, the similarities of other points are all 0, and they cannot be regarded as neighbors. In order to avoid this problem, the second regularization term in the above formula is introduced.

[0050] S203: Determine an initial similarity matrix corresponding to the minimum value of the similarity objective function as the similarity matrix.

[0051] In specific implementation, the optimized s can be obtained by solving the above formula i , that is, the similarity matrix S can be obtained, The S corresponding to the optimal value (minimum value) of is the similarity matrix.

[0052] In unsupervised clustering algorithms, objects are usually classified based on their similarity. The greater the similarity between objects of the same type, the smaller the similarity between different objects. Objects are used as vertices and their similarity measures, i.e., similarities, are used as edges to connect them to form a similarity graph. The edges of the similarity graph form a similarity matrix.

[0053] S204: Determine a degree matrix according to the similarity matrix.

[0054] Among them, the degree matrix D S is a diagonal matrix whose diagonal elements are Among them, d ii is the degree matrix D S The diagonal element in the i-th row and i-th column is the sum of the similarities of each vertex in the similarity graph. ji is the data in the j-th row and i-th column of the similarity matrix S.

[0055] S205: The difference between the degree matrix and the similarity matrix is ​​determined as a constraint matrix.

[0056] In one embodiment, the constraint matrix L may be determined according to the following formula to optimize the classification rules: L=DS-S.

[0057] S103: Determine a rank constraint condition according to the number of users and the number of user categories.

[0058] The rank constraint condition is rank(L) = nc, where n is the number of users and c is the number of user categories (the number of clusters).

[0059] S104: Determine whether the constraint matrix satisfies the rank constraint condition, and classify the users according to the determination result.

[0060] In one embodiment, S104 includes:

[0061] Determine whether the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories;

[0062] When the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories, the users are classified according to the similarity matrix; otherwise, the process returns to S202 until the constraint matrix satisfies the rank constraint condition.

[0063] The constraint matrix is ​​a Laplacian matrix, corresponding to the classification rules of the similarity graph. It has a special property, that is, the number of zero eigenvalues ​​in the Laplacian matrix is ​​equal to the number of maximal connected subgraphs in the corresponding similarity graph, and each maximal connected subgraph is a class. Since the rank of the constraint matrix is ​​the same as the number of its non-zero eigenvalues, when a rank constraint is imposed on the constraint matrix, the similarity graph corresponding to the similarity matrix can be divided into exactly c maximal connected subgraphs, that is, n samples can be divided into exactly c classes, completing the classification task without the need for subsequent K-means processing.

[0064] By arranging the above formulas and converting the similarity matrix formula from element mode to matrix mode according to the matrix properties, the overall model of the present invention can be obtained as follows:

[0065]

[0066] Among them, β>0, is the smoothing coefficient of the clean data matrix U.

[0067] Figure 1 The user classification method shown is executed by a computer. Figure 1 As can be seen from the process shown, the user classification method and system of the embodiment of the present invention first constructs a constraint matrix based on user characteristics, then determines the rank constraint condition based on the number of users and the number of user categories, and then determines whether the constraint matrix satisfies the rank constraint condition. The users are classified according to the judgment result. The users can be classified quickly and accurately, thereby providing personalized services for each type of users and improving user experience and user stickiness.

[0068] The specific process of the embodiment of the present invention is as follows:

[0069] 1. Obtain the number of users, number of user categories, and user characteristics.

[0070] For example, if 6 users are divided into two categories, the number of users n = 6, and the number of categories c = 2. The characteristics of the user may include the user's income, the user's age, and the user's region.

[0071] 2. Quantify each feature of the user to construct a classification input matrix.

[0072] For example, if the user's monthly income is less than or equal to 2,000 yuan, the user's income is quantified as 1; if the user's monthly income is greater than 2,000 yuan and less than or equal to 6,000 yuan, the user's income is quantified as 2; if the user's monthly income is greater than 6,000 yuan and less than or equal to 10,000 yuan, the user's income is quantified as 3; if the user's monthly income is greater than 10,000 yuan and less than or equal to 20,000 yuan, the user's income is quantified as 4; if the user's monthly income is greater than or equal to 20,000 yuan, the user's income is quantified as 5.

[0073] Similarly, the user's age and region can be quantified into 7 levels and 5 levels respectively. If the eigenvalues ​​of user A's income, age, and region are 3, 4, and 3 respectively, a column vector (3, 4, 3) can be obtained. When there are 6 users, a 3×6 dimensional matrix can be obtained, namely the classification input matrix X, as follows:

[0074]

[0075] 3. Determine the initial noise matrix and the initial clean data matrix based on the classification input matrix.

[0076] 4. Determine the noise separation objective function according to the initial noise matrix, the initial clean data matrix and the sparse coefficient, and determine that the initial clean data matrix corresponding to the minimum value of the noise separation objective function is the clean data matrix.

[0077] 5. Determine the similarity objective function based on the clean data matrix, the regularization parameter and the initial similarity matrix, and determine the initial similarity matrix corresponding to the minimum value of the similarity objective function as the similarity matrix.

[0078] 6. Determine the degree matrix based on the similarity matrix, and the difference between the degree matrix and the similarity matrix is ​​the constraint matrix.

[0079] 7. Determine the rank constraint based on the number of users and the number of categories.

[0080] The rank constraint condition in this embodiment is rank(L)=nc=6-2=4.

[0081] 8. Determine whether the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories. When the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories, classify the users according to the similarity matrix.

[0082] According to the special properties of the constraint matrix, the similarity graph that satisfies the rank constraint condition rank(L)=4 can have exactly two maximal connected subgraphs, that is, user segmentation can be completed directly without the need for other clustering methods.

[0083] The similarity matrix S is a diagonal matrix with diagonal elements of 0. According to the similarity matrix S, its corresponding similar graph can be obtained. Figure 5 is a similarity graph corresponding to the similarity matrix in the embodiment of the present invention. Figure 5 As shown in the figure, each vertex in the similarity graph represents a different user, and the value on the edge connecting the vertices represents the similarity of information between users. 11 =s 22 =0,s 12 =s 21 =0.8. The greater the similarity, the greater the possibility that the users are in the same category. The similar graph is segmented according to the similarity.

[0084] This embodiment divides the six users into two categories by segmentation: user 1, user 2 and user 3 can be classified into one category, and user 4, user 5 and user 6 can be classified into another category. The virtual frame G1 includes all users in a certain category, and G1 can be called a maximal connected subgraph.

[0085] In summary, the present invention performs dimensionality reduction processing on the classification input matrix X obtained by quantization, filters out the noise interference caused by the quantization factor, constructs the similarity matrix by adopting the strategy of adaptive nearest neighbor, improves the accuracy of the similarity between samples, and uses the rank constraint method to obtain c classification results, avoiding the uncertainty and instability brought by the K-means method. Therefore, compared with the existing clustering algorithms, the present invention has stronger robustness to noise, better clustering accuracy, simplifies the clustering model, and improves the user classification efficiency.

[0086] Based on the same inventive concept, an embodiment of the present invention further provides a user classification system. Since the principle of solving the problem by the system is similar to that of the user classification method, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be repeated.

[0087] Figure 6 is a structural block diagram of a user classification system in an embodiment of the present invention. Figure 6 As shown, the user classification system includes:

[0088] An acquisition unit, used to acquire the number of users, the number of user categories and user characteristics;

[0089] A constraint matrix construction unit, used to construct a constraint matrix according to user characteristics;

[0090] A rank constraint condition determination unit, used to determine the rank constraint condition according to the number of users and the number of user categories;

[0091] The user classification unit is used to determine whether the constraint matrix satisfies the rank constraint condition and classify the users according to the determination result.

[0092] In one embodiment, the constraint matrix construction unit is specifically used for:

[0093] Construct a classification input matrix based on user characteristics;

[0094] Determine a similarity objective function according to the classification input matrix, the regularization parameter and the initial similarity matrix;

[0095] Determine that the initial similarity matrix corresponding to the minimum value of the similarity objective function is the similarity matrix;

[0096] Determine the degree matrix according to the similarity matrix;

[0097] The difference between the certainty matrix and the similarity matrix is ​​the constraint matrix.

[0098] In one embodiment, the user classification unit is specifically used for:

[0099] Determine whether the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories;

[0100] When the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories, the users are classified according to the similarity matrix.

[0101] In one embodiment, the constraint matrix construction unit is specifically used for:

[0102] Determine an initial noise matrix and an initial clean data matrix according to the classification input matrix;

[0103] Determine a noise separation objective function according to an initial noise matrix, an initial clean data matrix and sparse coefficients;

[0104] Determine the initial clean data matrix corresponding to the minimum value of the noise separation objective function as the clean data matrix;

[0105] The similarity objective function is determined based on the clean data matrix, the regularization parameter and the initial similarity matrix.

[0106] In summary, the user classification system of the embodiment of the present invention first constructs a constraint matrix based on user characteristics, then determines the rank constraint conditions based on the number of users and the number of user categories, and then determines whether the constraint matrix satisfies the rank constraint conditions. The users are classified according to the judgment results. The users can be classified quickly and accurately, thereby providing personalized services for various types of users and improving user experience and user stickiness.

[0107] The embodiment of the present invention also provides a specific implementation of a computer device capable of implementing all the steps in the user classification method in the above embodiment. Figure 7 is a block diagram of a computer device in an embodiment of the present invention, see Figure 7 , the computer equipment specifically includes the following contents:

[0108] Processor (processor) 701 and memory (memory) 702.

[0109] The processor 701 is used to call the computer program in the memory 702. When the processor executes the computer program, all steps in the user classification method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0110] Obtain the number of users, number of user categories, and user characteristics;

[0111] Construct a constraint matrix based on user characteristics;

[0112] Determine the rank constraint condition according to the number of users and the number of user categories;

[0113] Determine whether the constraint matrix satisfies the rank constraint condition and classify the users according to the determination result.

[0114] In summary, the computer device of the embodiment of the present invention first constructs a constraint matrix according to user characteristics, and then determines the rank constraint conditions according to the number of users and the number of user categories, and then determines whether the constraint matrix satisfies the rank constraint conditions, and classifies the users according to the judgment results. It can quickly and accurately classify users, thereby providing personalized services for various types of users and improving user experience and user stickiness.

[0115] The embodiment of the present invention also provides a computer-readable storage medium capable of implementing all the steps in the user classification method in the above embodiment. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, all the steps of the user classification method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0116] Obtain the number of users, number of user categories, and user characteristics;

[0117] Construct a constraint matrix based on user characteristics;

[0118] Determine the rank constraint condition according to the number of users and the number of user categories;

[0119] Determine whether the constraint matrix satisfies the rank constraint condition and classify the users according to the determination result.

[0120] In summary, the computer-readable storage medium of an embodiment of the present invention first constructs a constraint matrix based on user characteristics, then determines the rank constraint conditions based on the number of users and the number of user categories, and then determines whether the constraint matrix satisfies the rank constraint conditions, and classifies the users according to the judgment results. This allows users to be classified quickly and accurately, thereby providing personalized services for each type of user and improving user experience and user stickiness.

[0121] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0122] Those skilled in the art may also understand that the various illustrative logical blocks, units, and steps listed in the embodiments of the present invention may be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly demonstrate the interchangeability of hardware and software, the various illustrative components, units, and steps described above have generally described their functions. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be understood as exceeding the scope of protection of the embodiments of the present invention.

[0123] The various illustrative logic blocks, or units, or devices described in the embodiments of the present invention can be implemented or operated by a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component, or any combination of the above. The general-purpose processor can be a microprocessor, and optionally, the general-purpose processor can also be any conventional processor, controller, microcontroller or state machine. The processor can also be implemented by a combination of computing devices, such as a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.

[0124] The steps of the method or algorithm described in the embodiments of the present invention can be directly embedded in hardware, a software module executed by a processor, or a combination of the two. The software module can be stored in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or other storage media of any form in the art. Exemplarily, the storage medium can be connected to the processor so that the processor can read information from the storage medium and can write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. The processor and the storage medium can be arranged in an ASIC, and the ASIC can be arranged in a user terminal. Optionally, the processor and the storage medium can also be arranged in different components in the user terminal.

[0125] In one or more exemplary designs, the above functions described in the embodiments of the present invention can be implemented in hardware, software, firmware or any combination of the three. If implemented in software, these functions can be stored on a computer-readable medium, or transmitted in the form of one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media that facilitate the transfer of computer programs from one place to another. The storage medium can be any available medium that can be accessed by any general or special computer. For example, such computer-readable media can include but are not limited to RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage devices, or any other medium that can be used to carry or store program codes in the form of instructions or data structures and other forms that can be read by general or special computers, or general or special processors. In addition, any connection can be appropriately defined as a computer-readable medium, for example, if the software is transmitted from a website site, server or other remote resource through a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL) or wirelessly, such as infrared, wireless and microwave, it is also included in the defined computer-readable medium. The disk and disc include compact disk, laser disk, optical disk, DVD, floppy disk and Blu-ray disk. Disks usually copy data magnetically, while discs usually copy data optically with lasers. The above combination can also be included in computer readable media.

Claims

1. A user classification method, characterized in that: include: Obtain the number of users, number of user categories, and user characteristics; constructing a constraint matrix according to the user characteristics; The constructing a constraint matrix according to the user characteristics comprises: constructing a classification input matrix according to the user characteristics; Determining a similarity objective function according to the classification input matrix, the regularization parameter and the initial similarity matrix; The similarity objective function is u i is the i-th column of the clean data matrix U, u j is the jth column of the clean data matrix U, s ij is the data of the i-th row and j-th column of the initial similarity matrix S, and γ is the regularization parameter; Determine an initial similarity matrix corresponding to the minimum value of the similarity objective function as a similarity matrix; The similarity matrix is ​​S corresponding to the optimal value of the similarity objective function; Determine a degree matrix according to the similarity matrix; Determine the difference between the degree matrix and the similarity matrix as a constraint matrix; The determining of the similarity objective function according to the classification input matrix, the regularization parameter and the initial similarity matrix comprises: Determining an initial noise matrix and an initial clean data matrix according to the classification input matrix; Determine a noise separation objective function according to the initial noise matrix, the initial clean data matrix and the sparse coefficients; The noise separation objective function is ||U|| * +α||E||1, U is the initial clean data matrix without noise of m×n dimensions, E is the initial noise matrix, and α is the sparse coefficient; Determine an initial clean data matrix corresponding to the minimum value of the noise separation objective function as a clean data matrix; Determining a similarity objective function according to the clean data matrix, the regularization parameter and the initial similarity matrix; Determining a rank constraint condition according to the number of users and the number of user categories; Determining whether the constraint matrix satisfies a rank constraint condition, and classifying the user according to the determination result, including: Determining whether the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories; When the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories, the users are classified according to the similarity matrix.

2. A user classification system, characterized in that: include: An acquisition unit, used to acquire the number of users, the number of user categories and user characteristics; A constraint matrix construction unit, used to construct a constraint matrix according to the user characteristics; The constraint matrix construction unit is specifically used for: constructing a classification input matrix according to the user characteristics; Determining a similarity objective function according to the classification input matrix, the regularization parameter and the initial similarity matrix; The similarity objective function is u i is the i-th column of the clean data matrix U, u j is the jth column of the clean data matrix U, s ij is the data of the i-th row and j-th column of the initial similarity matrix S, and γ is the regularization parameter; Determine an initial similarity matrix corresponding to the minimum value of the similarity objective function as a similarity matrix; The similarity matrix is ​​S corresponding to the optimal value of the similarity objective function; Determine a degree matrix according to the similarity matrix; Determine the difference between the degree matrix and the similarity matrix as a constraint matrix; The constraint matrix construction unit is specifically used for: Determining an initial noise matrix and an initial clean data matrix according to the classification input matrix; Determine a noise separation objective function according to the initial noise matrix, the initial clean data matrix and the sparse coefficients; The noise separation objective function is ||U|| * +α||E||1, U is the initial clean data matrix without noise of m×n dimensions, E is the initial noise matrix, and α is the sparse coefficient; Determine an initial clean data matrix corresponding to the minimum value of the noise separation objective function as a clean data matrix; Determining a similarity objective function according to the clean data matrix, the regularization parameter and the initial similarity matrix; A rank constraint condition determination unit, configured to determine a rank constraint condition according to the number of users and the number of user categories; A user classification unit, used to determine whether the constraint matrix satisfies the rank constraint condition, and classify the user according to the determination result; The user classification unit is specifically used for: Determining whether the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories; When the rank of the constraint matrix is ​​equal to the difference between the number of users and the number of user categories, the users are classified according to the similarity matrix.

3. A computer device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the user classification method according to claim 1 are implemented.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the user classification method according to claim 1 are implemented.

Images