Multi-dimensional evaluation model training method and device, recommendation method and device, equipment and medium

Abstract

The invention provides a multi-dimensional evaluation model training method and device, a recommendation method and device, equipment and a medium, and the method comprises the steps: obtaining standard user data of a plurality of users, including consumption behavior data, infrastructure data and social attribute data; each piece of standard user data is mapped to a three-dimensional space, and three-dimensional feature vectors of consumption behaviors, infrastructure use conditions and social attributes of corresponding users are obtained; performing multi-order clustering iterative training on the three-dimensional feature vectors of all the users to obtain user groups of different clusters; and aggregating the consumption behaviors of the user groups of the different clusters, the use conditions of the infrastructures and the feature vectors of the social attributes to obtain a multi-dimensional evaluation model of the user group portraits of the different clusters. According to the method, data of consumption behaviors, infrastructures and social attributes of the users are mapped to the three-dimensional space for comprehensive user behavior analysis and multi-order clustering, so that the user portraits of user groups of different clusters are obtained, and the accuracy of the user portraits is improved.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a multidimensional evaluation model training method, a user service recommendation method, a device, an electronic device, and a computer-readable storage medium. Background Art

[0002] In related technologies, with the development of network technology, there are many shortcomings in network business expansion. For example, the large amount of data accumulated in internal daily operations is not fully utilized, there is a lack of in-depth analysis of users' consumption habits, preferences and potential needs, or it is based on some simple static rule configurations, lacks flexibility and personalization, and is difficult to meet the specific needs of different users. The user conversion rate is low, resulting in low accuracy of user portraits, inability to achieve effective business expansion, and reduced business expansion efficiency.

[0003] Therefore, how to effectively improve the efficiency of business expansion is a technical problem that needs to be solved. Summary of the Invention

[0004] The present invention provides a multidimensional evaluation model training method, a user service recommendation method, an apparatus, an electronic device, and a computer-readable storage medium to at least address the problem in related technologies of being unable to effectively expand services and reducing business expansion efficiency due to a lack of flexibility and personalized analysis of users. The technical solutions of the present invention are as follows:

[0005] According to a first aspect of an embodiment of the present invention, a method for training a multidimensional evaluation model is provided, comprising:

[0006] Acquire standard user data of multiple users, wherein the standard user data includes: consumption behavior data, infrastructure data, and social attribute data;

[0007] Map each standard user data into three-dimensional space to obtain the corresponding three-dimensional feature vector of the user's consumption behavior, infrastructure usage and social attributes;

[0008] Perform iterative multi-stage clustering training on the three-dimensional feature vectors of all users to obtain user groups in different clusters;

[0009] The consumption behaviors, infrastructure usage and social attribute feature vectors of the user groups of different clusters are aggregated to obtain a multi-dimensional evaluation model of the three-dimensional user portraits of the user groups of different clusters.

[0010] Optionally, the acquisition of standard user data of multiple users includes: consumption behavior data, infrastructure data, and social attribute data, including:

[0011] Acquire user data of multiple enterprise users, wherein the user data includes: consumption behavior data, infrastructure data, and social attribute data;

[0012] Each user data is preprocessed to obtain preprocessed standard user data.

[0013] Optionally, the iterative training of multi-stage clustering is performed on the three-dimensional feature vectors of all users to obtain different user groups, including:

[0014] Perform iterative first-order clustering on the three-dimensional feature vectors of all users to obtain user groups in different clusters;

[0015] The boundary data of the user groups in different clusters after division are iteratively refined by second-order clustering to obtain refined user groups.

[0016] Optionally, performing first-order clustering iteration on the three-dimensional feature vectors of all users to obtain user groups in different clusters after division includes:

[0017] Based on the three-dimensional feature vectors of all users, iterative calculation is performed according to minimizing the square error to determine multiple cluster centers. Each cluster center is the cluster that minimizes the distance between each user's three-dimensional feature vector and the corresponding cluster center.

[0018] Based on the determined multiple cluster centers, cluster division is performed on the three-dimensional feature vectors of all users to obtain user groups in different clusters after division.

[0019] Optionally, performing a second-order clustering iterative optimization process on the boundary data of each divided user group to obtain the optimized user group includes:

[0020] For the user groups divided into different clusters, calculate the posterior probability that each user's three-dimensional feature vector belongs to the corresponding cluster;

[0021] Based on the posterior probability, the parameter mean, covariance matrix and prior probability in the posterior probability are iteratively updated by maximizing the log-likelihood function; until the parameter mean, covariance matrix and prior probability converge, an optimized user group is obtained.

[0022] Optionally, the method further includes:

[0023] Acquire a data set, the data set including sample data of multiple users, the sample data including: consumption behavior sample data, infrastructure sample data, and social attribute sample data;

[0024] The multidimensional evaluation model is cross-validated and A / B tested using the data set to obtain a tested multidimensional evaluation model.

[0025] According to a second aspect of an embodiment of the present invention, a user service recommendation method is provided, comprising:

[0026] Obtain consumption behavior data, infrastructure data, and social attribute data of the users to be tested;

[0027] Inputting the infrastructure data, the social attribute data, and the consumption behavior data into a trained multidimensional evaluation model for prediction to obtain the user group to which the user to be tested belongs, wherein the multidimensional evaluation model is based on mapping the obtained standard user data of multiple users into a three-dimensional space and performing iterative multi-order clustering training on the obtained three-dimensional feature vectors to obtain user groups in different clusters, and aggregating the feature vectors of the user groups in different clusters to obtain a model;

[0028] Based on the three-dimensional user portrait of the user group to which the user to be tested belongs, generating a personalized service package corresponding to the user group for the user to be tested;

[0029] The personalized service package is recommended to the user to be tested.

[0030] Optionally, obtaining consumption behavior data, infrastructure data, and social attribute data of the user to be tested includes:

[0031] Obtain infrastructure data and social attribute data of the user to be tested, perform regression analysis on the infrastructure data and the social attribute data, and predict potential consumption behavior data of the user to be tested; or

[0032] Directly obtain the consumption behavior data, infrastructure data and social attribute data of the user to be tested.

[0033] Optionally, performing regression analysis on the acquired infrastructure data and social attribute data to predict the user's potential consumption behavior data includes:

[0034] Performing principal component analysis on the infrastructure data and the social attribute data to obtain simplified infrastructure data and social attribute data;

[0035] The simplified infrastructure data and the simplified social attribute data are input into a regression model for prediction to obtain the potential consumption behavior data of the user to be tested.

[0036] According to a third aspect of an embodiment of the present invention, a multidimensional evaluation model training device is provided, comprising:

[0037] An acquisition module, configured to acquire standard user data of multiple users, wherein the standard user data includes consumption behavior data, infrastructure data, and social attribute data;

[0038] A mapping module, configured to map each standard user data into a three-dimensional space to obtain a three-dimensional feature vector of the corresponding user's consumption behavior, infrastructure usage, and social attributes;

[0039] The multi-stage clustering module performs iterative multi-stage clustering training on the three-dimensional feature vectors of all users to obtain user groups in different clusters;

[0040] The aggregation module is used to aggregate the consumption behaviors, infrastructure usage and social attribute feature vectors of the user groups of different clusters to obtain a multi-dimensional evaluation model of the user group portraits of different clusters.

[0041] Optionally, the acquisition module includes:

[0042] A multi-dimensional data acquisition module, configured to acquire user data of multiple enterprise users, wherein the user data includes consumption behavior data, infrastructure data, and social attribute data;

[0043] The preprocessing module is used to preprocess each user data to obtain preprocessed standard user data.

[0044] Optionally, the multi-stage clustering module includes:

[0045] The first-order clustering module is used to iteratively divide the three-dimensional feature vectors of all users into first-order clusters to obtain user groups in different clusters;

[0046] The second-order clustering module is used to perform iterative optimization processing of second-order clustering on the boundary data of the user groups of different clusters divided by the first-order clustering module to obtain the optimized user groups.

[0047] Optionally, the first-order clustering module includes:

[0048] A cluster center determination module is used to determine multiple cluster centers by iterative calculation based on the three-dimensional feature vectors of all users in accordance with minimization of squared error, where each cluster center is the cluster that minimizes the distance between each user's three-dimensional feature vector and the corresponding cluster center;

[0049] The cluster division module is used to perform cluster division on the three-dimensional feature vectors of all users based on the determined multiple cluster centers to obtain user groups in different clusters after division.

[0050] Optionally, the second-order clustering module includes:

[0051] A calculation module is used to calculate the posterior probability that each user's three-dimensional feature vector belongs to the corresponding cluster after the user groups are divided into different clusters;

[0052] An iterative update module is used to iteratively update the model parameter mean, covariance matrix and prior probability in the posterior probability based on the posterior probability and using the maximized log-likelihood function until the parameter mean, covariance matrix and prior probability converge to obtain an optimized user group.

[0053] Optionally, the device further includes:

[0054] A data acquisition module is used to acquire a verification data set, wherein the verification data set includes: sample data of multiple users, wherein the sample data includes: consumption behavior sample data, infrastructure sample data, and social attribute sample data;

[0055] The validation test module is used to perform cross-validation and A / B testing on the multidimensional evaluation model using the validation data set to obtain a tested multidimensional evaluation model.

[0056] According to a fourth aspect of an embodiment of the present invention, there is provided a user service recommendation device, comprising:

[0057] The acquisition module is used to obtain the consumption behavior data, infrastructure data and social attribute data of the user to be tested;

[0058] a prediction module, configured to input the infrastructure data, the social attribute data, and the consumer behavior data into a trained multidimensional evaluation model for prediction, thereby obtaining the user group to which the user to be tested belongs, wherein the multidimensional evaluation model is based on mapping the acquired standard user data of multiple users into a three-dimensional space and performing iterative multi-stage clustering training on the obtained three-dimensional feature vectors to obtain user groups in different clusters, and a model obtained by aggregating the feature vectors of the user groups in different clusters;

[0059] A generating module, configured to generate a personalized service package for the user to be tested, based on a three-dimensional user portrait of the user group to which the user to be tested belongs;

[0060] The recommendation module is used to recommend the personalized service package to the user to be tested.

[0061] Optionally, the acquisition module includes: a first acquisition module and a regression analysis module; and / or a second acquisition module, wherein:

[0062] The first acquisition module is used to acquire infrastructure data and social attribute data of the user to be tested;

[0063] The regression analysis module is used to perform regression analysis on the infrastructure data and the social attribute data to predict the potential consumption behavior data of the user to be tested;

[0064] The second acquisition module is used to directly acquire the consumption behavior data, infrastructure data and social attribute data of the user to be tested.

[0065] Optionally, the regression analysis module includes:

[0066] A principal component processing module, configured to perform principal component analysis on the infrastructure data and the social attribute data to obtain simplified infrastructure data and social attribute data;

[0067] The consumption behavior prediction module is used to input the simplified infrastructure data and the social attribute data into a regression model for prediction, so as to obtain the potential consumption behavior data of the user to be tested.

[0068] According to a fifth aspect of an embodiment of the present invention, there is provided an electronic device, including:

[0069] processor;

[0070] a memory for storing instructions executable by the processor;

[0071] The processor is configured to execute the instructions to implement the multi-dimensional evaluation model training method or the user service recommendation method as described above.

[0072] According to the sixth aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the multidimensional evaluation model training method as described above or the user service recommendation method as described above.

[0073] According to the seventh aspect of an embodiment of the present invention, a computer program product is provided, comprising a computer program or instructions, which, when executed by a processor of an electronic device, implements the multidimensional evaluation model training method described above or the user service recommendation method described above.

[0074] The technical solutions provided by the embodiments of the present invention bring at least the following beneficial effects:

[0075] In an embodiment of the present invention, standard user data of multiple users is obtained, and the standard user data includes: consumption behavior data, infrastructure data, and social attribute data; each standard user data is mapped to a three-dimensional space to obtain a three-dimensional feature vector of the corresponding user's consumption behavior, infrastructure usage, and social attributes; a multi-order clustering analysis is performed on the three-dimensional feature vectors of all users to obtain user groups of different clusters; the feature vectors of the consumption behavior, infrastructure usage, and social attributes of the user groups of different clusters are aggregated to generate a multi-dimensional evaluation model of the user group portraits of different clusters. That is, in an embodiment of the present invention, a comprehensive user behavior analysis is performed by mapping the user's consumption behavior, infrastructure usage, and social attribute data to different axes of a three-dimensional space, and the user data is subjected to multi-order clustering to obtain user portraits of user groups of different clusters, thereby improving the accuracy of the user portraits, and effectively expanding the business based on the user portraits to improve the efficiency of business expansion.

[0076] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present invention, and together with the specification are used to explain the principles of the present invention, and do not constitute an undue limitation of the present invention. In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

[0078] Figure 1 This is a flowchart of a multidimensional evaluation model training method provided by an embodiment of the present invention.

[0079] Figure 2 It is a schematic diagram of a first-order clustering result provided by an embodiment of the present invention.

[0080] Figure 3 It is a schematic diagram of a second-order clustering result provided by an embodiment of the present invention.

[0081] Figure 4 This is a schematic diagram of displaying results in three-dimensional space provided by an embodiment of the present invention.

[0082] Figure 5 This is a schematic diagram of an A / B test result provided by an embodiment of the present invention.

[0083] Figure 6This is a flowchart of a user service recommendation method provided by an embodiment of the present invention.

[0084] Figure 7 This is a user prediction regression diagram provided by an embodiment of the present invention.

[0085] Figure 8 This is a block diagram of a multidimensional evaluation model training device provided by an embodiment of the present invention.

[0086] Figure 9 This is a block diagram of a user service recommendation device provided by an embodiment of the present invention.

[0087] Figure 10 This is a block diagram of an electronic device provided by an embodiment of the present invention.

[0088] Figure 11 This is a block diagram of an apparatus for multi-dimensional evaluation model training or user service recommendation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0089] In order to enable ordinary persons in the art to better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0090] It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present invention, as detailed in the appended claims.

[0091] Technical terms:

[0092] Clustering is an unsupervised learning method that groups objects in a dataset into clusters based on similarity. Objects within the same cluster are highly similar, while objects in different clusters are highly diverse. Clustering algorithms are widely used in fields such as data mining, image segmentation, and social network analysis.

[0093] K-fold cross-validation: Divide the dataset into K subsets, using K-1 of them as the training set and the remaining one as the validation set. Repeat this process K times, using each subset as the validation set once. Finally, the performance of multiple validation sets is combined to evaluate the model's effectiveness.

[0094] A / B testing: A method for verifying the effectiveness of a model directly in a business environment. By randomly assigning users to different groups, the differences between new solutions and traditional methods are tested to evaluate the actual effect of the model.

[0095] Silhouette coefficient: measures the closeness and separation of clusters.

[0096] Regression model: A statistical analysis method used to explore and quantify the relationship between variables. It constructs a mathematical formula to predict the impact of one or more independent variables (features) on a dependent variable (target variable). Common regression models include linear regression and nonlinear regression. Linear regression assumes a linear relationship between variables, while nonlinear regression allows for more complex relationships.

[0097] Principal Component Analysis (PCA): A dimensionality reduction technique used to simplify a dataset while preserving as much of the important information as possible. It projects the original high-dimensional data onto a new coordinate system and identifies the directions of maximum variance (called principal components). These principal components are linear combinations of the original data features and are orthogonal (uncorrelated). The main purpose of PCA is to reduce data dimensionality, remove noise and redundancy, and improve the efficiency of analysis and modeling while still preserving the majority of the data's information.

[0098] On the basis of understanding the above technical terms, please refer to the following embodiments.

[0099] See also Figure 1 , is a flow chart of a multi-dimensional evaluation model training method provided by an embodiment of the present invention, such as Figure 1 As shown, the method includes the following steps:

[0100] Step 101: Acquire standard user data of multiple users, where the standard user data includes consumption behavior data, infrastructure data, and social attribute data.

[0101] Step 102: Map each standard user data into a three-dimensional space to obtain a three-dimensional feature vector of the corresponding user's consumption behavior, infrastructure usage, and social attributes.

[0102] Step 103: Perform iterative multi-stage clustering training on the three-dimensional feature vectors of all users to obtain user groups in different clusters.

[0103] Step 104: Aggregate the characteristic vectors of consumption behaviors, infrastructure usage, and social attributes of the user groups of different clusters to obtain a multi-dimensional evaluation model of the three-dimensional user portraits of the user groups of different clusters.

[0104] The multidimensional evaluation model training method described in the present invention can be applied to terminals, servers, etc., without limitation. The terminal implementation device can be an electronic device such as a smart phone, a laptop computer, a tablet computer, a desktop computer, a personal digital assistant (PDA) and a wearable device. The server can be an independent server or a server cluster, or a server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, intermediary services, domain name services, security services, content distribution networks, or a big data and artificial intelligence platform, etc., without limitation.

[0105] The following combination Figure 1 , the specific implementation steps of a multidimensional evaluation model training method provided by an embodiment of the present invention are described in detail.

[0106] In step 101, standard user data of a plurality of users is obtained, where the standard user data includes consumption behavior data, infrastructure data, and social attribute data.

[0107] In this step, user data of multiple corporate users (i.e., corporate customers) can be obtained, and the user data includes: consumption behavior data, infrastructure data, and social attribute data; and each user data is cleaned and standardized to obtain corresponding standard user data. Of course, the obtained user data can also be stored, etc., which is not limited in this implementation.

[0108] That is, in this step, by sorting out the massive amount of user-related data of various types accumulated in the current big data pool, we can determine the data used in this embodiment from a business perspective. The user data used in this embodiment is divided into three parts: consumption behavior data, infrastructure data, and social attribute data, as described below:

[0109] First, collect the consumption behavior data, infrastructure data, and social attribute data, including:

[0110] 1) Consumer behavior data

[0111] This consumer behavior data can be obtained through internal systems, such as telecommunications companies' internal systems, to capture consumer behavior data for government and enterprise customers. This consumer behavior data describes the communication behaviors and usage patterns of government and enterprise users. Using the data aggregated into the big data pool, typical consumer behavior data that matches user characteristics is extracted, as shown in Table 1, including categories and content. It should be noted that Table 1 is merely an example and is not limited to this in actual applications.

[0112] Table 1

[0113] category content Call data Average call duration, number of calls, and SMS volume for users of government and enterprise customers. Traffic data Monthly traffic usage, peak traffic, nighttime traffic, etc. of government and enterprise customers. Consumption data Monthly bills, call charges, activation of value-added services, etc. Terminal Usage Device model, operating system, and app usage duration.

[0114] 2) Infrastructure data

[0115] In this embodiment, infrastructure data is extracted based on base stations and queries various devices and network monitoring system logs within the base station coverage area, such as log data of base stations, routers, switches, and other equipment, device traffic data, grids, coordinates, and building surface HE area data synchronized by the GIS system. Typical infrastructure data extracted is shown in Table 2, which is only an example and is not limited to this in actual applications:

[0116] Table 2

[0117]

[0118]

[0119] 3) Social attribute data

[0120] Social attribute data collection is mainly obtained through external public information or commercial data sources, covering external social attributes such as the size, industry, and financial status of the enterprise. Social attribute data may include: the number of employees, number of branches, years of establishment, profit, income, investment amount, industry classification and ranking, and bidding project information, etc. Data collection methods may include: obtaining registration and operating data through government-disclosed industrial and commercial information, purchasing financial and industry data from third-party data service providers (such as Tianyancha, Qichacha, etc.), or obtaining the industry ranking and market performance of the enterprise through industry reports and business intelligence companies. Typical social attribute data are extracted, as shown in Table 3. It should be noted that Table 3 is only an example and is not limited to this in actual application:

[0121] Table 3

[0122] category content Enterprise scale Number of employees, number of branches, years of establishment Financial Data Corporate profits, revenue, and investment. Industry Information The industry classification and ranking of the enterprise. Bidding Information Bidding projects in which enterprises participate.

[0123] Secondly, the collected consumer behavior data, infrastructure data, and social attribute data are preprocessed. Preprocessing can include storage, cleaning, and standardization to ensure data quality and consistency, facilitating subsequent modeling and analysis.

[0124] 1) Data Storage

[0125] (1) Storage method

[0126] Business data is stored in a relational database (PostgreSQL), including structured data such as user behavior data and base station traffic.

[0127] The synchronization of large amounts of logs and unstructured data (such as network device logs, application logs, etc.) is processed using a big data platform (Spark).

[0128] Due to the large scale of data, the distributed storage service provided by China Telecom's cloud service platform (Tianyi Cloud) was used for storage and backup.

[0129] (2) Data table design

[0130] Consumption behavior data table: saves each user's behavior records, such as traffic, call duration, terminal device type, etc. The primary key is the enterprise user ID.

[0131] 2) Data cleaning

[0132] The purpose of data cleaning is to delete incomplete, erroneous, and duplicate records to ensure data consistency and accuracy. Specifically, it includes:

[0133] (1) Missing value processing:

[0134] Deleting records: If the proportion of missing data is small, delete these records directly.

[0135] Filling missing values: For some important fields, fill missing values according to business rules. For example, if the average call duration of a user is missing, fill it with the average value of the user's group.

[0136] (2) Duplicate value processing:

[0137] Duplicate check: Use fields such as user ID and device ID to check whether there are duplicate records in the data.

[0138] Merge duplicates: If there are multiple records with the same ID, merge them based on the latest record to avoid errors caused by data duplication.

[0139] (3) Outlier processing:

[0140] Detect extreme values: Detect extreme values that clearly do not conform to business logic. If a user's traffic usage is abnormally higher than the normal range, it will be considered as erroneous data collection and will be cleared.

[0141] Processing strategy: handle these outliers by deletion, correction, etc., and use rules to filter out abnormal records during modeling.

[0142] Infrastructure data table: stores information such as traffic flow and device access volume of each base station, and associates the base station ID and regional information.

[0143] Social attribute data table: stores enterprise information, including financial data, industry classification, etc. The primary key is the enterprise ID.

[0144] 3) Data standardization

[0145] Its purpose is to convert data of different magnitudes to the same scale through normalization, categorical data encoding and other techniques to avoid misleading the model due to different eigenvalue ranges. Specifically, it includes:

[0146] (1) Data normalization

[0147] Scale consumption data, traffic data, and other data in consumer behavior data to the range of [0,1]. The formula used is:

[0148]

[0149] (2) Data standardization

[0150] Convert data such as base station traffic fluctuations and user behavior patterns into a standard normal distribution with a mean of 0 and a standard deviation of 1. The formula used is:

[0151]

[0152] (3) Data category coding

[0153] Variables such as terminal device type, package type, and terminal device brand are converted into binary vectors using unique encoding.

[0154] In step 102, each standard user data is mapped to a three-dimensional space to obtain a three-dimensional feature vector corresponding to the user's consumption behavior, infrastructure usage, and social attributes.

[0155] In this step, each standard user data is mapped to three axes of a three-dimensional space, which correspond to the key feature vectors of the three dimensions of the user's consumption behavior, infrastructure usage, and social attributes.

[0156] (1) X-axis: Class A data (consumption behavior data, which is discrete data)

[0157] This type of data describes the average consumption behavior characteristics of individual enterprise users, including consumption, communication behavior, terminal device usage, etc. In this embodiment, a vector X can be used to represent the values of enterprise users on these characteristics.

[0158] (2) Y-axis: Category b data (infrastructure data, such as telecommunications infrastructure data)

[0159] This type of data is related to telecommunications infrastructure and reflects data in the telecommunications infrastructure network, such as base station traffic usage, device fluctuations, user distribution, etc. In this implementation, vector Y can be used to represent the performance of the enterprise user in these dimensions.

[0160] (3) Z-axis: C-type data (external social attribute data)

[0161] This type of data describes the external social attributes of the enterprise, such as enterprise size, industry, financial status, etc. In this embodiment, the social attribute characteristics of the enterprise can be represented by a vector Z.

[0162] Thus, in this embodiment, for each enterprise user, a three-dimensional vector can be used to describe its characteristic direction in three dimensions. Specifically, it can be expressed by the following formula:

[0163]

[0164] In this formula, Di represents the three-dimensional feature vector of user i. For each user i, there is a dimension n X +n Y +n Z The eigenvector of .

[0165] In step 103, iterative training of multi-stage cluster analysis is performed on the three-dimensional feature vectors of all users to obtain user groups in different clusters.

[0166] In this step, first, the three-dimensional feature vectors of all users are iteratively divided into first-order clusters to obtain user groups in different clusters after division; specifically, it includes: based on the three-dimensional feature vectors of all users, iterative calculation is performed according to minimizing the square error to determine multiple cluster centers, each cluster center is a cluster that minimizes the distance between the three-dimensional feature vector of each user and the corresponding cluster center; based on the determined multiple cluster centers, the three-dimensional feature vectors of all users are clustered to obtain user groups in different clusters after division.

[0167] Secondly, the boundary data of the user groups in different clusters after division are subjected to second-order clustering iterative optimization processing to obtain the optimized user groups, specifically including: calculating the posterior probability that the three-dimensional feature vector of each user belongs to the corresponding cluster for the user groups in different clusters after division; based on the posterior probability, using the maximized log-likelihood function, iteratively updating the parameter mean, covariance matrix and prior probability in the posterior probability; repeating the above iterative process until the parameter mean, covariance matrix and prior probability converge to obtain the optimized user group.

[0168] In other words, in order to meet the real-time requirements of large amounts of data (such as massive telecommunications data), this embodiment needs to be able to quickly cluster large numbers of users while also being able to handle fuzzy and complex data distributions. Therefore, this embodiment proposes two parts: preliminary first-order clustering and optimized (refined) second-order clustering. That is, first-order clustering is performed on the user's three-dimensional feature vector to find multiple preliminary cluster centers C. k , according to the cluster division, we can get the user groups of different clusters; for each cluster of user groups, we can further refine the model and estimate the parameters of the Gaussian distribution (center of the Kth cluster) μ kand covariance matrix ∑ k , through continuous iteration, adjust the cluster division and Gaussian distribution parameters until the Gaussian distribution parameters converge. The specific clustering process is as follows:

[0169] (1) First-order clustering, i.e. initial clustering

[0170] First, perform preliminary first-order clustering on the three-dimensional feature vectors of user data in the three-dimensional feature space. In this embodiment, P = (X, Y, Z) can be used as the feature vector of each user, and the following steps are performed:

[0171] Clustering goal:

[0172] Find K cluster centers μ1,μ2,...,μ k , so that each user i belongs to the cluster u with the smallest distance k , the optimization goal is to minimize the square error:

[0173]

[0174] Where Di represents the feature vector of user i; r ij is a binary variable. If data point i is assigned to cluster j, then r ij =1, otherwise r ij =0. That is:

[0175]

[0176] Iterate according to the above steps and update the center of each cluster:

[0177]

[0178] Based on this, this embodiment will find the optimal cluster division in three-dimensional space. j When convergence occurs, stop the iteration.

[0179] (2) Second-order clustering, i.e., refined clustering

[0180] After the initial first-order clustering, further processing is required for user groups with fuzzy boundaries, especially users that cross clusters. In this process, second-order clustering refines clusters through two steps: the E-step (expectation) and the M-step (maximization). The model parameters are estimated by maximizing the log-likelihood function of the data.

[0181] In step E, the feature vector D = (X, Y, Z) of each cluster is considered to be a user, which can be expressed as a mixture model composed of multiple Gaussian distributions. i The posterior probability γ that it belongs to the jth cluster ij , the posterior probability γ ijIt can also be called responsibility, which can be expressed by the explicit formula:

[0182]

[0183] Where: π j represents the prior probability of the jth cluster, N(D i |μ j ,∑ j ) is the probability density function of the j-th Gaussian distribution, μ j represents the mean of the jth cluster, ∑ j represents the covariance matrix of the jth cluster, π k represents the prior probability of the kth cluster, N(D i |μ k ,∑ k ) is the probability density function of the Gaussian distribution of the kth cluster, μ k represents the mean of the kth cluster, ∑ k represents the covariance matrix of the kth cluster. It should be noted that the jth cluster is any cluster center in the kth cluster.

[0184] In this embodiment, a three-dimensional Gaussian distribution is defined for each cluster, and the relationship between users in these three dimensions is captured from the covariance matrix. The covariance matrix is in the form of:

[0185]

[0186] Step M, based on the responsibility γi calculated in step E j , maximize the log-likelihood function and update the model parameter mean μ j , covariance matrix ∑ j , prior probability π j , specifically including:

[0187] 1) Update the mean μ j , the mean is calculated by weighted average, and the weight is the responsibility degree γ of the current user data j ij , update as follows:

[0188]

[0189] 2) Update the covariance matrix ∑ j In this embodiment, the variance of the current user data in each dimension and the covariance between the dimensions are updated. Specifically, the update is performed using the following formula:

[0190]

[0191] Among them, (D i -μ j )(D i-μ j ) T It is the outer product of the distance between user i and cluster center j, indicating the variance contribution of the data.

[0192] 3) Prior probability π j In this embodiment, the update mainly represents the ratio of the number of data points in the cluster to the total number of population points. The update method is as follows:

[0193]

[0194] in, is the effective number of all data points in cluster j.

[0195] In this embodiment, by continuously iterating the E step and the M step until all parameters converge, that is, the parameter change is less than the set threshold, at this time, the iteration is stopped, and a detailed probability distribution of each cluster (that is, a detailed probability model) can be obtained. Based on the detailed probability distribution, more detailed user classification can be performed according to the detailed probability distribution, that is, users are assigned to user groups in different clusters.

[0196] In step 104, the characteristic vectors of consumption behaviors, infrastructure usage, and social attributes of the user groups of different clusters are aggregated to generate a multi-dimensional evaluation model of the user group portraits of different clusters.

[0197] In this step, the characteristic vectors of consumption behavior, infrastructure usage and social attributes of user groups in different clusters are aggregated to generate a multi-dimensional user portrait evaluation model for different user groups, which can also be called a three-dimensional user portrait evaluation system model (i.e., 3D-UPES model).

[0198] Specifically, in this embodiment of the present invention, the acquired data for each standard user is mapped onto the three axes of a three-dimensional space and the feature vectors along these three axes are subjected to multi-stage clustering to construct a multi-dimensional evaluation model (also referred to as a fitting evaluation model, etc.) tailored to the business characteristics. In this embodiment, this model is also referred to as a Three-Dimensional User Portrait Evaluation System (3D-UPES). Subsequently, the three-dimensional feature vectors of all users are input into the multi-dimensional evaluation model for iterative multi-stage clustering training. The probability distributions obtained from this iterative training are used to determine the user groups in different clusters.

[0199] For example, in this embodiment, the above three types of data can be mapped to the three axes of the three-dimensional space respectively based on the data source in the telecommunications field, and the characteristic vectors on the three axes can be multi-stage clustered to construct a multi-dimensional evaluation model suitable for the characteristics of telecommunications business, and then iterative training of multi-stage clustering can be performed.

[0200] Among them, this paper uses telecommunications user data as an example to generate (construct) a multi-dimensional evaluation model. The core ideas include:

[0201] (1) Telecom user data is mapped to the three axes of three-dimensional space according to the data source category. The three axes correspond to the key feature vectors of the three dimensions respectively.

[0202] (2) For the key feature vectors of the three dimensions, the efficient clustering state of the data (i.e., first-order clustering and second-order clustering) is calculated, and probability density distribution analysis is performed. A new clustering evaluation method is proposed, that is, first-order clustering and second-order clustering are performed on the three-dimensional feature vectors of all users in turn, and users are divided into user groups of different clusters according to the probability distribution obtained by clustering.

[0203] (3) Aggregate the behavioral, network, and social attribute characteristics of different users in different clusters (or aggregate them into a unified portrait model) to form a continuous and dynamic evaluation mechanism, that is, to obtain a multidimensional evaluation model.

[0204] Among them, generating a multidimensional evaluation model is to learn the distribution probability of the input samples and generate new sample data through this distribution probability.

[0205] In an embodiment of the present invention, standard user data of multiple users is obtained, and the standard user data includes: consumption behavior data, infrastructure data, and social attribute data; each standard user data is mapped to a three-dimensional space to obtain a three-dimensional feature vector of the corresponding user's consumption behavior, infrastructure usage, and social attributes; a multi-order clustering analysis is performed on the three-dimensional feature vectors of all users to obtain user groups of different clusters; the feature vectors of the consumption behavior, infrastructure usage, and social attributes of the user groups of different clusters are aggregated to generate a multi-dimensional evaluation model of the user group portraits of different clusters. That is, in an embodiment of the present invention, a comprehensive user behavior analysis is performed by mapping the user's consumption behavior, infrastructure usage, and social attribute data to different axes of a three-dimensional space, and the user data is subjected to multi-order clustering to obtain user portraits of user groups of different clusters, thereby improving the accuracy of the user portraits, and effectively expanding the business based on the user portraits to improve the efficiency of business expansion.

[0206] Optionally, in another embodiment, based on the above embodiment, the method may further include:

[0207] Acquire a verification data set, the verification data set including sample data of multiple users, the sample data including: consumption behavior data, infrastructure data, and social attribute data;

[0208] The multidimensional evaluation model is cross-validated and A / B tested using a validation dataset to obtain a tested multidimensional evaluation model.

[0209] That is, in this embodiment of the present invention, the performance and effectiveness of the trained multidimensional evaluation model (i.e., the 3D-UPES model) need to be further verified. This embodiment simulates the training effect of user profiles of different groups and compares the differences between the experimental group and the control group to ensure the performance of the multidimensional evaluation model in a real environment.

[0210] First, model validation

[0211] This embodiment proposes to use cross-validation and A / B testing to verify the multi-dimensional evaluation model, specifically including:

[0212] 1) Cross-validation

[0213] In this embodiment, the performance of the 3D-UPES model can be verified by K-fold cross validation to evaluate the indicators such as the silhouette score, accuracy, and log likelihood of the 3D-UPES model.

[0214] 2) A / B testing

[0215] This embodiment tests different service strategies or network resource optimization strategies in the experimental group and the control group respectively. By comparing the results of the experimental group and the control group, the improvement effect of the 3D-UPES model in user classification, service return rate, customer satisfaction, resource optimization, etc. is evaluated.

[0216] Experimental group: Use user portraits built using the 3D-UPES model to deliver precise service push or network optimization.

[0217] Control group: Use traditional user segmentation services (such as simple grouping strategies based on spending or user activity, etc.), and then use the same service strategies or network optimization.

[0218] Secondly, the model experimental design

[0219] 1) Experimental dataset:

[0220] This embodiment randomly obtains 1,000 telecommunication enterprise users from the database, and obtains the consumption behavior data, infrastructure data, and social attribute data (i.e., three types of data features) of these 1,000 telecommunication enterprise users through the above-mentioned acquisition method.

[0221] The dimensions of each type of data are as follows:

[0222] X-axis (consumption behavior data): consumption amount, data usage, call duration, etc., a total of 10 features.

[0223] Y-axis (infrastructure data): base station coverage area, equipment usage, etc., a total of 8 features.

[0224] Z-axis (social attribute data): enterprise size, industry, financial status, etc., a total of 5 characteristics.

[0225] 2) Experimental environment

[0226] The tools used in this embodiment include

[0227] sklearn: used for clustering, data normalization, cross-validation, etc.

[0228] matplotlib: for data visualization.

[0229] seaborn: used to generate charts.

[0230] 3) Evaluation Metrics

[0231] Silhouette Score: measures the closeness and separation of clustering results.

[0232] Log Likelihood: Used for model evaluation, it indicates how well the data point fits the probability model.

[0233] Accuracy: Comparison of the marketing effectiveness or resource optimization effects of the application in the experimental group and the control group.

[0234] 4) Verification steps

[0235] Data preprocessing: Data normalization is done so that all features are at the same level. Please refer to the above for details and will not be repeated here.

[0236] First-order clustering: The data set is divided into K clusters, and the clustering effect is recorded.

[0237] Second-order clustering: further clustering is performed within each cluster to optimize the clustering results.

[0238] Cross-validation: Each time, use the data of 800 users for training and the data of the remaining 200 users for validation. Repeat this process 5 times, using a different set of 200 users for validation each time.

[0239] A / B testing: Apply the constructed user profiles to the experimental and control groups to observe the differences in their performance in service strategies.

[0240] 5) Verify the results

[0241] Using the above verification process, the schematic diagram of the first-order clustering results is as follows Figure 2 As shown, Figure 2 A schematic diagram of a first-order clustering result (i.e., preliminary clustering result) provided by an embodiment of the present invention. Figure 2 It can be seen that the horizontal axis is the simulated behavior data and the vertical axis is the simulated network data. The silhouette coefficient result is 0.55, the clustering effect is good, and the differences between user groups are large.

[0242] The schematic diagram of the second-order clustering results is as follows Figure 3 As shown, Figure 3 A schematic diagram of a second-order clustering result (i.e., refined clustering result) provided by an embodiment of the present invention is shown in FIG. Figure 3 It can be seen that the horizontal axis is the simulated behavior data and the vertical axis is the simulated network data (i.e., infrastructure data). The log likelihood is -1100, and the degree of fitting of the user data distribution is high, which can refine the user groups with fuzzy boundaries.

[0243] The result diagram is shown in three-dimensional space. Figure 4 As shown, Figure 4 A schematic diagram of displaying results in a three-dimensional space provided by an embodiment of the present invention, that is, mapping three types of user data into the three-dimensional space.

[0244] The schematic diagram of the A / B test results is as follows Figure 5 As shown, Figure 5 A schematic diagram of an A / B test result (experimental control result of the experimental group VS the control group) provided in an embodiment of the present invention, Figure 5 The results show that the horizontal axis represents the service success rate (i.e., marketing success rate) and the vertical axis represents the number of users. The service success rate (i.e., marketing success rate) of the experimental group increased by approximately 35% compared to the control group, demonstrating the advantages of user profiling based on 3D-UPES in providing precise services.

[0245] See also Figure 6 , is a flowchart of a user service recommendation method provided by an embodiment of the present invention, the method comprising:

[0246] Step 601: Obtain consumption behavior data, infrastructure data, and social attribute data of the user to be tested.

[0247] In this embodiment, there are two ways to obtain data. One is that the user to be tested is a telecom user. If the user is a telecom user, the user's consumption behavior data can be obtained from the telecom content system, and the telecom user's infrastructure data and social attribute data can be obtained from other systems, servers, or databases. The specific acquisition process is described above and will not be repeated here.

[0248] Alternatively, if the user to be tested is a non-telecom user, the consumption behavior data of the user cannot be obtained from the telecom internal system. However, the infrastructure data and social attribute data of the user to be tested (i.e., the non-telecom user) can be obtained from other systems, servers, or databases. The specific acquisition process is detailed above and will not be repeated here.

[0249] Based on the infrastructure data and social attribute data of the user to be tested (i.e., non-telecom user), regression analysis is performed on the infrastructure data and social attribute data to predict the user's potential consumption behavior data. Specifically, it includes:

[0250] First, principal component analysis is performed on the infrastructure data and social attribute data to obtain simplified infrastructure data and social attribute data. Second, the simplified infrastructure data and social attribute data are input into a regression model for prediction to obtain the potential consumption behavior data of the user to be tested. The specific process is described below.

[0251] In this embodiment, for telecom users, since they already have the above three types of data, the three types of data of the telecom users (i.e., consumption behavior data, infrastructure data, and social attribute data) can be directly collected and input into a trained multi-dimensional evaluation model (such as a three-dimensional user portrait evaluation system 3D-UPES model, etc.) to generate clustered users for the corresponding user groups.

[0252] For non-telecom users, only infrastructure data and social attribute data are collected, but consumption behavior data is missing. Therefore, this embodiment needs to infer the potential consumption behavior data of non-telecom users to form three-dimensional data, that is, to find potential users and match them with appropriate service recommendations (such as marketing recommendations). To this end, this example uses a prediction method based on a regression model to infer consumption behavior data. The process includes:

[0253] (1) Mostly feature processing

[0254] Since each dimension contains multiple feature variables, directly using all features may lead to overfitting and the computational cost is too high. Therefore, this embodiment uses the principal component analysis (PCA) method to help simplify the feature space while retaining most of the information. Among them, PCA reduces the dimensionality of the original high-dimensional data through linear transformation and finds some new "principal components". These principal components can explain most of the variation in the original data. After PCA processing, the known two-dimensional data (i.e., infrastructure data and social attribute data) will be simplified into several principal components respectively. Its implementation steps include:

[0255] 1) Standardized data: Since the dimensions of each feature are different, we first standardize each feature so that they have the same mean and variance.

[0256] 2) Calculate the covariance matrix: Construct the covariance matrix of the data to represent the relationship between each feature.

[0257] 3) Eigenvalue Decomposition: Eigenvalue decomposition is used to find the eigenvalues and eigenvectors of the covariance matrix. Each eigenvector corresponds to a principal component, and the eigenvalue indicates the degree to which the principal component explains the variance of the data.

[0258] 4) Principal component selection: The first few principal components that explain most of the variance are selected to replace the original features, thereby reducing the dimensionality.

[0259] (2) Multiple regression model

[0260] In this step, after obtaining the two-dimensional features simplified in step (1), prediction is performed using a multiple linear regression model. That is, the multiple regression model processes multiple input features and finds the best fit by minimizing the error.

[0261] The regression model is in the form of:

[0262] X i =β0+β1Y pca1 +β2Y pca2 +β3Y pca3 +β4Z pca1 +β5Z pca2 +β6Z pca3 +ε i

[0263] Among them, Y pca1 ,Y pca2 ,Y pca3 is the principal component of the infrastructure data after PCA simplification, Z pca1 ,Z pca2 ,Z pca3 is the principal component of social attribute data after PCA simplification, X i It refers to multiple characteristics of consumer behavior data, such as call duration, data usage, etc.

[0264] (3) Solving the regression coefficient

[0265] In this step, the least squares method is used to minimize the sum of squared errors and find the optimal regression coefficient of the regression model. That is, by minimizing the objective function, the optimal β parameter can be found and then prediction can be made. The prediction results are as follows: Figure 7 As shown, Figure 7 A user prediction regression diagram provided by an embodiment of the present invention, that is, using a regression model to predict VS actual. Figure 7 As shown, the horizontal axis is the actual call duration (minutes), and the vertical axis is the predicted call duration (minutes). Based on the certain data (i.e. potential consumption behavior data), the consumption behavior data, infrastructure data and social attribute data of the non-telecom user can be input into the multidimensional evaluation model for prediction, so as to obtain the corresponding telecom user group to which the non-telecom user belongs, that is, services classified according to telecom users can be recommended to the non-telecom user.

[0266] Step 602: Input the infrastructure data, social attribute data and consumption behavior data into a trained multidimensional evaluation model for prediction to obtain a three-dimensional user portrait of the corresponding user group to which the user to be tested belongs, wherein the multidimensional evaluation model is based on mapping the standard user data of multiple users obtained into a three-dimensional space, and performing iterative training of multi-order clustering on the obtained three-dimensional feature vectors to obtain user groups in different clusters, as well as a model obtained by aggregating the feature vectors of user groups in different clusters.

[0267] In this step, the multidimensional evaluation model is pre-trained. The training process is detailed in the above corresponding embodiment and will not be described in detail here.

[0268] Step 603: Based on the three-dimensional user portrait of the user group to which the user to be tested belongs, generate a personalized service package corresponding to the user group.

[0269] In this embodiment of the present invention, the 3D-UPES model allows for precise segmentation of different types of telecom users (i.e., into distinct user groups). These groups exhibit unique characteristics in terms of consumption behavior, infrastructure usage, and social attributes. Therefore, this implementation allows for the development of tailored pricing and service (marketing) strategies tailored to each user group to meet the needs of enterprise users, while also improving customer satisfaction and enterprise revenue through value-added services and customized packages.

[0270] Among them, the classification standard

[0271] In this embodiment, the collected actual user data is used to generate characteristics of different user groups through the 3D-UPES model, and the standards of the corresponding groups are improved based on different user groups, as shown in Table 4:

[0272] Table 4

[0273]

[0274]

[0275] Based on the actual classification, the user portrait data of different user groups are shown in Table 5:

[0276] Table 5

[0277]

[0278] Therefore, in this step, based on the three-dimensional user portrait of the user group to which the user to be tested belongs, the personalized service package of the corresponding user group is allocated to the user to be tested, wherein the personalized service package can be one set or multiple sets, which is not limited in this embodiment.

[0279] Step 604: Recommend the personalized service package to the user to be tested.

[0280] In this step, the recommendation method can be recommended to the user to be tested through various methods such as text messages, emails, APP push notifications, etc., wherein it should be noted that the user to be tested can be a telecom user, or multiple users of a telecom enterprise (referred to as the enterprise, etc.), of course, it can also be a non-telecom user, or multiple users of a non-telecom enterprise (also referred to as the enterprise, etc.).

[0281] In this embodiment, based on the classification criteria generated by existing telecom users and combined with the current user's data, one or more recommended plans are recommended to the enterprise user to be converted. The recommended plans can be sorted from high to low by score, displaying the plans that best meet the customer's needs. As shown in Table 6, Table 6 uses Enterprises A, B, and C as examples, but in actual applications, this is not limited to these.

[0282] Table 6

[0283]

[0284] In other words, in this embodiment of the present invention, package recommendations can be pushed to users via various methods, including SMS, email, and app push notifications. The push notifications are concise and clear, offering quick response mechanisms such as "one-click ordering." Service activities are designed to reach users through multiple channels, and service strategies are adjusted in real time based on user responses. If a user expresses interest in a package but doesn't purchase it, further discounts or package upgrades can be offered later.

[0285] Furthermore, the method can also track and evaluate the effectiveness of service activities recommended to users, analyze key indicators such as user conversion rate and purchase rate, and optimize recommendation algorithms and service strategies based on the evaluation results to improve user satisfaction.

[0286] In an embodiment of the present invention, the consumption behavior data, infrastructure data, and social attribute data of the user to be tested are obtained; the infrastructure data, social attribute data, and consumption behavior data are input into a trained multidimensional evaluation model for prediction to obtain the user group to which the user to be tested belongs; based on the three-dimensional user portrait of the user group to which the user to be tested belongs, a personalized service package corresponding to the user group is generated for the user to be tested; and the personalized service package is recommended to the user to be tested. In other words, the present invention obtains and integrates the user's consumption behavior data, infrastructure data, and social attribute data, uses multi-order clustering analysis to identify the behavior patterns and communication needs of different customers, and through data processing and advanced algorithm models, accurately and efficiently recommends personalized services to users. It can also convert non-telecom users into telecom users, thereby achieving precise business expansion and personalized marketing.

[0287] The embodiments of the present invention include: data collection and processing, model construction (model training), model verification, customer prediction, etc. First, message behavior data from a large amount of basic data accumulated in daily operations is obtained from a telecommunications shared data pool, and infrastructure data and social attribute data from enterprise operation and management data are obtained from an external public data access website, and saved in a data inventory through a message interface. At the same time, the three types of data obtained are preprocessed (such as cleaning and standardization, etc.), and then the preprocessed three types of data are used to create a data model from multiple dimensions (i.e., a multidimensional evaluation model). The creation can be specifically achieved by first-order clustering of the three types of data to divide different user groups, and then second-order clustering of the different user groups obtained from the first-order clustering is performed, that is, the complex boundary data of different user groups are converged through second-order clustering refinement processing, and finally a telecommunications enterprise customer data model system (i.e., a multidimensional evaluation model with three-dimensional user portraits of user groups with different clusters) is established. By using this multi-dimensional evaluation model, we analyze and predict the infrastructure data and social attribute data of non-telecom users, match potential users with telecom user groups based on the prediction results, generate corresponding personalized packages for potential users, and push personalized tariff and package recommendations to potential customers, achieving accurate user identification and marketing services, thereby improving the efficiency of business expansion.

[0288] It should be noted that for the method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that this disclosure is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions involved are not necessarily required by the present invention.

[0289] Figure 8This is a block diagram of a multi-dimensional evaluation model training device provided by an embodiment of the present invention. The device includes: an acquisition module 801, a mapping module 802, a multi-order clustering module 803 and an aggregation module 804, wherein:

[0290] An acquisition module 801 is configured to acquire standard user data of multiple users, wherein the standard user data includes consumption behavior data, infrastructure data, and social attribute data;

[0291] A mapping module 802 is configured to map each standard user data into a three-dimensional space to obtain a three-dimensional feature vector of the corresponding user's consumption behavior, infrastructure usage, and social attributes;

[0292] The multi-stage clustering module 803 performs iterative multi-stage clustering training on the three-dimensional feature vectors of all users to obtain user groups in different clusters;

[0293] Aggregation module 804 is used to aggregate the characteristic vectors of consumption behavior, infrastructure usage and social attributes of the user groups of different clusters to obtain a multi-dimensional evaluation model of the user group portraits of different clusters.

[0294] Optionally, in another embodiment, based on the above embodiment, the acquisition module includes:

[0295] A multi-dimensional data acquisition module, configured to acquire user data of multiple enterprise users, wherein the user data includes consumption behavior data, infrastructure data, and social attribute data;

[0296] The preprocessing module is used to preprocess each user data to obtain preprocessed standard user data.

[0297] Optionally, in another embodiment, based on the above embodiment, the multi-stage clustering module includes:

[0298] The first-order clustering module is used to iteratively divide the three-dimensional feature vectors of all users into first-order clusters to obtain user groups in different clusters;

[0299] The second-order clustering module is used to perform iterative optimization processing of second-order clustering on the boundary data of the user groups of different clusters divided by the first-order clustering module to obtain the optimized user groups.

[0300] Optionally, in another embodiment, based on the above embodiment, the first-order clustering module includes:

[0301] A cluster center determination module is used to determine multiple cluster centers by iterative calculation based on the three-dimensional feature vectors of all users in accordance with minimization of squared error, where each cluster center is the cluster that minimizes the distance between each user's three-dimensional feature vector and the corresponding cluster center;

[0302] The cluster division module is used to perform cluster division on all three-dimensional feature vectors of users based on the determined multiple cluster centers to obtain user groups in different clusters after division.

[0303] Optionally, in another embodiment, based on the above embodiment, the second-order clustering module includes:

[0304] A calculation module is used to calculate the posterior probability that each user's three-dimensional feature vector belongs to the corresponding cluster after the user groups are divided into different clusters;

[0305] An iterative update module is used to iteratively update the model parameter mean, covariance matrix and prior probability in the posterior probability based on the posterior probability and using the maximized log-likelihood function until the parameter mean, covariance matrix and prior probability converge to obtain an optimized user group.

[0306] Optionally, in another embodiment, based on the above embodiment, the device further includes:

[0307] A data acquisition module is used to acquire a verification data set, wherein the verification data set includes: sample data of multiple users, wherein the sample data includes: consumption behavior sample data, infrastructure sample data, and social attribute sample data;

[0308] The validation test module is used to perform cross-validation and A / B testing on the multidimensional evaluation model using the validation data set to obtain a tested multidimensional evaluation model.

[0309] See also Figure 9 , is a block diagram of a user service recommendation device according to an embodiment of the present invention, the device comprising: an acquisition module 901, a prediction module 902, a generation module 903 and a recommendation module 904, wherein,

[0310] Acquisition module 901, used to acquire consumption behavior data, infrastructure data and social attribute data of the user to be tested;

[0311] Prediction module 902 is configured to input the infrastructure data, the social attribute data, and the consumption behavior data into a trained multidimensional evaluation model for prediction, thereby obtaining the user group to which the user to be tested belongs. The multidimensional evaluation model is based on mapping the acquired standard user data of multiple users into a three-dimensional space and iteratively training the obtained three-dimensional feature vectors using multi-stage clustering to obtain user groups in different clusters, and a model obtained by aggregating the feature vectors of the user groups in different clusters.

[0312] A generating module 903 is configured to generate a personalized service package for the user to be tested, based on the three-dimensional user portrait of the user group to which the user to be tested belongs;

[0313] The recommendation module 904 is configured to recommend the personalized service package to the user to be tested.

[0314] Optionally, in another embodiment, based on the above embodiment, the acquisition module includes: a first acquisition module and a regression analysis module; and / or a second acquisition module, wherein:

[0315] The first acquisition module is used to acquire infrastructure data and social attribute data of the user to be tested;

[0316] The regression analysis module is used to perform regression analysis on the infrastructure data and the social attribute data to predict the potential consumption behavior data of the user to be tested;

[0317] The second acquisition module is used to directly acquire the consumption behavior data, infrastructure data and social attribute data of the user to be tested.

[0318] Optionally, in another embodiment, based on the above embodiment, the regression analysis module includes:

[0319] A principal component processing module, configured to perform principal component analysis on the infrastructure data and the social attribute data to obtain simplified infrastructure data and social attribute data;

[0320] The consumption behavior prediction module is used to input the simplified infrastructure data and the social attribute data into a regression model for prediction, so as to obtain the potential consumption behavior data of the user to be tested.

[0321] Optionally, an embodiment of the present invention further provides an electronic device, including:

[0322] processor;

[0323] a memory for storing instructions executable by the processor;

[0324] The processor is configured to execute the instructions to implement the multi-dimensional evaluation model training method or the user service recommendation method as described above.

[0325] An embodiment of the present invention also provides a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute the multidimensional evaluation model training method or the user service recommendation method as described above.

[0326] An embodiment of the present invention also provides a computer program product, including a computer program or instructions, which, when executed by a processor of an electronic device, implements the multidimensional evaluation model training method or the user service recommendation method as described above.

[0327] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0328] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0329] See also Figure 10 , is a block diagram of an electronic device 1000 provided in an embodiment of the present invention. For example, electronic device 1000 can be a mobile terminal or a server. The present invention uses a mobile terminal as an example for description. For example, electronic device 1000 can be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0330] like Figure 10 As shown, electronic device 1000 may include one or more of the following components: a processing component 1002 , a memory 1004 , a power component 1006 , a multimedia component 1008 , an audio component 1010 , an input / output (I / O) interface 1012 , a sensor component 1014 , and a communication component 1016 .

[0331] The processing component 1002 generally controls the overall operation of the electronic device 1000, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 1002 may include one or more processors 1020 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 1002 may include one or more modules to facilitate interaction between the processing component 1002 and other components. For example, the processing component 1002 may include a multimedia module to facilitate interaction between the multimedia component 1008 and the processing component 1002.

[0332] The memory 1004 is configured to store various types of data to support the operation of the device 1000. Examples of such data include instructions for any application or method operating on the electronic device 1000, contact data, phone book data, messages, pictures, videos, etc. The memory 1004 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0333] The power supply assembly 1006 provides power to the various components of the electronic device 1000. The power supply assembly 1006 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 1000.

[0334] The multimedia component 1008 includes a screen that provides an output interface between the electronic device 1000 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 1008 includes a front camera and / or a rear camera. When the device 1000 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have a focal length and optical zoom capability.

[0335] The audio component 1010 is configured to output and / or input audio signals. For example, the audio component 1010 includes a microphone (MIC), which is configured to receive external audio signals when the electronic device 1000 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 1004 or transmitted via the communication component 1016. In some embodiments, the audio component 1010 also includes a speaker for outputting audio signals.

[0336] I / O interface 1012 provides an interface between processing component 1002 and peripheral interface modules, such as a keyboard, click wheel, buttons, etc. These buttons may include but are not limited to: a home button, volume buttons, a start button, and a lock button.

[0337] The sensor assembly 1014 includes one or more sensors for providing various aspects of the status assessment of the electronic device 1000. For example, the sensor assembly 1014 can detect the open / closed state of the device 1000, the relative positioning of components, such as the display and keypad of the electronic device 1000. The sensor assembly 1014 can also detect changes in the position of the electronic device 1000 or a component of the electronic device 1000, the presence or absence of user contact with the electronic device 1000, the orientation or acceleration / deceleration of the electronic device 1000, and changes in the temperature of the electronic device 1000. The sensor assembly 1014 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 1014 can also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 1014 can also include an accelerometer, a gyroscope, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0338] The communication component 1016 is configured to facilitate wired or wireless communication between the electronic device 1000 and other devices. The electronic device 1000 can access a wireless network based on a communication standard, such as WiFi, an operator network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 1016 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 1016 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0339] In an embodiment, the electronic device 1000 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the multidimensional evaluation model training method or user service recommendation method shown above.

[0340] In an embodiment, a computer-readable storage medium is further provided. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device 1000 is enabled to perform the multi-dimensional evaluation model training method or user service recommendation method shown above. For example, the computer-readable storage medium can be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

[0341] In an embodiment, a computer program product is also provided, including a computer program or instructions. When the computer program or instructions are executed by the processor 1020 of the electronic device 1000, the electronic device 1000 executes the above-mentioned multidimensional evaluation model training method or user service recommendation method.

[0342] Figure 11 1 is a block diagram of a device 1100 for a multi-dimensional evaluation model training method or a user service recommendation method provided by an embodiment of the present invention. For example, the device 1100 can be provided as a server. Figure 11 The apparatus 1100 includes a processing component 1122, which further includes one or more processors, and memory resources represented by a memory 1132 for storing instructions, such as applications, that can be executed by the processing component 1122. The application stored in the memory 1132 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1122 is configured to execute the instructions to perform the above-described method.

[0343] The device 1100 may also include a power supply component 1126 configured to perform power management of the device 1100, a wired or wireless network interface 1150 configured to connect the device 1100 to a network, and an input / output (I / O) interface 1158. The device 1100 may operate based on an operating system stored in the memory 1132, such as Windows Server™, MacOS X™, Unix™, Linux™, FreeBSD™, or the like.

[0344] The user data involved in the present invention (including but not limited to the user's device information, user personal information, etc.) and related data are all information authorized by the user or authorized by all parties.

[0345] Other embodiments of the present invention will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

[0346] It should be understood that the present invention is not limited to the exact construction described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A training method for a multidimensional evaluation model, characterized in that: include: Acquire standard user data of multiple users, wherein the standard user data includes: consumption behavior data, infrastructure data, and social attribute data; Map each standard user data into three-dimensional space to obtain the corresponding three-dimensional feature vector of the user's consumption behavior, infrastructure usage and social attributes; Perform iterative multi-stage clustering training on the three-dimensional feature vectors of all users to obtain user groups in different clusters; The consumption behaviors, infrastructure usage and social attribute feature vectors of the user groups of different clusters are aggregated to obtain a multi-dimensional evaluation model of the three-dimensional user portraits of the user groups of different clusters.

2. The training method of the multidimensional evaluation model according to claim 1, characterized in that: The iterative training of multi-stage clustering is performed on the three-dimensional feature vectors of all users to obtain different user groups, including: Perform iterative first-order clustering on the three-dimensional feature vectors of all users to obtain user groups in different clusters; The boundary data of the user groups in different clusters after division are iteratively refined by second-order clustering to obtain refined user groups.

3. The training method of the multidimensional evaluation model according to claim 2, characterized in that: The three-dimensional feature vectors of all users are iteratively divided into first-order clusters to obtain user groups in different clusters, including: Based on the three-dimensional feature vectors of all users, iterative calculation is performed according to minimizing the square error to determine multiple cluster centers. Each cluster center is the cluster that minimizes the distance between each user's three-dimensional feature vector and the corresponding cluster center. Based on the determined multiple cluster centers, cluster division is performed on the three-dimensional feature vectors of all users to obtain user groups in different clusters after division.

4. The training method of the multidimensional evaluation model according to claim 2, characterized in that: The second-order clustering iterative optimization process is performed on the boundary data of each divided user group to obtain the optimized user group, including: For the user groups divided into different clusters, calculate the posterior probability that each user's three-dimensional feature vector belongs to the corresponding cluster; Based on the posterior probability, the parameter mean, covariance matrix and prior probability in the posterior probability are iteratively updated by maximizing the log-likelihood function; until the parameter mean, covariance matrix and prior probability converge, an optimized user group is obtained.

5. A user service recommendation method, characterized in that: include: Obtain consumption behavior data, infrastructure data, and social attribute data of the users to be tested; Inputting the infrastructure data, the social attribute data, and the consumption behavior data into a trained multidimensional evaluation model for prediction to obtain the user group to which the user to be tested belongs, wherein the multidimensional evaluation model is based on mapping the obtained standard user data of multiple users into a three-dimensional space and performing iterative multi-order clustering training on the obtained three-dimensional feature vectors to obtain user groups in different clusters, and aggregating the feature vectors of the user groups in different clusters to obtain a model; Based on the three-dimensional user portrait of the user group to which the user to be tested belongs, generating a personalized service package corresponding to the user group for the user to be tested; The personalized service package is recommended to the user to be tested.

6. The user service recommendation method according to claim 5, characterized in that: The acquisition of consumption behavior data, infrastructure data, and social attribute data of the user to be tested includes: Obtain infrastructure data and social attribute data of the user to be tested, perform regression analysis on the infrastructure data and the social attribute data, and predict potential consumption behavior data of the user to be tested; or Directly obtain the consumption behavior data, infrastructure data and social attribute data of the user to be tested.

7. The user service recommendation method according to claim 6, characterized in that: The performing of regression analysis on the acquired infrastructure data and social attribute data to predict the user's potential consumption behavior data includes: Performing principal component analysis on the infrastructure data and the social attribute data to obtain simplified infrastructure data and social attribute data; The simplified infrastructure data and the simplified social attribute data are input into a regression model for prediction to obtain the potential consumption behavior data of the user to be tested.

8. A multidimensional evaluation model training device, characterized in that: include: An acquisition module, configured to acquire standard user data of multiple users, wherein the standard user data includes consumption behavior data, infrastructure data, and social attribute data; A mapping module, configured to map each standard user data into a three-dimensional space to obtain a three-dimensional feature vector of the corresponding user's consumption behavior, infrastructure usage, and social attributes; The multi-stage clustering module performs iterative multi-stage clustering training on the three-dimensional feature vectors of all users to obtain user groups in different clusters; The aggregation module is used to aggregate the consumption behaviors, infrastructure usage and social attribute feature vectors of the user groups of different clusters to obtain a multi-dimensional evaluation model of the user group portraits of different clusters.

9. A user service recommendation device, characterized in that: include: The acquisition module is used to obtain the consumption behavior data, infrastructure data and social attribute data of the user to be tested; a prediction module, configured to input the infrastructure data, the social attribute data, and the consumer behavior data into a trained multidimensional evaluation model for prediction, thereby obtaining the user group to which the user to be tested belongs, wherein the multidimensional evaluation model is based on mapping the acquired standard user data of multiple users into a three-dimensional space and performing iterative multi-stage clustering training on the obtained three-dimensional feature vectors to obtain user groups in different clusters, and a model obtained by aggregating the feature vectors of the user groups in different clusters; A generating module, configured to generate a personalized service package for the user to be tested, based on a three-dimensional user portrait of the user group to which the user to be tested belongs; The recommendation module is used to recommend the personalized service package to the user to be tested.

10. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the multidimensional evaluation model training method according to any one of claims 1 to 4 or the user service recommendation method according to any one of claims 5 to 7.

11. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the multidimensional evaluation model training method as described in any one of claims 1 to 4 or the user service recommendation method as described in any one of claims 5 to 7.

Images