User classification method, apparatus and computer readable storage medium

Abstract

This disclosure relates to a user classification method, apparatus, and computer-readable storage medium, and pertains to the field of computer technology. The method includes: calculating the correlation between various types of user feature data; generating user feature vectors based on feature data whose correlation is less than a threshold; and performing clustering processing on the feature vectors of each user to classify each user into multiple user categories.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a user classification method, a user classification device, and a computer-readable storage medium. Background Technology

[0002] To provide better services to users, various industries need to mine user data. For example, in the telecommunications industry, incremental data mining and storage to match user needs with the services provided is an important research direction.

[0003] In related technologies, the Hard C Means algorithm is used to mine user data in order to determine user classification. Summary of the Invention

[0004] The inventors of this disclosure have discovered the following problem in the aforementioned related technologies: the data mining is not accurate enough, resulting in poor accuracy in user classification.

[0005] In view of this, this disclosure proposes a user classification technology solution that can improve the accuracy of user classification.

[0006] According to some embodiments of this disclosure, a user classification method is provided, including: calculating the correlation between various types of user feature data; generating user feature vectors based on feature data with correlation less than a threshold; and performing clustering processing on the feature vectors of each user to classify each user into multiple user categories.

[0007] In some embodiments, generating a user's feature vector based on feature data whose relevance is less than a threshold includes: generating an initial feature vector for the user based on feature data whose relevance is less than a threshold; and performing dimensionality-upgrading processing on the initial feature vector using a Gaussian kernel function to generate a new feature vector.

[0008] In some embodiments, clustering the feature vectors of each user and dividing each user into multiple user categories includes: determining the number of clustering categories based on the number of service categories that can be provided to the user; selecting the feature vectors of a corresponding number of users as initial cluster centers based on the number of categories; and performing clustering based on the initial cluster centers and the feature vectors of each user.

[0009] In some embodiments, clustering based on initial cluster centers and feature vectors of each user includes: optimizing the initial cluster centers using a particle swarm optimization algorithm to determine optimized cluster centers; and performing clustering based on the optimized cluster centers and feature vectors of each user.

[0010] In some embodiments, the method further includes: determining the business category corresponding to each user category based on the feature data of each user in each user category, so as to provide business services according to the corresponding business category of each user.

[0011] According to some other embodiments of this disclosure, a user classification apparatus is provided, comprising: a calculation unit for calculating the correlation between various types of user feature data; a generation unit for generating user feature vectors based on feature data whose correlation is less than a threshold; and a clustering unit for performing clustering processing on the feature vectors of each user to classify each user into multiple user categories.

[0012] In some embodiments, the generation unit generates an initial feature vector for the user based on feature data with a correlation level less than a threshold, and uses a Gaussian kernel function to perform dimensionality increase processing on the initial feature vector to generate a new feature vector.

[0013] In some embodiments, the clustering unit determines the number of clustering categories based on the number of service categories that can be provided to users, selects the feature vectors of a corresponding number of users as initial cluster centers, and performs clustering processing based on the initial cluster centers and the feature vectors of each user.

[0014] In some embodiments, the clustering unit uses the particle swarm optimization algorithm to optimize the initial cluster centers, determines the optimized cluster centers, and performs clustering based on the optimized cluster centers and the feature vectors of each user.

[0015] In some embodiments, the apparatus further includes a determining unit, configured to determine the service category corresponding to each user category based on the characteristic data of each user in each user category, so as to provide service according to the corresponding service category of each user.

[0016] According to further embodiments of the present disclosure, a user classification apparatus is provided, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute the user classification method of any of the above embodiments based on instructions stored in the memory device.

[0017] According to further embodiments of the present disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the user classification method of any of the above embodiments.

[0018] In the above embodiments, relatively independent feature data are selected as the basis for user classification based on the degree of correlation between feature data. This reduces data redundancy and interference, allows for in-depth mining of user characteristics, and improves the accuracy of user classification. Attached Figure Description

[0019] The accompanying drawings, which form part of this specification, illustrate embodiments of this disclosure and, together with the specification, serve to explain the principles of this disclosure.

[0020] This disclosure will become clearer with reference to the accompanying drawings and the following detailed description:

[0021] Figure 1 Flowcharts illustrating some embodiments of the user classification method of this disclosure;

[0022] Figure 2 Show Figure 1 Flowcharts of some embodiments of step 120;

[0023] Figure 3 Flowcharts illustrating other embodiments of the user classification method of this disclosure;

[0024] Figure 4 Block diagrams illustrating some embodiments of the user classification device of this disclosure;

[0025] Figure 5 Block diagrams illustrating other embodiments of the user classification device of this disclosure;

[0026] Figure 6 Block diagrams illustrating further embodiments of the user classification apparatus of this disclosure are shown. Detailed Implementation

[0027] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present disclosure.

[0028] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0029] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this disclosure or its application or use.

[0030] Techniques, methods, and equipment known to a person skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the license specification.

[0031] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0032] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0033] As mentioned above, this disclosure addresses the aforementioned technical problems by classifying users based on their usage behavior data using optimized machine learning algorithms. This user classification allows for precise targeting of specific users. For example, network operators can use user classification to match users' service plans with their actual usage behavior. This enables accurate user service while reducing wasted time and resources. The technical solution of this disclosure can be implemented, for example, through the following embodiments.

[0034] Figure 1 Flowcharts illustrating some embodiments of the user classification method of this disclosure are shown.

[0035] like Figure 1 As shown, the method includes: step 110, calculating the correlation degree; step 120, generating feature vectors; and step 130, clustering processing.

[0036] In step 110, the correlation between various user characteristic data is calculated. For example, special diagnostic data is data that reflects user behavior, such as called call duration, number of called persons, called call duration, number of called persons, local called persons, average monthly data usage, data usage duration, and mobile terminal usage duration.

[0037] In some embodiments, the degree of correlation between feature data can be determined by calculating the covariance between feature data.

[0038] In step 120, a feature vector for the user is generated based on the feature data whose relevance is less than a threshold.

[0039] In some embodiments, it can be achieved through Figure 2 The embodiment in the example implements step 120.

[0040] Figure 2 Show Figure 1 A flowchart of some embodiments of step 120.

[0041] like Figure 2 As shown, step 120 includes: step 1210, generating an initial feature vector; and step 1220, performing dimensionality increase processing.

[0042] In step 1210, an initial feature vector for the user is generated based on the feature data whose relevance is less than a threshold.

[0043] In step 1220, the initial feature vector is increased in dimensionality using a Gaussian kernel function to generate the feature vector.

[0044] After generating the feature vector, it can be used Figure 1 The remaining steps continue to be categorized.

[0045] In step 130, the feature vectors of each user are clustered to divide each user into multiple user categories.

[0046] In some embodiments, the number of clustering categories is determined based on the number of service categories that can be offered to users. For example, if a network operator can offer users six different categories of service packages, the number of categories can be set to six.

[0047] In some embodiments, a number of feature vectors are randomly selected as initial cluster centers based on the number of categories.

[0048] In some embodiments, clustering is performed based on the initial cluster centers and the feature vectors of each user. For example, the particle swarm optimization algorithm can be used to optimize the initial cluster centers to determine optimized cluster centers; then, clustering is performed based on the optimized cluster centers and the feature vectors of each user.

[0049] In some embodiments, a service category corresponding to each user category is determined based on the characteristic data of each user in each user category, so as to provide service according to the corresponding service category of each user. For example, a network operator can determine the closest service package based on the characteristic data of each user in a user category, and then provide or recommend the service package of that category to users in that user category.

[0050] In some embodiments, it can be achieved through Figure 3 The embodiments described herein involve the selection and processing of user feature data.

[0051] Figure 3 Flowcharts illustrating some other embodiments of the user classification method of this disclosure are shown.

[0052] like Figure 3 As shown, the method may include: a data acquisition step; a data preprocessing step; and a feature field extraction step.

[0053] In the data acquisition step, feature data that reflects user behavior can be selected as the basis for user classification.

[0054] In some embodiments, within a user system, user data may include massive amounts of feature data such as user identity data, business and product data, behavioral data, and consumption data. Key feature data of user behavior can be selected from this data.

[0055] For example, the feature data may include multiple items such as the plan usage level, number usage duration, whether 4G is activated, terminal usage month, whether online payment is activated, online payment consumption amount, average monthly consumption amount, voice call duration, incoming call duration, number of incoming calls, outgoing call duration, number of outgoing calls, number of local outgoing calls, number of domestic outgoing calls, average monthly data usage, data usage duration, and mobile terminal usage duration.

[0056] In the data preprocessing step, feature data can be preprocessed by standardization and other methods.

[0057] In some embodiments, to avoid the influence of different units of measurement on user classification, all independent variable feature data can be standardized. This normalizes the feature data to a mean of 0, with values ​​ranging between 0 and 1; and the standardized data are then used to form a feature vector, thereby obtaining a first feature vector representing the behavior of each user (each user can correspond to one feature vector).

[0058] In the feature field extraction step, feature fields (feature data contained in the feature vector) can be extracted by determining the correlation of feature data to generate a second feature vector.

[0059] In some embodiments, the correlation between two feature data can be represented by the magnitude of the covariance. For example, the covariance formula is:

[0060]

[0061] X and Y are two feature fields, and n is the total number of data points corresponding to the feature fields. These are the average values ​​of data from multiple users corresponding to the two feature fields.

[0062] Based on the formula above, the correlation between the feature fields and the first feature vectors of multiple users can be calculated. The larger the covariance value, the greater the correlation between the two feature fields, and vice versa.

[0063] For example, the first feature vector of the collected user group includes three feature fields: feature field a, "average monthly consumption amount"; feature field b, "duration of using the telecom number"; and feature field c, "duration of voice calls".

[0064] Using the covariance formula described above, the correlation between each pair of the three feature fields (a and b, a and c, b and c) is calculated. The covariance values ​​obtained are 0.25, 0.045, and 0.57, respectively.

[0065] In this case, the correlation between "average monthly spending" and "duration of using the telecom number" is the least, so "average monthly spending" and "duration of using the telecom number" can be selected as the feature fields of the second feature vector.

[0066] In the above steps, the covariance formula is used to compare the feature fields in the first feature vector of the collected users pairwise to determine the feature field with the smallest covariance value; based on the covariance, multiple feature fields that can independently describe user behavior can be determined (the specific number is set as needed); these feature fields are combined into a second feature vector corresponding to each user in the user group.

[0067] In some embodiments, the feature fields of the second feature vector may include average monthly cost, data usage, duration of use of the telecommunications number, voice call duration, mobile terminal usage duration, online payment consumption amount, number of domestic outgoing calls, etc.

[0068] In some embodiments, the particle swarm optimization algorithm can be used to optimize the clustering path. For example, optimization can be achieved through the following steps.

[0069] Determine the number K of clusters (user categories) and the central value of each cluster. For example, the number K of clusters can be determined based on the number of service packages that the operator can offer. Furthermore, based on the service characteristics of various categories, the data characteristics of the cluster centers corresponding to different clusters can be determined, that is, the central value corresponding to each cluster can be determined.

[0070] For example, the operator can offer six service package tiers: 10 yuan, 19 yuan, 49 yuan, 69 yuan, 99 yuan, and 199 yuan. In this case, the number of clusters can be set to 6; based on the core customers of different package tiers (which can be randomly selected), the initial center value corresponding to the package tier cluster can be determined.

[0071] The particle swarm optimization (PSO) algorithm is used to optimize the center values ​​of each cluster, thereby obtaining an optimized clustering path. For example, the initial center values ​​of each cluster can be fed into the PSO algorithm for optimization, thus obtaining the optimal cluster center values. This improves the accuracy of subsequent clustering results.

[0072] The particle swarm optimization algorithm is characterized by its speed and accuracy, resulting in more accurate clustering and higher precision in user location. This optimization algorithm combines the local search capabilities of search algorithms with the global optimization capabilities of particle swarm optimization, enabling rapid target user location. This reduces runtime and effectively optimizes the entire model.

[0073] In some embodiments, a fuzzy C-means clustering method can be used, combined with optimized cluster center values, for clustering. For example, the clustering process may include the following steps.

[0074] A Gaussian kernel function is used to increase the dimensionality of the second eigenvector, introducing it into a higher-dimensional space to obtain the corresponding third eigenvector. For example, if the second eigenvector has at least three dimensions, the third eigenvector can be eight-dimensional after the dimensionality increase process.

[0075] After using the Gaussian kernel function for dimensionality enhancement, the differences between the data fields in the third feature vector increase and the mutual interference decreases, thus making the clustering results more accurate.

[0076] The fuzzy C-means algorithm, combined with optimized cluster center values, is used to cluster the third feature vectors. In this way, the clustering result of each third feature vector can be represented as the corresponding user classification result.

[0077] In some embodiments, for network operators, after classifying users, the user characteristic data in the classification results can be adjusted to services corresponding to the package, such as voice minutes, data traffic, etc.

[0078] For example, the cluster with K=1 in the clustering results includes users A, B, and C. The voice call durations for users A, B, and C are 100 minutes, 120 minutes, and 90 minutes respectively, which is close to the voice service provided by the operator's 10 yuan plan. In this case, the service plan for all users in the K=1 cluster can be upgraded to the 10 yuan plan. Thus, if the actual plan for users in the K=1 cluster is not the 10 yuan plan, this model can accurately recommend users to that plan, thereby achieving precise service.

[0079] In some embodiments, for scenarios where network operators provide communication services to users, firstly, user data spanning more than six months can be collected; the data of this user group is preprocessed to obtain a first feature vector corresponding to each user in the user group.

[0080] For example, the collected user data needs to meet the following criteria: the user data does not include user data from government or enterprise units, and only includes individual or family user data; user data with missing user numbers or a large number of missing independent characteristic data needs to be removed.

[0081] Secondly, the covariance formula can be used to calculate the correlation between the feature fields in the first feature vector of the collected multiple users; extract the feature fields with the least correlation with the user group; and use these feature fields to form the second feature vector corresponding to each user.

[0082] Then, based on the different levels of services that network operators can provide and their specific business content, the number of clusters K and the center value of each cluster are determined; the particle swarm optimization algorithm is used to optimize the center value of each cluster to obtain the optimized center value of each cluster.

[0083] Finally, the Gaussian kernel function is used to introduce each second feature vector corresponding to the user group into a high dimension, transforming it into a high-dimensional feature vector, which is the third feature vector corresponding to each user in the user group. Using the fuzzy C-means algorithm, combined with the optimized center value of each cluster, the third feature vector is clustered to obtain the classification result for each user.

[0084] For example, based on the classification results, a service package matched to a user can be compared with the package the user actually subscribed to. If the user's actual package is lower in quality than the service package matched in the classification results, a more precise service upgrade recommendation can be made to the user.

[0085] In the above embodiments, based on unique and irreplaceable feature fields corresponding to users, customers with similar behaviors are grouped into one category; and corresponding value-based service packages are recommended to each category of customers, making it more accurate in targeting customers and more suitable for precision services. This significantly reduces the number of orders dispatched by the system, lowers the cost of manual outbound calls, and allows for faster model updates without long waiting periods.

[0086] In the above embodiments, the target user is accurately located and the profile is accurate; due to the fast model operation speed, all data can be used in the model; abnormal behavior of the target user is immediately apparent, reducing the need for support from other systems and data reports; if the user is not satisfied, the first-tier package can be recommended first, and alternative packages are provided.

[0087] Figure 4 Block diagrams illustrating some embodiments of the user classification device of this disclosure are shown.

[0088] like Figure 4 As shown, the user classification device 4 includes a calculation unit 41, a generation unit 42, and a clustering unit 43.

[0089] The calculation unit 41 calculates the correlation between various user feature data. The generation unit 42 generates user feature vectors based on the feature data whose correlation is less than a threshold. The clustering unit 43 performs clustering processing on the feature vectors of each user, dividing each user into multiple user categories.

[0090] In some embodiments, the generation unit 42 generates an initial feature vector for the user based on feature data whose correlation is less than a threshold; and uses a Gaussian kernel function to perform dimensionality-upgrading on the initial feature vector to generate a new feature vector.

[0091] In some embodiments, clustering unit 43 determines the number of clustering categories based on the number of service categories that can be provided to users; selects the feature vectors of a corresponding number of users as initial cluster centers based on the number of categories; and performs clustering processing based on the initial cluster centers and the feature vectors of each user.

[0092] In some embodiments, clustering unit 43 uses particle swarm optimization to optimize the initial cluster centers and determine optimized cluster centers; then, based on the optimized cluster centers and the feature vectors of each user, it performs clustering processing.

[0093] In some embodiments, the user classification device 4 further includes a determination unit 44, which is used to determine the business category corresponding to each user category based on the feature data of each user in each user category, so as to provide business services according to the corresponding business category of each user.

[0094] In the above embodiments, relatively independent feature data are selected as the basis for user classification based on the degree of correlation between feature data. This reduces data redundancy and interference, allows for in-depth mining of user characteristics, and improves the accuracy of user classification.

[0095] Figure 5 Block diagrams illustrating other embodiments of the user classification device of this disclosure are shown.

[0096] like Figure 5 As shown, the user classification device 5 of this embodiment includes a memory 51 and a processor 52 coupled to the memory 51. The processor 52 is configured to execute the user classification method of any embodiment of this disclosure based on instructions stored in the memory 51.

[0097] The memory 51 may include, for example, system memory, fixed non-volatile storage media, etc. The system memory stores, for example, the operating system, application programs, bootloader, database, and other programs.

[0098] Figure 6 Block diagrams illustrating further embodiments of the user classification apparatus of this disclosure are shown.

[0099] like Figure 6 As shown, the user classification device 6 in this embodiment includes a memory 610 and a processor 620 coupled to the memory 610. The processor 620 is configured to execute the user classification method in any of the foregoing embodiments based on instructions stored in the memory 610.

[0100] The memory 610 may include, for example, system memory, fixed non-volatile storage media, etc. The system memory stores, for example, the operating system, application programs, bootloaders, and other programs.

[0101] The user's classification device 6 may also include an input / output interface 630, a network interface 640, and a storage interface 650. These interfaces 630, 640, and 650, as well as the memory 610 and processor 620, can be connected, for example, via a bus 660. The input / output interface 630 provides a connection interface for input / output devices such as a monitor, mouse, keyboard, and touchscreen. The network interface 640 provides a connection interface for various networked devices. The storage interface 650 provides a connection interface for external storage devices such as SD cards and USB flash drives.

[0102] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable non-transitory storage media containing computer-usable program code.

[0103] The user classification method, user classification device, and computer-readable storage medium according to this disclosure have been described in detail. To avoid obscuring the concept of this disclosure, some details known in the art have not been described. Those skilled in the art will fully understand how to implement the technical solutions disclosed herein based on the above description.

[0104] The methods and systems of this disclosure may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the methods is for illustrative purposes only, and the steps of the methods of this disclosure are not limited to the order specifically described above unless otherwise specifically stated. Furthermore, in some embodiments, this disclosure may also be implemented as a program recorded on a recording medium, the program including machine-readable instructions for implementing the methods according to this disclosure. Thus, this disclosure also covers recording media storing programs for performing the methods according to this disclosure.

[0105] While specific embodiments of this disclosure have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of this disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments without departing from the scope and spirit of this disclosure. The scope of this disclosure is defined by the appended claims.

Claims

1. A method for classifying users, comprising: Calculate the correlation between various user characteristic data; Generate the user's feature vector based on the feature data whose relevance is less than a threshold. Clustering is performed on the feature vectors of each user to classify each user into multiple user categories; The specific method of clustering is to determine the corresponding business category for each user category based on the feature data of each user in each user category, so as to provide business services according to the corresponding business category of each user. The step of clustering the feature vectors of each user and classifying each user into multiple user categories includes: The number of clustering categories is determined based on the number of service categories that can be provided to users, and the number of categories is determined according to the number of service package tiers provided by the operator. Based on the number of categories, a corresponding number of user feature vectors are selected as initial cluster centers. The center value of the initial cluster centers is determined based on the feature data of users who select different service packages. Clustering is performed based on the initial cluster centers and the feature vectors of each user.

2. The classification method of claim 1, wherein, The process of generating a user's feature vector based on feature data with a correlation level less than a threshold includes: Generate the user's initial feature vector based on the feature data whose relevance is less than a threshold; The initial feature vector is increased in dimensionality using a Gaussian kernel function to generate the feature vector.

3. The classification method of claim 1, wherein, The clustering process based on the initial cluster centers and the feature vectors of each user includes: The initial cluster centers are optimized using the particle swarm optimization algorithm to determine the optimal cluster centers; Clustering is performed based on the optimized cluster centers and the feature vectors of each user.

4. A user classification device, comprising: The calculation unit is used to calculate the correlation between various user feature data; The generation unit is used to generate the user's feature vector based on the feature data whose relevance is less than a threshold. Clustering unit, used to cluster the feature vectors of each user and divide each user into multiple user categories; The determining unit is used to determine the business category corresponding to each user category based on the feature data of each user in each user category, so as to provide business services according to the corresponding business category of each user. The clustering unit determines the number of clustering categories based on the number of service categories that can be provided to users. The number of categories is determined based on the number of service package tiers offered by the operator. Based on the number of categories, the feature vectors of a corresponding number of users are selected as initial cluster centers. Clustering is performed based on the initial cluster centers and the feature vectors of each user. The center value of the initial cluster centers is determined based on the feature data of users who have selected different service package tiers.

5. The sorting device according to claim 4, wherein, The generation unit generates an initial feature vector for the user based on feature data whose correlation is less than a threshold, and then uses a Gaussian kernel function to perform dimensionality increase processing on the initial feature vector to generate the feature vector.

6. The sorting device according to claim 4, wherein, The clustering unit uses the particle swarm optimization algorithm to optimize the initial cluster centers, determines optimized cluster centers, and performs clustering based on the optimized cluster centers and the feature vectors of each user.

7. A user classification device, comprising: Memory; and A processor coupled to the memory, the processor being configured to execute the user classification method of any one of claims 1-3 based on instructions stored in the memory.

8. A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the user classification method according to any one of claims 1-3.

Images