Telecommunication user recommendation method and device based on latent variables, electronic equipment and medium

Abstract

The invention discloses a telecom user recommendation method and device based on latent variables, electronic equipment and a medium, and the method comprises the steps: carrying out the group division of telecom user data, and obtaining the group data of a plurality of user groups; determining a latent variable vector according to the group data and a preset latent variable model; constructing a logic regression model based on latent variables; training the logistic regression model according to the group data and the latent variable vector to obtain a multi-service handling willingness prediction model corresponding to each user group; carrying out integrated learning on the multi-service handling willingness prediction model to obtain a telecommunication user service handling willingness prediction model; and according to the telecommunication user service handling willingness prediction model and the target user data, obtaining a handling willingness prediction result of each preset service, and further determining a target promotion service and a corresponding target recommendation user. The method can realize screening of users with high handling willingness for specific services, improves the accuracy of telecom user recommendation, and can be widely applied to the technical field of machine learning.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and in particular to a method, device, electronic device and medium for recommending telecom users based on latent variables. Background Art

[0002] With the wide application of Internet and smart home technologies, users' personalized needs for telecom services are increasing day by day. Although existing user profiling technologies can classify and recommend users, they do not fully capture users' latent preferences and subtle differences. Therefore, the current technology has certain limitations in identifying users' deep-seated needs and providing accurate recommendations, which affects the accuracy of telecom user recommendations.

[0003] Term Explanation:

[0004] Categorical data: refers to data with clear classification attributes, such as user type, handling type, etc.

[0005] Numerical data: refers to data that can be quantitatively measured, such as age, income, rate, contract points, etc.

[0006] Latent variable: a variable that reflects the deep-seated needs and preferences of user groups that cannot be captured by direct behavior data.

[0007] Deep learning model: refers to a method of feature extraction and prediction of data through a multi-layer neural network.

[0008] Ensemble Learning: is a machine learning method that improves the overall prediction performance by combining the prediction results of multiple models. It constructs a strong model by integrating multiple weak models (usually called base models), thereby improving accuracy, robustness and generalization ability. The basic idea of ensemble learning is "collective wisdom", that is, by combining the prediction results of multiple models, reducing the bias or variance of a single model, and improving the overall prediction effect. Summary of the Invention

[0009] The purpose of the present invention is to solve at least to some extent one of the technical problems existing in the prior art.

[0010] To this end, an object of an embodiment of the present invention is to provide a method for recommending telecom users based on latent variables, which can realize the screening of users with high willingness to handle specific services, support operators to quickly carry out business promotion activities, and improve the accuracy of telecom user recommendations.

[0011] Another object of an embodiment of the present invention is to provide a device for recommending telecom users based on latent variables.

[0012] In order to achieve the above technical objectives, the technical solutions adopted in the embodiments of the present invention include:

[0013] On the one hand, an embodiment of the present invention provides a method for recommending telecom users based on latent variables, including the following steps:

[0014] Obtain telecom user data, perform population division on the telecom user data, and obtain population data of multiple user groups;

[0015] Determine a latent variable vector according to the population data and a preset latent variable model;

[0016] Construct a logistic regression model based on latent variables, and the logistic regression model is used to predict the willingness of users to handle multiple preset services;

[0017] Train the logistic regression model according to the population data and the latent variable vector to obtain a multi-service handling willingness prediction model corresponding to each user group;

[0018] Perform ensemble learning on the multi-service handling willingness prediction models corresponding to each user group to obtain a telecom user service handling willingness prediction model;

[0019] Obtain target user data, obtain the handling willingness prediction results of each preset service according to the telecom user service handling willingness prediction model and the target user data, and then determine the target promotion service and the corresponding target recommended users according to the handling willingness prediction results.

[0020] Further, in an embodiment of the present invention, the performing population division on the telecom user data to obtain population data of multiple user groups specifically includes:

[0021] Determine the optimal number of clusters of the telecom user data by the elbow method or the silhouette coefficient method;

[0022] Cluster the telecom user data according to the optimal number of clusters to obtain multiple user groups, and divide the telecom user data into population data of each user group.

[0023] Further, in an embodiment of the present invention, the determining a latent variable vector according to the population data and a preset latent variable model specifically includes:

[0024] Statistically calculate the characteristic values of each user group according to the population data to obtain a user characteristic matrix and a characteristic mean vector;

[0025] Calculate a covariance matrix according to the user characteristic matrix and the characteristic mean vector;

[0026] Perform eigen decomposition on the covariance matrix to obtain an eigenvector matrix and a diagonal matrix;

[0027] Determine a load matrix according to the feature vector matrix and the diagonal matrix;

[0028] Determine a score factor matrix according to the load matrix, the user feature matrix, and the feature mean vector;

[0029] Substitute the feature mean vector, the load matrix, and the factor score matrix into a preset latent variable model to obtain the latent variable vector corresponding to the user group.

[0030] Further, in an embodiment of the present invention, the covariance matrix is calculated by the following formula:

[0031]

[0032] X centered = X - μ

[0033] where S represents the covariance matrix, n represents the number of user groups, X represents the user feature matrix, μ represents the feature mean vector, and X centered represents the centralized data matrix;

[0034] Perform eigenvalue decomposition on the covariance matrix by the following formula:

[0035] S = VDV T

[0036] where V represents the feature vector matrix and D represents the diagonal matrix;

[0037] Determine the load matrix by the following formula:

[0038]

[0039] where L represents the load matrix, and V m represents the feature vector matrix corresponding to the largest m eigenvalues in the group data, represents the diagonal matrix formed by the square roots of the largest m eigenvalues in the group data;

[0040] Determine the score factor matrix by the following formula:

[0041] F = (L T L) -1 L T (X - μ)

[0042] where F represents the score factor matrix;

[0043] The latent variable model is:

[0044] Z = μ + LF + ∈

[0045] Among them, Z represents the latent variable vector, and ∈ represents the preset error term.

[0046] Further, in an embodiment of the present invention, training the logistic regression model according to the group data and the latent variable vector specifically includes:

[0047] Initialize the service weight matrix and the bias matrix, and construct a multi-service joint loss function by combining the cross-entropy loss and the L2 regularization term;

[0048] Use the user feature matrix and the latent variable vector as inputs, and predict the user handling probabilities of each preset service through the logistic regression model;

[0049] Determine the loss value according to the user handling probability and the multi-service joint loss function;

[0050] Update the service weight matrix and the bias matrix according to the loss value, and evaluate the prediction performance of the logistic regression model through accuracy, recall rate or F1 score, and then adjust the learning rate and regularization strength of the logistic regression model.

[0051] Further, in an embodiment of the present invention, the logistic regression model is:

[0052]

[0053] Among them, Y = 1 indicates that the user chooses to handle the service, X represents the user feature matrix, z represents the latent variable vector, P represents the user handling probability, α represents the learning rate, β represents the bias matrix, and γ represents the influence factor matrix, which is used to capture the influence of the latent variable vector Z on the user's willingness to handle the service;

[0054] The multi-service joint loss function is:

[0055] L(W,β) = -[ylog(P)+(1 - y)log(1 - P)]+λ‖W‖ 2

[0056] Among them, L(W,β) represents the loss value, W represents the service weight matrix, y represents the sample label, and λ represents the regularization strength;

[0057] Update the service weight matrix and the bias matrix through the following formula:

[0058]

[0059] Where W new and W old respectively represent the service weight matrix after and before update, represents the gradient of the multi-service joint loss function with respect to the service weight matrix, and βnew and β old respectively represent the bias matrix after and before the update, indicating the gradient of the multi-service combined loss function with respect to the bias matrix.

[0060] Furthermore, in an embodiment of the present invention, the integrated learning of the multi-service handling willingness prediction models corresponding to each of the user groups is performed to obtain a telecommunications user service handling willingness prediction model, specifically as follows:

[0061] Taking the multi-service handling willingness prediction models corresponding to each of the user groups as sub-models, and using the gradient boosting decision tree as a non-linear fitting tool for integrated learning to obtain the telecommunications user service handling willingness prediction model;

[0062] wherein, the gradient boosting decision tree fits the residuals of the previous decision tree by gradually adding decision trees, thereby reducing the prediction error.

[0063] On the other hand, an embodiment of the present invention provides a telecommunications user recommendation device based on latent variables, including:

[0064] A group division module, configured to obtain telecommunications user data, perform group division on the telecommunications user data to obtain group data of multiple user groups;

[0065] A latent variable vector determination module, configured to determine a latent variable vector according to the group data and a preset latent variable model;

[0066] A model construction module, configured to construct a logistic regression model based on latent variables, and the logistic regression model is used to predict the handling willingness of users for multiple preset services;

[0067] A model training module, configured to train the logistic regression model according to the group data and the latent variable vector to obtain multi-service handling willingness prediction models corresponding to each of the user groups;

[0068] An integrated learning module, configured to perform integrated learning on the multi-service handling willingness prediction models corresponding to each of the user groups to obtain a telecommunications user service handling willingness prediction model;

[0069] A user recommendation module, configured to obtain target user data, obtain prediction results of the handling willingness of each of the preset services according to the telecommunications user service handling willingness prediction model and the target user data, and further determine a target promoted service and corresponding target recommended users according to the handling willingness prediction results.

[0070] On the other hand, an embodiment of the present invention provides an electronic device, which includes a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for implementing connection communication between the processor and the memory. When the program is executed by the processor, it implements the method for recommending telecommunications users based on latent variables as described above.

[0071] On the other hand, an embodiment of the present invention further provides a storage medium, which is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method for recommending telecommunications users based on latent variables as described above.

[0072] The advantages and beneficial effects of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention:

[0073] In the embodiment of the present invention, telecommunications user data is obtained, the telecommunications user data is grouped to obtain group data of multiple user groups, a latent variable vector is determined according to the group data and a preset latent variable model, a logistic regression model based on latent variables is constructed, and the logistic regression model is used to predict the willingness of users to handle multiple preset services. The logistic regression model is trained according to the group data and the latent variable vector to obtain a multi-service handling willingness prediction model corresponding to each user group. Ensemble learning is performed on the multi-service handling willingness prediction models corresponding to each user group to obtain a telecommunications user service handling willingness prediction model. Target user data is obtained, and the handling willingness prediction results of each preset service are obtained according to the telecommunications user service handling willingness prediction model and the target user data. Furthermore, the target promotion service and the corresponding target recommended users are determined according to the handling willingness prediction results. In the embodiment of the present invention, the extraction of the latent variable vector can enrich the user portrait, capture the potential needs and implicit behavior characteristics of users, use the logistic regression model based on latent variables to predict the willingness of users to handle each preset service, and through the training and ensemble learning of the logistic regression model, finally obtain a telecommunications user service handling willingness prediction model that integrates multiple services, which can realize the screening of users with high handling willingness for specific services, support the operator to quickly carry out business promotion activities, and improve the accuracy of telecommunications user recommendation. Description of the Drawings

[0074] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following introduces the drawings required to be used in the embodiments of the present invention. It should be understood that the drawings introduced below only facilitate the clear expression of some embodiments of the technical solutions in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative efforts.

[0075] Figure 1 It is a flowchart of steps for a method for recommending telecom users based on latent variables provided by an embodiment of the present invention;

[0076] Figure 2 It is a flowchart of steps for step S101 provided by an embodiment of the present invention;

[0077] Figure 3 It is a schematic diagram for determining the optimal number of clusters by the elbow method provided by an embodiment of the present invention;

[0078] Figure 4 It is a schematic diagram of the result of group division provided by an embodiment of the present invention;

[0079] Figure 5 It is a flowchart of steps for step S102 provided by an embodiment of the present invention;

[0080] Figure 6 It is a flowchart of steps for step S104 provided by an embodiment of the present invention;

[0081] Figure 7 It is a schematic flowchart of the process for training a logistic regression model provided by an embodiment of the present invention;

[0082] Figure 8 It is a flowchart of steps for step S105 provided by an embodiment of the present invention;

[0083] Figure 9 It is a schematic overall flowchart of the method for recommending telecom users based on latent variables provided by an embodiment of the present invention;

[0084] Figure 10 It is a schematic structural diagram of a device for recommending telecom users based on latent variables provided by an embodiment of the present invention;

[0085] Figure 11 It is a schematic hardware structure diagram of an electronic device provided by an embodiment of the present invention;

[0086] Figure 12 It is a schematic structural diagram of a storage medium provided by an embodiment of the present invention. Detailed implementation manners

[0087] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where like or similar reference numerals denote like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present application and should not be construed as a limitation to the present application. It should be noted that although functional module division is performed in the system schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different module division from that in the system schematic diagram or a different order from that in the flowchart. For the step numbers in the following embodiments, they are only set for the convenience of elaboration and explanation, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0088] In the description of the present invention, the meaning of "a plurality of" is two or more. If the first and second are described, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or the sequence of the indicated technical features. In addition, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0089] The method for recommending telecom users based on latent variables provided by the embodiments of the present application can be applied to a terminal, or to a server side, or can also be software running on a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a set-top box, etc.; the server side can be configured as an independent physical server, or can be configured as a server cluster or a distributed system composed of multiple physical servers, or can also be configured as a cloud server providing 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, CDN, and big data and artificial intelligence platforms; the software can be an application for implementing the method for recommending telecom users based on latent variables, etc., but is not limited to the above forms.

[0090] This application can be used in numerous general-purpose or special-purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This application can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0091] It should be noted that in each specific embodiment of this application, when it comes to relevant processing based on data related to the user's identity or characteristics, such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first. Moreover, the collection, use, and processing of these data will comply with the relevant laws, regulations, and standards of relevant countries and regions. In addition, when the embodiments of this application need to obtain sensitive personal information of the user, the user's separate permission or separate consent will be obtained through methods such as pop-up windows or redirecting to a confirmation page. After clearly obtaining the user's separate permission or separate consent, the necessary user-related data for the normal operation of the embodiments of this application will be obtained.

[0092] With the wide application of Internet and smart home technologies, users' personalized demands for telecommunications services are increasing day by day. Although existing user profiling technologies can classify and recommend users, they do not fully capture users' potential preferences and subtle differences. Therefore, the current technology has certain limitations in identifying users' deep-seated needs and providing accurate recommendations, which affects the accuracy of telecommunications user recommendations. To better solve this problem, the present invention customizes user profile latent variables (variables that reflect the deep-seated needs and preferences of user groups that cannot be captured by direct behavior data), and introduces latent variables into the recommendation system. Then, the service weights are dynamically adjusted according to the latent variables, enabling the algorithm to more autonomously adapt to the changing needs of user groups, thereby improving the accuracy, personalization level, and system response ability of recommendations.

[0093] In the prior art "Personalized Recommendation System and Method for Plastic Products Combining User Portraits", the user portrait module extracts multi-dimensional features of users, constructs user portraits and updates them dynamically, and matches them in the plastic product database through a matching module to generate a candidate set, and then uses the user portrait for personalized recommendation. However, the user portrait construction method of this patent mainly relies on direct user behavior data, such as search conditions and historical data, etc. This model based on explicit data may not fully capture more subtle differences in users' potential preferences, especially fails to consider users' potential and unmanifested needs and interests. The recommendation module relies more on existing data features. For some users without obvious behavior trajectories, or users with more periodic and non-linear behaviors, the updated recommended content may not fully reflect the changing interests and needs of users.

[0094] In the prior art "Method, Device, Electronic Device and Storage Medium for Generating Recommendation Model", the demand theme is identified by analyzing the text data of users, tags are constructed based on this, and the initial weights are adjusted according to the user behavior data to obtain the weight information corresponding to the target tags. However, this demand analysis method may rely on the text data input by users, ignoring potential needs and implicit behavior characteristics. For users who do not clearly express their needs, the system may not be able to comprehensively capture their needs. In addition, the dynamic adjustment of weights depends on the timely response of user behavior. However, in practice, users' needs and behaviors often have periodic and sporadic fluctuations. When this patent deals with complex and changeable user behaviors, the response may not be flexible enough.

[0095] Therefore, in the current business promotion work, a recommendation algorithm that fully analyzes the specificity and potential needs of user groups is needed. The present invention effectively captures users' potential preferences and subtle differences by introducing custom latent variables, and combines the latent variable dynamic optimization algorithm weights, so that the result can respond more quickly to changes in user needs and provide more personalized recommendations. The present invention can help enterprises quickly screen high-willingness customer groups for handling, efficiently promote specific businesses, and assist front-line personnel in precise marketing.

[0096] Such as Figure 1 shown is a step flow chart of a method for recommending telecom users based on latent variables provided by an embodiment of the present invention. Referring to Figure 1 , the embodiment of the present invention provides a method for recommending telecom users based on latent variables, which specifically includes the following steps:

[0097] S101. Obtain telecom user data, divide the telecom user data into groups, and obtain group data of multiple user groups;

[0098] S102. Determine a latent variable vector according to the group data and a preset latent variable model;

[0099] S103. Construct a logistic regression model based on latent variables, where the logistic regression model is used to predict the willingness of users to handle multiple preset services;

[0100] S104. Train the logistic regression model according to the group data and the latent variable vector to obtain a multi-service handling willingness prediction model corresponding to each user group;

[0101] S105. Perform ensemble learning on the multi-service handling willingness prediction models corresponding to each user group to obtain a telecom user service handling willingness prediction model;

[0102] S106. Obtain target user data, and according to the telecom user service handling willingness prediction model and the target user data, obtain the handling willingness prediction results of each preset service, and then determine the target promoted service and the corresponding target recommended users according to the handling willingness prediction results.

[0103] In the embodiment of the present invention, the extraction of the latent variable vector can enrich the user portrait, capture the potential needs and implicit behavior characteristics of users, use the logistic regression model based on latent variables to predict the willingness of users to handle each preset service, and through the training and ensemble learning of the logistic regression model, finally obtain a telecom user service handling willingness prediction model that integrates multiple services, which can realize the screening of users with high handling willingness for specific services, support the operator to quickly carry out business promotion activities, and improve the accuracy of telecom user recommendation.

[0104] As Figure 2 shown is a step flow chart of step S101 provided by the embodiment of the present invention. Referring to Figure 2 , further as an optional implementation manner, group the telecom user data to obtain group data of multiple user groups, which specifically includes:

[0105] S1011. Determine the optimal number of clusters of the telecom user data by the elbow method or the silhouette coefficient method;

[0106] S1012. Cluster the telecom user data according to the optimal number of clusters to obtain multiple user groups, and divide the telecom user data into group data of each user group.

[0107] Specifically, collect the existing telecom user data, including user type, monthly traffic, package type, region, consumption situation, marketing acceptance situation, whether to handle a certain service, etc., and use one-hot encoding or label encoding to numerically process the categorical data, and at the same time standardize the numerical data.

[0108] To accurately explore the differences among different user groups, it is necessary to first divide the user data into groups. The present invention uses the elbow method and the silhouette coefficient method to determine the optimal number of clusters, that is, the number of user groups.

[0109] The elbow method is a method for determining the optimal number of clusters in a clustering algorithm. This method calculates the sum of squared errors (SSE) under different numbers of clusters, observes the trend of the error decrease, and thus selects a suitable number of clusters. The sum of squared errors (SSE) represents the sum of the squares of the Euclidean distances from all sample points x i to its nearest cluster center (centroid) The formula is as follows:

[0110]

[0111] Taking K-means clustering and the elbow method as an example, Figure 3 shows the curve of calculating a certain data set and using the elbow method to calculate the number of K-means clusters. The number 3 corresponding to the elbow point on the curve is the optimal clustering k value, that is, the optimal number of clusters.

[0112] The silhouette coefficient method calculates a score based on the similarity within and between groups, is used to evaluate the quality of clustering, and measures the clustering effect of each data point. The silhouette coefficient formula is as follows:

[0113]

[0114] where s is the silhouette coefficient of a data point, and its value range is [-1, 1]. a is the average distance from data point i to other points in the same cluster, which is used to calculate the within-group compactness. b is the average distance from data point i to all users in the nearest classification group, which is called the between-group separation.

[0115] The silhouette coefficient can measure the quality of the clustering result. The closer the value is to 1, the better the sample points are clustered; the closer the value is to -1, the more likely the sample points are misclustered. By calculating the average silhouette coefficient under different k values, the k value corresponding to the maximum silhouette coefficient is selected as the optimal number of clusters.

[0116] After determining the optimal number of clusters, the probability that user n belongs to category k is calculated through multinomial logistic regression:

[0117]

[0118] In the formula, C k is category k, X n is the feature vector of user n, and β k is the coefficient vector related to category k. According to this formula, all users are divided into k groups. For exampleFigure 4 The following is a schematic diagram of the group division result provided by the embodiment of the present invention.

[0119] As shown in Figure 5 The following is a flowchart of a step of step S102 provided by the embodiment of the present invention. Referring to Figure 5 , as a further optional implementation manner, a latent variable vector is determined according to the group data and a preset latent variable model, which specifically includes:

[0120] S1021. Statistically calculate the characteristic values of each user group according to the group data to obtain a user characteristic matrix and a characteristic mean vector;

[0121] S1022. Calculate a covariance matrix according to the user characteristic matrix and the characteristic mean vector;

[0122] S1023. Perform eigenvalue decomposition on the covariance matrix to obtain an eigenvector matrix and a diagonal matrix;

[0123] S1024. Determine a load matrix according to the eigenvector matrix and the diagonal matrix;

[0124] S1025. Determine a score factor matrix according to the load matrix, the user characteristic matrix, and the characteristic mean vector;

[0125] S1026. Substitute the characteristic mean vector, the load matrix, and the factor score matrix into the preset latent variable model to obtain the latent variable vector of the corresponding user group.

[0126] As a further optional implementation manner, the covariance matrix is calculated by the following formula:

[0127]

[0128] X centered = X - μ

[0129] where S represents the covariance matrix, n represents the number of user groups, X represents the user characteristic matrix, μ represents the characteristic mean vector, and X centered represents the centralized data matrix;

[0130] The covariance matrix is eigen-decomposed by the following formula:

[0131] S = VDV T

[0132] where V represents the eigenvector matrix and D represents the diagonal matrix;

[0133] The load matrix is determined by the following formula:

[0134]

[0135] Among them, L represents the load matrix, and V m represents the eigenvector matrix corresponding to the largest m eigenvalues in the group data, and

[0136] represents the diagonal matrix formed by the square roots of the largest m eigenvalues in the group data;

[0136] The score factor matrix is determined by the following formula:

[0137] F = (L T L) -1 L T (X - μ)

[0138] where F represents the score factor matrix;

[0139] The latent variable model is:

[0140] Z = μ + LF + ∈

[0141] where Z represents the latent variable vector, and ∈ represents a preset error term.

[0142] Specifically, first, according to the user clustering result, the characteristic values of the group data of each user group are statistically calculated to obtain an n×p user characteristic matrix (assuming n user groups and p customer characteristics), denoted as X; the mean value of each characteristic value is calculated to obtain a p×1 mean vector μ. Next, a latent variable model is constructed. This model is used to extract latent variables from the observed data of users, and these variables can help explain the psychological and emotional factors behind user behavior. The formula of the latent variable model of the present invention is as follows:

[0143] Z = μ + LF + ∈

[0144] where Z is the latent variable vector, μ is the mean vector mentioned above, L is the load matrix, F is the latent factor score, and ∈ is the error term (the error value is set through analysis and estimation). The calculation of the parameters in the formula will be specifically introduced below.

[0145] Based on X and μ, the load matrix L is calculated. The load matrix L describes the relationship between the observed variables and the latent factors in factor analysis. Before calculating the load matrix, the covariance matrix needs to be calculated first and the principal components are extracted. The calculation process for constructing the load matrix L is as follows:

[0146] First, centralize the data matrix X:

[0147] X centered = X - μ

[0148] Next, calculate the covariance matrix of the centralized data:

[0149]

[0150] The covariance matrix S is eigen-decomposed by the following formula to extract the matrix V containing the eigenvectors and the diagonal matrix D:

[0151] S = VDV T

[0152] Finally, the number of factors m is determined according to the magnitudes of the eigenvalues and the percentage of the total variance to be explained, and the loading matrix is calculated:

[0153]

[0154] Here, V m is the matrix containing the eigenvectors corresponding to the m largest eigenvalues of the data set, while is the diagonal matrix formed by the square roots of these eigenvalues.

[0155] Next, the factor score matrix F is calculated through L, representing the scores of each observation (or individual) on each factor. Through factor score analysis, a complex data set containing many variables can be reduced to several representative factors, revealing the latent structure that is not easily observable directly in the data. These latent factors represent the common driving forces behind the observed variables (for example, the score factors may reveal that certain user groups tend to use a large amount of traffic but make few calls, and this information can help the operator design customized data plans or promotional activities for these groups). The factor score matrix F is calculated from the loading matrix calculated above:

[0156] F = (L T L) -1 L T (X - μ)

[0157] So far, the parameters of the latent variables have all been calculated. All the parameters are substituted into the latent variable model formula above to obtain the corresponding latent variable vector Z. The latent variables reflect the internal differences in the behavior of user groups and are important parameters for estimating the group decision coefficient.

[0158] As Figure 6 shown is a flowchart of a step of step S104 provided by an embodiment of the present invention. Referring to Figure 6 , further as an optional implementation manner, the logistic regression model is trained according to the group data and the latent variable vector, which specifically includes:

[0159] S1041. Initialize the service weight matrix and the bias matrix, and construct a multi-service joint loss function by combining the cross-entropy loss and the L2 regularization term;

[0160] S1042. Use the user feature matrix and the latent variable vector as inputs, and predict the user handling probabilities of each preset service through the logistic regression model;

[0161] S1043. Determine the loss value according to the user's handling probability and the multi-service joint loss function;

[0162] S1044. Update the service weight matrix and the bias matrix according to the loss value, and evaluate the prediction performance of the logistic regression model through accuracy, recall rate or F1 score, and then adjust the learning rate and regularization strength of the logistic regression model.

[0163] Further as an optional implementation manner, the logistic regression model is:

[0164]

[0165] where Y = 1 indicates that the user selects to handle the service, X represents the user feature matrix, Z represents the latent variable vector, P represents the user's handling probability, α represents the learning rate, β represents the bias matrix, and γ represents the influence factor matrix, which is used to capture the influence of the latent variable vector Z on the user's handling willingness;

[0166] The multi-service joint loss function is:

[0167] L(W,β) = -[ylog(P)+(1 - y)log(1 - P)]+λ‖W‖ 2

[0168] where L(W,β) represents the loss value, W represents the service weight matrix, y represents the sample label, and λ represents the regularization strength;

[0169] Update the service weight matrix and the bias matrix through the following formula:

[0170]

[0171] where W new and W old respectively represent the service weight matrix after and before the update, represents the gradient of the multi-service joint loss function with respect to the service weight matrix, β new and β old respectively represent the bias matrix after and before the update, represents the gradient of the multi-service joint loss function with respect to the bias matrix.

[0172] Specifically, use user data (user type, monthly traffic, package type, etc.) and the clustering result as the input of the customer data set, and use whether the user handles the service as the output of the target variable to construct a logistic regression model. As Figure 7 shown in the flow chart of training the logistic regression model provided by the embodiment of the present invention, the training process thereof will be described below.

[0173] First, initialize the business weight matrix W and the bias matrix β. The weights are based on prior business knowledge or randomly initialized, and the initial weight w of each business b (assuming the total number of businesses is B) is set. These weights represent the predicted importance of each business for decision-making at the beginning of the model, and the initialized business weight matrix is obtained: n , which represents the predicted importance of each business for decision-making at the beginning of the model, and the initialized business weight matrix is obtained:

[0174] W = [w1, w2, …, w B

[0175] After that, the latent variable Z and logistic regression are used to predict the user's willingness to handle the business. The logistic regression model designed in the present invention is as follows:

[0176]

[0177] where Y = 1 indicates that the user chooses to handle the business, X is the data matrix mentioned above, Z is the latent variable vector, α, β, and γ are model parameters, and γ captures the impact of the latent variable Z on the user's willingness to handle the business. The present invention uses maximum likelihood estimation or other optimization algorithms to fit the above logistic regression model and estimate the parameters α, β, and γ. This logistic regression model can predict the probability that a user will handle the business given the observed features and latent variables, transform the linear predictor into a probability through a logistic function, and output the percentage probability that the user will handle the business.

[0178] The present invention designs a loss function by combining cross-entropy loss (measuring the gap between the model prediction and the actual label) and L2 regularization term (preventing overfitting). The loss function is defined as follows:

[0179] L(W, β) = -[ylog(P) + (1 - y)log(1 - P)] + λ‖W‖ 2

[0180] During the training process, the loss function is optimized by calculating the gradients of the loss function L with respect to the weight W and the bias β, and the weights are updated according to the following rules (where α is the learning rate):

[0181]

[0182] Finally, the model performance is evaluated using accuracy, recall rate, F1 score, etc. Hyperparameters such as the learning rate α and the regularization strength λ are adjusted according to the performance of the model on the validation set. Thus, a multi-business handling prediction model based on a single group of latent variables for k (assuming all users are divided into k groups) is obtained.

[0183] As Figure 8 shown is a step flow chart of step S105 provided by an embodiment of the present invention. Referring to Figure 8 ​, further as an optional implementation, perform ensemble learning on the multi-service handling willingness prediction models corresponding to each user group to obtain a telecom user service handling willingness prediction model, specifically as follows:

[0184] S1051. Use the multi-service handling willingness prediction models corresponding to each user group as sub-models, and use gradient boosting decision trees as non-linear fitting tools for ensemble learning to obtain a telecom user service handling willingness prediction model;

[0185] Among them, the gradient boosting decision tree reduces the prediction error by gradually adding decision trees to fit the residuals of the previous decision tree.

[0186] Specifically, use the obtained multi-service handling prediction models of multiple single groups as sub-models, and use gradient boosting decision trees (GBDT) as non-linear fitting tools for ensemble learning to obtain the final telecom user service handling willingness prediction model that integrates user subjective factors. GBDT reduces the prediction error by gradually adding decision trees to fit the residuals of the previous tree. The formula is as follows:

[0187] F t (x) = F t-1 (x) + γ t h t (x)

[0188] Among them, F t-1 (x) is the cumulative prediction of all previous trees, h t (x) is the contribution of the current tree, and γ t is the optimization step size. In this way, the model becomes more accurate with each additional tree.

[0189] In the final fusion process, evaluate the model performance through cross-validation, and optimize the weight allocation and hyperparameter settings of GBDT to generate a more accurate final prediction result. The final model can be used for new customer data to predict their service handling willingness, or can directly screen high-willingness handling groups based on existing users.

[0190] In some optional embodiments, integrate the fused telecom user service handling willingness prediction model into the business system, and use the real-time collected target user data for prediction. The model dynamically generates the handling willingness prediction results of each preset service according to the input data, and determines the target promotion service and the corresponding target recommended users according to the handling willingness prediction results to support marketing decision-making and user behavior analysis. In addition, integrating the fused telecom user service handling willingness prediction model into the business system can realize the screening of high-willingness users for specific services, support the company to quickly carry out business promotion activities, and the system realizes the automated application of the algorithm, has the ability of real-time update and expansion, and can meet the multi-scenario business requirements.

[0191] As Figure 9 shown in the overall process schematic diagram of the latent variable-based telecom user recommendation method provided by the embodiment of the present invention, it can be recognized that the present invention combines user clustering and latent variable models to analyze the internal preference factors that cannot be directly observed by users, can comprehensively describe the multi-dimensional needs of users, provide more accurate personalized recommendations, enhance the algorithm's understanding of the heterogeneity of user personalized needs and preferences. In addition, it can more comprehensively capture the multi-dimensional characteristics of users, and can also flexibly adapt to changes in user needs, further improving the intelligence and real-time performance of the recommendation system; in the process of initial weight setting and adjustment, the present invention introduces a custom latent variable, making the initial weight adjustment mechanism more precise. In the process of weight adjustment, it combines the potential behavior and needs of users, and more flexibly adapts to the complex changes of user behavior. Through the latent variable, the present invention can effectively identify and adjust the changes in user needs at different time periods or in different situations, thus ensuring the real-time performance and accuracy of the recommendation.

[0192] As Figure 10 shown in the structural schematic diagram of the latent variable-based telecom user recommendation device provided by the embodiment of the present invention, referring to Figure 10 , the embodiment of the present invention provides a latent variable-based telecom user recommendation device, including:

[0193] A group division module, configured to obtain telecom user data, divide the telecom user data into groups, and obtain group data of multiple user groups;

[0194] A latent variable vector determination module, configured to determine a latent variable vector according to the group data and a preset latent variable model;

[0195] A model construction module, configured to construct a latent variable-based logistic regression model, and the logistic regression model is used to predict the willingness of users to handle multiple preset services;

[0196] A model training module, configured to train the logistic regression model according to the group data and the latent variable vector to obtain a multi-service handling willingness prediction model corresponding to each user group;

[0197] An ensemble learning module, configured to perform ensemble learning on the multi-service handling willingness prediction models corresponding to each user group to obtain a telecom user service handling willingness prediction model;

[0198] A user recommendation module, configured to obtain target user data, obtain a handling willingness prediction result of each preset service according to the telecom user service handling willingness prediction model and the target user data, and then determine a target promotion service and corresponding target recommended users according to the handling willingness prediction result.

[0199] The content in the above method embodiments is applicable to the device embodiments of the present invention. The functions specifically implemented by the device embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0200] An embodiment of the present invention further provides an electronic device, which includes: a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing the connection and communication between the processor and the memory. When the program is executed by the processor, it realizes the above-mentioned latent variable-based telecommunications user recommendation method. The electronic device can be any intelligent terminal including a tablet computer, an in-vehicle computer, etc.

[0201] As Figure 11 shown is a schematic diagram of the hardware structure of the electronic device provided by the embodiment of the present invention. Referring to Figure 11 , an embodiment of the present invention provides an electronic device, including:

[0202] A processor 1101, which can be implemented by using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided by the embodiments of the present invention;

[0203] A memory 1102, which can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), etc. The memory 1102 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1102, and the processor 1101 is called to execute the latent variable-based telecommunications user recommendation method of the embodiments of the present invention;

[0204] An input / output interface 1103, which is used to realize information input and output;

[0205] A communication interface 1104, which is used to realize the communication interaction between this device and other devices, and can realize communication through a wired method (such as USB, network cable, etc.) or through a wireless method (such as a mobile network, WIFI, Bluetooth, etc.);

[0206] A bus 1105, which transmits information between the various components of the device (such as the processor 1101, the memory 1102, the input / output interface 1103, and the communication interface 1104);

[0207] Among them, the processor 1101, the memory 1102, the input / output interface 1103, and the communication interface 1104 are communicatively connected to each other inside the device through the bus 1105.

[0208] As Figure 12 shown is a schematic structural diagram of the storage medium provided by an embodiment of the present invention. Referring to Figure 12 , an embodiment of the present invention further provides a storage medium. The storage medium is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs 1201, and the one or more programs 1201 can be executed by one or more processors to implement the above-mentioned latent variable-based telecommunication user recommendation method.

[0209] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include memories remotely disposed relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above-mentioned network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0210] An embodiment of the present invention also discloses a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 1 the method shown.

[0211] In some alternative embodiments, the functions / operations mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the functions / operations involved, two consecutive blocks shown can actually be executed substantially simultaneously, or the above-mentioned blocks can sometimes be executed in the reverse order. In addition, the embodiments presented and described in the flowcharts of the present invention are provided by way of example for the purpose of providing a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated, where the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0212] In addition, although the present invention has been described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the above functions and / or features may be integrated in a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It should also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. Rather, considering the attributes, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the modules will be understood within the ordinary skills of an engineer. Therefore, those skilled in the art can implement the present invention as set forth in the claims without undue experimentation. It should also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0213] If the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the above methods in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0214] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a predefined sequence of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0215] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection (electronic device) having one or more wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable media can even be paper or other suitable media on which the above programs can be printed, because the above programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then storing them in a computer memory.

[0216] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0217] In the above description of this specification, descriptions with reference to the terms "one embodiment / example", "another embodiment / example", or "certain embodiments / examples", etc., mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0218] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and purposes of the present invention, and the scope of the present invention is defined by the claims and their equivalents.

[0219] The above has specifically described the preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for recommending telecommunications users based on latent variables, characterized in that, It includes the following steps: Obtain telecommunication user data, perform population division on the telecommunication user data to obtain population data of multiple user groups; Determine a latent variable vector according to the population data and a preset latent variable model; Construct a logistic regression model based on the latent variable, and the logistic regression model is used to predict the willingness of users to handle multiple preset services; Train the logistic regression model according to the population data and the latent variable vector to obtain a multi-service handling willingness prediction model corresponding to each user group; Perform ensemble learning on the multi-service handling willingness prediction models corresponding to each user group to obtain a telecommunication user service handling willingness prediction model; Obtain target user data, and obtain a handling willingness prediction result of each preset service according to the telecommunication user service handling willingness prediction model and the target user data, and then determine a target promotion service and corresponding target recommended users according to the handling willingness prediction result.

2. The method for recommending telecommunications users based on latent variables according to claim 1, wherein The performing population division on the telecommunication user data to obtain population data of multiple user groups specifically includes: Determine the optimal number of clusters of the telecommunication user data by the elbow method or the silhouette coefficient method; Cluster the telecommunication user data according to the optimal number of clusters to obtain multiple user groups, and divide the telecommunication user data into population data of each user group.

3. A method for recommending telecommunications users based on latent variables according to claim 1, characterized in that, The determining a latent variable vector according to the population data and a preset latent variable model specifically includes: Statistically calculate the characteristic values of each user group according to the population data to obtain a user characteristic matrix and a characteristic mean vector; Calculate a covariance matrix according to the user characteristic matrix and the characteristic mean vector; Perform eigen decomposition on the covariance matrix to obtain an eigenvector matrix and a diagonal matrix; Determine a load matrix according to the eigenvector matrix and the diagonal matrix; Determine a score factor matrix according to the load matrix, the user characteristic matrix, and the characteristic mean vector; Substitute the characteristic mean vector, the load matrix, and the factor score matrix into the preset latent variable model to obtain the latent variable vector corresponding to the user group.

4. A method for recommending telecom users based on latent variables according to claim 3, characterized in that, The covariance matrix is calculated by the following formula: X centered = X - μ Among them, S represents the covariance matrix, n represents the number of user groups, X represents the user feature matrix, μ represents the feature mean vector, and X centered represents the centralized data matrix; The eigen decomposition of the covariance matrix is performed by the following formula: S = VDV T Where, V represents the eigenvector matrix, and D represents the diagonal matrix; The load matrix is determined by the following formula: Among them, L represents the load matrix, and V m represents the eigenvector matrix corresponding to the largest m eigenvalues in the population data, represents the diagonal matrix formed by the square roots of the largest m eigenvalues in the population data; The score factor matrix is determined by the following formula: F = (L T L) -1 L T (X - μ) Where, F represents the score factor matrix; The latent variable model is: Z = μ + LF + ∈ Where, Z represents the latent variable vector, and ∈ represents a preset error term.

5. The method for recommending telecommunications users based on latent variables according to claim 3, characterized in that The training the logistic regression model according to the population data and the latent variable vector specifically includes: Initialize a service weight matrix and a bias matrix, and construct a multi-service joint loss function by combining cross-entropy loss and an L2 regularization term; Use the user characteristic matrix and the latent variable vector as inputs, and predict the user handling probability of each preset service through the logistic regression model; Determine a loss value according to the user handling probability and the multi-service joint loss function; Update the service weight matrix and the bias matrix according to the loss value, and evaluate the prediction performance of the logistic regression model through accuracy, recall rate, or F1 score, and then adjust the learning rate and regularization strength of the logistic regression model.

6. The method for recommending a telecommunications user based on latent variables according to claim 5, wherein The logistic regression model is: where Y = 1 indicates that the user chooses to handle the service, X represents the user feature matrix, Z represents the latent variable vector, P represents the user's handling probability, α represents the learning rate, β represents the bias matrix, and γ represents the influence factor matrix, which is used to capture the influence of the latent variable vector Z on the user's willingness to handle the service; The multi-service joint loss function is: L(W,β) = -[ylog(P) + (1 - y)log(1 - P)] + λ‖W‖ 2 where L(W,β) represents the loss value, W represents the service weight matrix, y represents the sample label, and λ represents the regularization strength; Update the service weight matrix and the bias matrix through the following formula: Among them, W new and W old respectively represent the service weight matrices after and before the update, represents the gradient of the multi-service joint loss function with respect to the service weight matrix, and β new and β old respectively represent the bias matrices after and before the update, represents the gradient of the multi-service joint loss function with respect to the bias matrix.

7. A method for recommending telecommunications users based on latent variables according to any one of claims 1 to 6, characterized in that Perform ensemble learning on the multi-service handling willingness prediction models corresponding to each of the user groups to obtain a telecom user service handling willingness prediction model, specifically: Use the multi-service handling willingness prediction models corresponding to each of the user groups as sub-models, and use gradient boosting decision trees as non-linear fitting tools for ensemble learning to obtain the telecom user service handling willingness prediction model; where the gradient boosting decision tree reduces the prediction error by gradually adding decision trees to fit the residuals of the previous decision tree.

8. A telecommunications user recommendation device based on latent variables, characterized in that It includes: A group division module, which is used to obtain telecom user data, divide the telecom user data into groups, and obtain the group data of multiple user groups; A latent variable vector determination module, which is used to determine the latent variable vector according to the group data and a preset latent variable model; A model construction module, which is used to construct a logistic regression model based on latent variables, and the logistic regression model is used to predict the user's willingness to handle multiple preset services; A model training module, which is used to train the logistic regression model according to the group data and the latent variable vector to obtain a multi-service handling willingness prediction model corresponding to each of the user groups; An ensemble learning module, which is used to perform ensemble learning on the multi-service handling willingness prediction models corresponding to each of the user groups to obtain a telecom user service handling willingness prediction model; A user recommendation module, which is used to obtain target user data, obtain the prediction results of the handling willingness of each of the preset services according to the telecom user service handling willingness prediction model and the target user data, and then determine the target promoted service and the corresponding target recommended users according to the handling willingness prediction results.

9. An electronic device, characterized in that, The electronic device includes a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing the connection and communication between the processor and the memory. When the program is executed by the processor, the steps of the latent variable-based telecom user recommendation method according to any one of claims 1 to 7 are realized.

10. A storage medium, which is a computer-readable storage medium for computer-readable storage, characterized in that, The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to realize the steps of the latent variable-based telecom user recommendation method according to any one of claims 1 to 7.

Images