Communication service evaluation method and system

Abstract

The present invention belongs to the field of communication technology, and provides a communication service evaluation method and system. The method includes: automatically collecting user data of each service provider, encrypting the user data, and uploading the initial model parameters of each service provider; each service provider trains a sub-communication service evaluation model according to each local user data and initial model parameters; according to the model parameters uploaded by each service provider, federated learning is performed to obtain an optimized global communication service evaluation model; the user data to be processed is input into the global communication service evaluation model to obtain a communication service evaluation result. The present invention can make full use of the user data of each communication service provider, and adopt the established global communication service evaluation model to obtain a more accurate communication service evaluation result.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a communication service evaluation method and system. Background Art

[0002] With the rapid development of the telecommunications industry, the quality of telecommunications services has become the focus of consumers. Traditional telecommunications service evaluation systems and methods usually rely on a single data center for data collection and processing, which has problems such as one-sided data and low processing efficiency due to only manual participation. Existing telecommunications service evaluation systems usually have data privacy protection issues and cannot fully utilize the data resources of various service providers, resulting in incomplete and inaccurate evaluation results. Federated learning is a distributed machine learning technology that allows participants to train and update models on local data while maintaining data security and privacy. In addition, there is still much room for improvement in how to achieve model accuracy for multiple participants and further improve data security and privacy in specific scenarios.

[0003] Therefore, it is necessary to provide a communication service evaluation method and system to solve the above problems. Summary of the invention

[0004] The present invention intends to provide a communication service evaluation method and system to solve the technical problems in the prior art that the communication service data of a single service provider is incomplete, resulting in the inability to provide customized communication services to each user, the low accuracy of the existing communication service evaluation method, and how to further improve data security and privacy in specific scenarios. The technical problems to be solved by the present invention are achieved through the following technical solutions.

[0005] The first aspect of the present invention proposes a communication service evaluation method, including: automatically collecting user data of each service provider, encrypting the user data, and uploading the initial model parameters of each service provider; each service provider performs sub-communication service evaluation model training based on the local user data and initial model parameters, specifically including establishing each local training data for training the corresponding sub-communication service evaluation model based on the following feature data: user number, communication service time, cost, call quality, network coverage, and evaluation data corresponding to each feature data; group training corresponding to the user group, training once to obtain updated model parameters; performing federated learning based on the model parameters uploaded by each service provider to obtain an optimized global communication service evaluation model; inputting the user data to be processed into the global communication service evaluation model to obtain a communication service evaluation result.

[0006] The second aspect of the present invention proposes a communication service evaluation system, which adopts the communication service evaluation method described in the first aspect of the present invention, and the communication service evaluation system includes: a data collection module, which is used to automatically collect user data of each service provider, encrypt the user data, and upload the initial model parameters of each service provider; a sub-model training module, each service provider performs sub-communication service evaluation model training based on local user data and initial model parameters, specifically including establishing local training data based on the following feature data for training corresponding sub-communication service evaluation models: user number, communication service time, cost, call quality, network coverage, and evaluation data corresponding to each feature data; group training corresponding to user groups, each training obtains updated model parameters; a federated learning module, which performs federated learning based on the model parameters uploaded by each service provider to obtain an optimized global communication service evaluation model; a prediction module, which inputs the user data to be processed into the global communication service evaluation model to obtain a communication service evaluation result.

[0007] The third aspect of the present invention provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the communication service evaluation method described in the first aspect of the present invention.

[0008] A fourth aspect of the present invention provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the communication service evaluation method described in the first aspect of the present invention.

[0009] The embodiments of the present invention include the following advantages:

[0010] Compared with the prior art, the present invention automatically collects user data from each service provider, divides the user data into user groups, and establishes a sub-communication service evaluation model corresponding to the user group. By optimizing the model parameters of the global communication service evaluation model using the model parameters of each sub-communication service evaluation model, a global communication service evaluation model with more comprehensive user data can be obtained, and thus a more accurate communication service evaluation result can be obtained, which effectively overcomes the problem of incomplete data and insufficient data processing of a single communication service provider; according to the communication service evaluation results determined by the global communication service evaluation model, customized communication services that are more in line with user characteristics can be provided for various user groups more effectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 is a flowchart of an example of the communication service evaluation method of the present invention;

[0012] Figure 2It is a partial flow chart of determining a user group in the communication service evaluation method of the present invention;

[0013] Figure 3 is a structural block diagram of the communication service evaluation system of the present invention;

[0014] Figure 4 is a schematic structural diagram of an electronic device embodiment according to the present invention;

[0015] Figure 5 It is a schematic diagram of the structure of a computer-readable medium embodiment according to the present invention. DETAILED DESCRIPTION

[0016] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0017] In view of the above problems, the present invention provides a communication service evaluation method, which automatically collects user data of each service provider, encrypts the user data, and uploads the initial model parameters of each service provider; each service provider performs sub-communication service evaluation model training based on the local user data and initial model parameters, specifically including establishing local training data for training corresponding sub-communication service evaluation models based on the following feature data: user number, communication service time, fee, call quality, network coverage, and evaluation data corresponding to each feature data; group training corresponding to the user group, and obtaining updated model parameters after one round of training; performing federated learning based on the model parameters uploaded by each service provider to obtain an optimized global communication service evaluation model; inputting the user data to be processed into the global communication service evaluation model to obtain a communication service evaluation result.

[0018] Example 1

[0019] Refer to the following Figure 1 , Figure 2 , the contents of the present invention will be described in detail.

[0020] Figure 1 It is a flowchart of an example of the communication service evaluation method of the present invention.

[0021] like Figure 1 As shown, in step S101, user data of each service provider is automatically collected, encrypted, and the initial model parameters of each service provider are uploaded.

[0022] In a specific application example, a data center system is included, and the data center system includes a software system. The software system integrates a data collection module, a sub-model training module, a federated learning module, and a prediction module. Specifically, the user communication service is evaluated based on a distributed model, and factors such as service time, cost, call quality, and network coverage are comprehensively considered to form an evaluation result, and the evaluation result is fed back to the user to provide optimization suggestions for telecommunications services. At the same time, user feedback information, such as service quality evaluation, new service requirements, etc., is collected for further optimization of the software system.

[0023] Optionally, a hardware system is also included. The terminal devices in the hardware system include various types of mobile phones, tablet computers, smart watches, etc., which are mainly used to collect user usage data. The data transmission equipment in the hardware system includes routers and switches, which are used to transmit the data collected by the terminal devices to the data center without loss; the data storage devices in the hardware system include servers and disk arrays, etc., which are used to store and back up data, as well as run the software system. The data collection module in the software system is connected to the terminal device through various interfaces, automatically collects the user's usage data, and performs encryption processing; the federated learning algorithm module in the software system uses a federated learning algorithm to distribute data to each device participating in the training, and each device generates a model based on its own data, and shares the updated model with other devices to further update the data and model.

[0024] Specifically, the data collection module automatically collects user data of each service provider, encrypts the user data, and uploads the initial model parameters of each service provider to the corresponding data collection module of the data center system.

[0025] There are at least three service providers, and the user data of each service provider cannot be shared. Each service provider provides communication services to the user.

[0026] It should be noted that in this example, the communication services include fixed-line telephone, mobile phone, broadband Internet and other services. In another embodiment, the communication services also include value-added communication services, specifically including SMS, MMS, voice mail, Internet phone, etc. In another embodiment, cloud computing services are also included, specifically including the provision of cloud hosting, cloud storage, cloud desktop and other services. The above are only described as optional examples and cannot be understood as limitations on the present invention.

[0027] Specifically, the user data includes user usage data disclosed on the Internet (such as shopping orders, monthly order quantity), user phone numbers (such as telephone numbers, WeChat numbers, etc.), communication service time, fees, call quality, network coverage, user gender, annual income or monthly income, network connection length, residence information, credit repayment record data, etc.

[0028] Processing of user data of each service provider, including removal of outliers.

[0029] Specifically, the following expression is used to normalize the numerical features and remove the sample data that is not within (μ-3σ, μ+3σ):

[0030] in, It represents the feature data of the i-th feature data in the sample data of the same user after being normalized by numerical features; represents the i-th feature data value in the same user sample data without numerical feature standardization, where i is a positive integer; μ represents the mean of the i-th feature data of all user sample data; σ represents the variance of the i-th feature data of all user sample data.

[0031] When the frequency of the numerical feature corresponding to a sample data in the data set is less than 0.05, the sample data is removed.

[0032] For discrete features, one-hot encoding is used for processing, and (1,0,0), (0,1,0), and (0,0,1) are used to represent first-tier cities, second-tier cities, and third-tier cities respectively.

[0033] After processing different types of data, merge them into a new data set, and then use the K-Means clustering method to remove outliers, such as Figure 2 As shown, the specific steps include:

[0034] Step S201: Determine the K value.

[0035] In a specific implementation, the silhouette coefficient method is selected to determine the value of K. The silhouette coefficient method is a commonly used indicator for evaluating the effect of clustering analysis, which can help determine whether the data points (ie, the data points corresponding to each user sample) are correctly assigned to the corresponding clusters.

[0036] The calculation steps for determining the K value using the silhouette coefficient method are as follows:

[0037] S(i)=b(i)-a(i)max{a(i),b(i)};

[0038] in:

[0039] S(i) is the silhouette coefficient of the data point corresponding to the i-th user sample;

[0040] a(i) represents the average distance between the data point corresponding to the i-th user sample and other data points in the cluster to which it belongs, which is used to reflect the compactness of the current cluster;

[0041] b(i) represents the average distance between the data point corresponding to the i-th user sample and the data point of the nearest cluster in the non-cluster, which is used to reflect the separation between clusters;

[0042] When b(i)>a(i), it means that the sample point corresponding to the user sample is closer to the cluster to which it belongs, and the silhouette coefficient is positive; conversely, if a(i)>b(i), it means that the sample point may be misclassified, and the silhouette coefficient is negative.

[0043] When the silhouette coefficient is positive, it means that b(i) is much larger than a(i), that is, the data point i corresponding to the user sample is dense in the cluster to which it belongs and well separated from other clusters.

[0044] When the silhouette coefficient is close to 0, it means that a(i) is close to b(i), that is, the data point i corresponding to the user sample is at the cluster boundary of the cluster to which it belongs, or the distance between points within the cluster is equivalent to the distance between clusters.

[0045] When the silhouette coefficient is negative, it means that a(i) is much larger than b(i), that is, the data point i corresponding to the user sample may be incorrectly assigned to the wrong cluster.

[0046] In the invention, K=2, 3, 4, 5, 6 are used in turn to cluster the user groups corresponding to the sub-communication service evaluation model to divide the user groups. For the clustering results under each K value, the silhouette coefficients of all data points are calculated in turn, and then the global silhouette coefficients under each K value are averaged to obtain the global silhouette coefficient. The K value with the largest global silhouette coefficient is selected as the optimal clustering number, and the K value when the global silhouette coefficient is the largest is set to 3.

[0047] When K=3, it means that the user group is clustered into three clusters, representing high-consumption users, medium-consumption users, and low-consumption users respectively.

[0048] Step S202: Select the centroid.

[0049] Randomly select K data points as initial centroids.

[0050] Step S203: User sample allocation.

[0051] Specifically, user samples are assigned to corresponding clusters: the distance from each point in the data set sample to the centroid is calculated, and then the sample point is assigned to the cluster with the nearest centroid.

[0052] Step S204: Update the centroid.

[0053] For each cluster, recalculate its centroid, specifically calculate the average value of all samples in the cluster on each feature, and use the average value of each feature in each cluster as the new centroid.

[0054] Step S205: Repeat the steps of user sample allocation and centroid update until the centroids no longer change, and obtain the clustering result.

[0055] For example, in a specific implementation, user samples are clustered into three categories, namely three clusters, which specifically correspond to the first user group (ie, high-consumption user group), the second user group (ie, medium-consumption user group), and the third user group (ie, low-consumption user group).

[0056] It should be noted that the above is only described as an optional example and should not be understood as a limitation to the present invention.

[0057] Next, in step S102, each service provider performs sub-communication service evaluation model training based on each local user data and initial model parameters, specifically including establishing each local training data for training the corresponding sub-communication service evaluation model based on the following feature data: user number, communication service time, cost, call quality, network coverage, and evaluation data corresponding to each feature data; group training corresponding to the user group, and obtaining updated model parameters each time the training is performed.

[0058] Specifically, according to the different importance of different communication service dimensions for users, the users are grouped according to the specified ratio to obtain multiple user groups. When three user groups (specifically a high-consumption user group, a medium-consumption user group, and a low-consumption user group) are obtained, the specified ratio is, for example, 1:2:7.

[0059] According to different user groups, multiple sub-communication service evaluation models are established, and group training corresponding to the user groups is performed, and updated model parameters are obtained each time the training is performed.

[0060] Based on the following characteristic data formed by each communication service of each user group, a training data set is established to be used for training the respective sub-communication service evaluation models.

[0061] It should be noted that, in this example, the characteristic data includes service time, cost, call quality, network coverage, network access time, and residence information. However, it is not limited to this. In other examples, user usage data (such as shopping orders, monthly order quantity), user phone number (such as telephone number, WeChat number, etc.), user gender, annual income or monthly income, network access time, credit repayment record data, etc. The above are only used as optional examples for explanation and cannot be understood as limiting the present invention.

[0062] Different groups of people have different priorities for service importance. For example, high-consumption users pay more attention to call quality, while low-consumption users pay more attention to cost. That is, different groups of people attach different importance to different dimensions of communication services. Therefore, it is necessary to divide the groups of people and model them. Specifically, divide the user groups first, and then build the sub-communication service evaluation model corresponding to each user group.

[0063] Next, the construction of the sub-communication service evaluation model will be described.

[0064] Specifically, a neural network algorithm is used to establish a sub-communication service evaluation model corresponding to each user group. The sub-communication service evaluation model is expressed by the following expression:

[0065] Y= [A1 +A2 +...+An ],

[0066] Wherein, Y represents the evaluation score output by the sub-communication service evaluation model, that is, the model calculation score or the model prediction score; , ,..., are multiple feature data selected from user data, including service time, cost, call quality, network coverage, network access time, and place of residence information. A1, A2, ..., An represent the weight parameters corresponding to the first feature data, the second feature data, and the nth feature data, respectively, representing the importance of different feature data in the group. The larger An is, the more important it is. It will be continuously updated during the model training process. Add up all items and multiply by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n is a positive integer, and n represents the number of feature data.

[0067] By adjusting and calculating characteristic parameters corresponding to different user groups, the sub-communication service evaluation model is trained for multiple rounds to minimize the difference between the model calculation score and the actual score, so as to determine to stop training the sub-communication service evaluation model and obtain a trained sub-communication service evaluation model.

[0068] In the process of training the sub-communication service evaluation model, the model parameters of the sub-communication service evaluation model are updated every time the training is performed, and the model parameters include characteristic parameters A1, A2, ...An.

[0069] The difference between the model-calculated score and the actual score of the sub-communication service evaluation model is expressed as follows:

[0070] =

[0071] Where E represents the difference between the model calculation score Y of the sub-communication service evaluation model and the real score given by the user The difference between represents the real score given by the user, and Y represents the model calculation score of the sub-communication service evaluation model, that is, the score predicted by the model; , ,..., Respectively represent the first feature data, the second feature data, and the nth feature data; A1, A2, ..., An represent the weight parameters corresponding to the first feature data, the second feature data, and the nth feature data, respectively, representing the importance of different feature data in the group. The larger the An, the more important it is. It will be continuously updated with the model training process. The sub-communication service evaluation model is trained in each group. By collecting multiple groups of user data, each group of data includes feature data , ,..., , add up the items and multiply by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n is a positive integer, n represents the number of feature data, through the model training of the sub-communication service evaluation model, the value of the loss function E is close to 0, that is, the score Y predicted by the model is different from the actual score given by the user When E<0.05, the model converges and a set of trained weight parameters is obtained. , training completed.

[0072] In order to speed up the training, the weight parameters are initialized according to the different importance of each user group to each feature of the communication service. When the user questionnaire survey is conducted in advance, the scores of importance of different features for each group are collected. For example, for the high-spending group, the questionnaire results show that the importance of network coverage, call quality, and cost for this group (for example, the predetermined score is 10) is recorded as 8, 7, and 3. When initializing the parameters for this user group, appropriate fine-tuning is performed. For example, the initialization parameters for the high-spending user group are (0.6, 0.5, 0.2). For example, the training convergence speed of the parameters is accelerated by using empirical judgment.

[0073] For example, for a sub-communication service provider Q, its high, medium, and low consumer groups are denoted as Q i ,Q j ,Q k , for Q i In the group, each user generates a set of data (x 1 ,x 2 ,...,x n ), a total of M groups of data are obtained, that is, for Qi The size of the population dataset is M*(x 1 ,x 2 ,...,x n ).

[0074] For example, in the above example, after each use of the service provided by the operator, the evaluation results of each feature of the service (service time, cost, call quality, network coverage, etc.) will be returned to the user, and a set of results will be recorded as , ,..., For different user groups, the initialization parameters of the corresponding communication service evaluation model are different, and the initialization parameters of the communication service evaluation model corresponding to the same user group are the same.

[0075] In an optional implementation, before grouping users according to a specified ratio, an initial classification is performed based on user gender, annual income, network connection length, and place of residence information to obtain a first user group, a second user group, and a third user group corresponding to high consumption, medium consumption, and low consumption. The number of first user groups: the number of second user groups: the number of third user groups is 1:2:7.

[0076] It should be noted that the above is only described as an optional example and should not be understood as a limitation to the present invention.

[0077] Next, in step S103, federated learning is performed based on the model parameters uploaded by each service provider to obtain an optimized global communication service evaluation model.

[0078] When the number of the sub-communication service evaluation models is m, after the m sub-communication service evaluation models have all completed convergence, the model parameters of the respective m sub-communication service evaluation models are uploaded to the central server.

[0079] Specifically, in this scenario example, the federated learning specifically refers to learning the sensitivity of different user groups to various features in the communication service, and then combining the model data of different service providers to build a global communication service evaluation model, which is applied to various communication service scenarios.

[0080] The following expression is used to represent the global communication service evaluation model:

[0081] ;

[0082] in, represents the model calculation score of the global communication service evaluation model, t represents the t-th user group, and t is a positive integer; c 1 ,c 2 ,...c mThey represent the number of data sets of m sub-communication service evaluation models when they are uploaded to the central server in each round to train the global communication service evaluation model. They are used to adjust the weight ratios of sub-communication service evaluation models of different sizes. Each sub-communication service evaluation model corresponds to a different service provider. m is a positive integer. Specifically, c 1 represents the number of data sets uploaded to the central server for the first round of training of the first sub-communication service evaluation model, c 2 represents the number of data sets uploaded to the central server for the first round of training of the second sub-communication service evaluation model, c m represents the number of data sets uploaded to the central server for the first round of training of the mth sub-communication service evaluation model; , ,..., Respectively represent the first feature data, the second feature data, and the nth feature data, where n is a positive integer; a 1 Indicates the first feature data X under the global communication service evaluation model 1 The weight parameter, a 2 Indicates the second feature data X under the global communication service evaluation model 2 The weight parameter, a m Represents the nth feature data X under the global communication service evaluation model n The weight parameter is: add each item and multiply it by , used to obtain the arithmetic mean of the evaluation scores of each feature data, and n represents the number of feature data.

[0083] The updated global communication service evaluation model will then send the relevant parameters to each sub-communication service evaluation model for training again, and upload the parameters obtained from this training after the model training is completed. The above steps are repeated until the final global communication service evaluation model is obtained, and the number of parameter uploads between each sub-communication service evaluation model and the global communication service evaluation model is greater than or equal to 50 times, or the number of parameter sends between each sub-communication service evaluation model and the global communication service evaluation model is greater than or equal to 50 times.

[0084] For the training of each sub-communication service evaluation model, in order to prevent the model from overfitting, the loss error is not higher than 0.05 and the number of training rounds is not higher than 2000 echos.

[0085] For each service provider, based on the final global communication service evaluation model and cost model, we use linear programming to solve the optimal solution and seek the balance point between high service score and low cost. The cost of the feature data of each dimension is set to be in the range of [1,10], where 1 represents the lowest cost and 10 represents the highest cost. The total cost that each service provider can bear is recorded as u, and the objective function F = Y t-C, let C = u / n, which represents the arithmetic mean of the cost of each feature, so that t Aligned in terms of dimensions, the objective function is F = Y t -C, by solving the following system of equations, so that F is maximized:

[0086]

[0087] Wherein, F represents the global communication service performance index under the t-th user group. The larger the global communication service performance index F is, the higher the user satisfaction is and the lower the cost is. t is a positive integer; Y t The model calculation score of the global communication service evaluation model is used to represent user satisfaction, t represents the t-th user group, c 1 ,c 2 ,...c m They represent the number of data sets of m sub-communication service evaluation models when they are uploaded to the central server in each round to train the global communication service evaluation model. They are used to adjust the weight ratios of sub-communication service evaluation models of different sizes. Each sub-communication service evaluation model corresponds to a different service provider. m is a positive integer. Specifically, c 1 represents the number of data sets uploaded to the central server for the first round of training of the first sub-communication service evaluation model, c 2 represents the number of data sets uploaded to the central server for the first round of training of the second sub-communication service evaluation model, c m represents the number of data sets uploaded to the central server for the first round of training of the mth sub-communication service evaluation model; , ,..., Respectively represent the first feature data, the second feature data, and the nth feature data, where n is a positive integer; a 1 Indicates the first feature data X under the global communication service evaluation model 1 The weight parameter, a 2 Indicates the second feature data X under the global communication service evaluation model 2 The weight parameter, a m Represents the nth feature data X under the global communication service evaluation model n The weight parameter is to add each item and multiply it by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n represents the number of feature data;

[0088] C represents the cost model of the service provider corresponding to the t-th user group, and u represents the total cost that each service provider can bear, that is, 0≤ + +...+ ≤u, i is a positive integer, specifically 1, 2, 3, ..., n, n represents the number of feature data, u i Indicates the cost value of the feature data of the first latitude, 1≤ ≤10, Represents the cost value of the feature data of the second dimension, 1≤ ≤10, Represents the cost value of the feature data at the nth latitude, 1≤ ≤10;

[0089] By solving the problem, we can finally get a set of optimal solutions. , ,..., , is used to represent the cost coefficient of the t-th user group.

[0090] In an optional implementation, before the user data to be processed is input into the global communication service evaluation model, a user group corresponding to the user data to be processed is determined, and model parameters and comparison thresholds of the global communication service evaluation model are selected based on the determined user group.

[0091] Specifically, the comparison threshold of the global communication service evaluation model is selected according to the ratio of the number of user groups. For example, the comparison threshold of the global communication service evaluation model is selected according to the ratio of the number of the first user group: the number of the second user group: the number of the third user group (for example, 1:2:7).

[0092] It should be noted that the above is only described as an optional example and should not be understood as a limitation to the present invention.

[0093] Next, in step S104, the user data to be processed is input into the global communication service evaluation model to obtain a communication service evaluation result.

[0094] For example, the input data is a communication service data feature, such as the communication service data feature of the current number of times, and the communication service data feature of the current number of times is input into the global communication service evaluation model to obtain a communication service evaluation result.

[0095] According to the communication service evaluation result obtained, the service level corresponding to the user data to be processed is determined, for example, the score of the user group corresponding to the communication service data feature of the current number of times is output.

[0096] It should be noted that the above is only described as an optional example and should not be understood as a limitation to the present invention.

[0097] In addition, the drawings are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the drawings do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0098] Compared with the prior art, the present invention automatically collects user data from each service provider, divides the user data into user groups, and establishes a sub-communication service evaluation model corresponding to the user group. By optimizing the model parameters of the global communication service evaluation model using the model parameters of each sub-communication service evaluation model, a global communication service evaluation model with more comprehensive user data can be obtained, and thus a more accurate communication service evaluation result can be obtained, which effectively overcomes the problem of incomplete data and insufficient data processing of a single communication service provider; according to the communication service evaluation results determined by the global communication service evaluation model, customized communication services that are more in line with user characteristics can be provided for various user groups more effectively.

[0099] Example 2

[0100] The following is a system embodiment of the present invention, which can be used to implement the method embodiment of the present invention. For details not disclosed in the system embodiment of the present invention, please refer to the method embodiment of the present invention.

[0101] Figure 3 FIG. 4 is a schematic diagram of the structure of an example of a communication service evaluation system according to the present invention.

[0102] The second aspect of the present disclosure provides a communication service evaluation system 300, which adopts the communication service evaluation method described in the first aspect of the present disclosure.

[0103] Reference Figure 3 The communication service evaluation system 300 includes a data collection module 310, a sub-model training module 320, a federated learning module 330 and a prediction module 340.

[0104] In a specific implementation, the data collection module 310 is used to automatically collect user data of each service provider, encrypt the user data, and upload the initial model parameters of each service provider.

[0105] In the sub-model training module 320, each service provider performs sub-communication service evaluation model training based on each local user data and initial model parameters, specifically including establishing each local training data for training the corresponding sub-communication service evaluation model based on the following feature data: user number, communication service time, cost, call quality, network coverage, and evaluation data corresponding to each feature data; group training corresponding to the user group, and obtaining updated model parameters after each training; the federated learning module 330 performs federated learning based on the model parameters uploaded by each service provider to obtain an optimized global communication service evaluation model; the prediction module 340 inputs the user data to be processed into the global communication service evaluation model to obtain the communication service evaluation result.

[0106] According to an optional implementation manner, based on the different importance of different dimensions of communication services to users, the users are grouped in a specified ratio to obtain multiple user groups; multiple sub-communication service evaluation models are established based on different user groups, and group training corresponding to the user groups is performed, and the model parameters are updated once the training is performed; wherein, based on the following feature data formed by each communication service of each user group, a training data set is established for training the respective sub-communication service evaluation models: service time, cost, call quality, network coverage, network access time, and residence information.

[0107] According to an optional implementation, before grouping users according to a specified ratio, a preliminary classification is performed according to user gender, annual income, network access time, and place of residence information to obtain a first user group, a second user group, and a third user group corresponding to high consumption, medium consumption, and low consumption. The specified ratio is, for example, 1:2:7. Specifically, the number of first user groups: the number of second user groups: the number of third user groups is, for example, 1:2:7.

[0108] According to an optional implementation manner, a neural network algorithm is used to establish a sub-communication service evaluation model corresponding to each user group, and the sub-communication service evaluation model is expressed by the following expression:

[0109] Y= [A1 +A2 +...+An ], wherein Y represents the evaluation score output by the sub-communication service evaluation model; , ,..., are multiple feature data selected from user data, including service time, cost, call quality, network coverage, network access time, and place of residence information. A1, A2, ..., An represent the weight parameters corresponding to the first feature data, the second feature data, and the nth feature data, respectively, representing the importance of different feature data in the group. The larger An is, the more important it is. It will be continuously updated during the model training process. Add up all items and multiply by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n is a positive integer, and n represents the number of feature data.

[0110] By adjusting and calculating characteristic parameters corresponding to different user groups, the sub-communication service evaluation model is trained for multiple rounds to minimize the difference between the model calculation score and the actual score, so as to determine to stop training the sub-communication service evaluation model and obtain a trained sub-communication service evaluation model.

[0111] According to an optional implementation manner, the difference between the model-calculated score of the sub-communication service evaluation model and the actual score given by the user is expressed as follows:

[0112] = Where E represents the difference between the model calculation score Y of the sub-communication service evaluation model and the real score given by the user , Y represents the model calculation score of the sub-communication service evaluation model, that is, the score predicted by the model, Represents the real rating given by the user; , ,..., They represent the first feature data, the second feature data, and the nth feature data respectively. A1, A2, ..., An represent the weight parameters corresponding to the first feature data, the second feature data, and the nth feature data respectively, representing the importance of different feature data in the group. The larger the An, the more important it is. It will be continuously updated during the model training process. The sub-communication service evaluation model is trained in each group. By collecting multiple groups of user data, each group of data includes feature data. , ,..., , add up the items and multiply by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n is a positive integer, n represents the number of feature data, through the model training of the sub-communication service evaluation model, the value of the loss function E is close to 0, that is, the score Y predicted by the model is different from the actual score given by the user When E<0.05, the model converges and a set of trained weight parameters is obtained. , training completed.

[0113] The following expression is used to represent the global communication service evaluation model:

[0114] ;

[0115] in, represents the model calculation score of the global communication service evaluation model, t represents the tth user group; c 1 ,c 2 ,...c m They represent the number of data sets of m sub-communication service evaluation models when they are uploaded to the central server in each round to train the global communication service evaluation model. They are used to adjust the weight ratios of sub-communication service evaluation models of different sizes. Each sub-communication service evaluation model corresponds to a different service provider. m is a positive integer. Specifically, c 1 represents the number of data sets uploaded to the central server for the first round of training of the first sub-communication service evaluation model, c 2 represents the number of data sets uploaded to the central server for the first round of training of the second sub-communication service evaluation model, c m represents the number of data sets uploaded to the central server for the first round of training of the mth sub-communication service evaluation model; , ,..., Respectively represent the first feature data, the second feature data, and the nth feature data, where n is a positive integer; a 1 Indicates the first feature data X under the global communication service evaluation model 1 The weight parameter, a 2 Indicates the second feature data X under the global communication service evaluation model 2 The weight parameter, a m Represents the nth feature data X under the global communication service evaluation model n The weight parameter is to add each item and multiply it by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n represents the number of feature data;

[0116] The updated global communication service evaluation model then sends the relevant parameters to each sub-communication service evaluation model for training again, and uploads the parameters obtained from this training after the model training is completed, repeating the above steps until the final global communication service evaluation model is obtained, and the number of parameter uploads between each sub-communication service evaluation model and the global communication service evaluation model is greater than or equal to 50 times, or the number of parameter sends between each sub-communication service evaluation model and the global communication service evaluation model is greater than or equal to 50 times;

[0117] For the training of each sub-communication service evaluation model, in order to prevent the model from overfitting, the loss error is not higher than 0.05 and the number of training rounds is not higher than 2000 echos.

[0118] According to an optional implementation method, for each service provider, based on the final global communication service evaluation model and cost model, a linear programming solution is used to find the optimal solution to seek a balance between high service score and low cost. The cost of the feature data of each dimension is set to be in the interval of [1,10], 1 represents the lowest cost, 10 represents the highest cost, and the total cost that each service provider can bear is recorded as u. Let C = u / n, which represents the arithmetic mean of the feature costs, so that it is consistent with Y t Aligned in terms of dimensions, the objective function is F = Y t -C, by solving the following system of equations to maximize F:

[0119]

[0120] Wherein, F represents the global communication service performance index under the t-th user group. The larger the global communication service performance index F is, the higher the user satisfaction is and the lower the cost is. t is a positive integer; Y t The model calculation score of the global communication service evaluation model is used to represent user satisfaction, t represents the t-th user group, c 1 ,c 2 ,...c m They represent the number of data sets of m sub-communication service evaluation models when they are uploaded to the central server in each round to train the global communication service evaluation model. They are used to adjust the weight ratios of sub-communication service evaluation models of different sizes. Each sub-communication service evaluation model corresponds to a different service provider. m is a positive integer. Specifically, c 1 represents the number of data sets uploaded to the central server for the first round of training of the first sub-communication service evaluation model, c 2 represents the number of data sets uploaded to the central server for the first round of training of the second sub-communication service evaluation model, c m represents the number of data sets uploaded to the central server for the first round of training of the mth sub-communication service evaluation model; , ,..., Respectively represent the first feature data, the second feature data, and the nth feature data, where n is a positive integer; a 1 Indicates the first feature data X under the global communication service evaluation model 1 The weight parameter, a 2 Indicates the second feature data X under the global communication service evaluation model 2 The weight parameter, a mRepresents the nth feature data X under the global communication service evaluation model n The weight parameter is to add each item and multiply it by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n represents the number of feature data;

[0121] C represents the cost model of the service provider corresponding to the t-th user group, and u represents the total cost that each service provider can bear, that is, 0≤ + +...+ ≤u, i is a positive integer, specifically 1, 2, 3, ..., n, n represents the number of feature data, u i Indicates the cost value of the feature data of the first latitude, 1≤ ≤10, Represents the cost value of the feature data of the second dimension, 1≤ ≤10, Represents the cost value of the feature data at the nth latitude, 1≤ ≤10.

[0122] By solving the problem, we can finally get a set of optimal solutions. , ,..., , is used to represent the cost coefficient of the t-th user group.

[0123] According to an optional implementation manner, removing outliers specifically includes:

[0124] Use the following expression to normalize the numerical features and remove the sample data that is not within (μ-3σ, μ+3σ):

[0125] in, It represents the feature data of the i-th feature data in the sample data of the same user after being normalized by numerical features; represents the i-th feature data value in the sample data of the same user without numerical feature standardization, i is a positive integer; μ represents the mean of the i-th feature data of all user sample data; σ represents the variance of the i-th feature data of all user sample data;

[0126] When the frequency of the numerical feature corresponding to a sample data in the data set is less than 0.05, the sample data is removed;

[0127] For discrete features, one-hot encoding is used for processing, and (1,0,0), (0,1,0), and (0,0,1) are used to represent first-tier cities, second-tier cities, and third-tier cities respectively;

[0128] After processing different types of data, merge them into a new data set, and then use the K-Means clustering method to remove outliers.

[0129] Compared with the prior art, the present invention automatically collects user data from each service provider, divides the user data into user groups, and establishes a sub-communication service evaluation model corresponding to the user group. By optimizing the model parameters of the global communication service evaluation model using the model parameters of each sub-communication service evaluation model, a global communication service evaluation model with more comprehensive user data can be obtained, and thus a more accurate communication service evaluation result can be obtained, which effectively overcomes the problem of incomplete data and insufficient data processing of a single communication service provider; according to the communication service evaluation results determined by the global communication service evaluation model, customized communication services that are more in line with user characteristics can be provided for various user groups more effectively.

[0130] Figure 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

[0131] like Figure 4 As shown, the electronic device is presented in the form of a general computing device. The processor may be one or more and work in coordination. The present invention does not exclude distributed processing, that is, the processor may be dispersed in different physical devices. The electronic device of the present invention is not limited to a single entity, but may also be the sum of multiple physical devices.

[0132] The memory stores a computer executable program, which is usually a machine-readable code. The computer-readable program can be executed by the processor to enable the electronic device to perform the method of the present invention, or at least part of the steps in the method.

[0133] The memory includes a volatile memory, such as a random access memory unit (RAM) and / or a cache memory unit, and may also be a non-volatile memory, such as a read-only memory unit (ROM).

[0134] Optionally, in this embodiment, the electronic device further includes an I / O interface, which is used for the electronic device to exchange data with an external device. The I / O interface can represent one or more of several types of bus structures, including a storage unit bus or a storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of a variety of bus structures.

[0135] It should be understood that Figure 4The electronic device shown is only an example of the present invention, and the electronic device of the present invention may also include elements or components not shown in the above examples. For example, some electronic devices also include display units such as display screens, and some electronic devices also include human-computer interaction elements such as buttons, keyboards, etc. As long as the electronic device can execute the computer-readable program in the memory to implement the method of the present invention or at least part of the steps of the method, it can be considered as an electronic device covered by the present invention.

[0136] Through the above description of the implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by combining software with necessary hardware. Figure 5 As shown, the technical solution according to the embodiment of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of commands to enable a computing device (which can be a personal computer, a server, or a network device, etc.) to execute the above method according to the embodiment of the present invention.

[0137] The software product may be any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0138] The computer readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with a command execution system, device, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0139] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0140] The computer-readable medium carries one or more programs (eg, computer-executable programs). When the one or more programs are executed by a device, the computer-readable medium implements the method of the present disclosure.

[0141] Those skilled in the art will appreciate that the above modules can be distributed in the device according to the description of the embodiment, or can be changed accordingly and only used in one or more devices different from the embodiment. The modules of the above embodiments can be combined into one module, or further divided into multiple sub-modules.

[0142] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the exemplary embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several commands to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the embodiment of the present invention.

[0143] The exemplary embodiments of the present invention are specifically shown and described above. It should be understood that the present invention is not limited to the detailed structure, configuration or implementation method described herein; on the contrary, the present invention is intended to cover various modifications and equivalent configurations included in the spirit and scope of the appended claims.

Claims

1. A communication service evaluation method, characterized in that: include: Automatically collect user data from each service provider, encrypt the user data, and upload the initial model parameters of each service provider; Each service provider trains the sub-communication service evaluation model based on each local user data and initial model parameters, specifically including establishing each local training data for training the corresponding sub-communication service evaluation model based on the following feature data: user number, communication service time, cost, call quality, network coverage, and evaluation data corresponding to each feature data; group training corresponding to the user group, and obtaining updated model parameters after each training round; Federated learning is performed based on the model parameters uploaded by each service provider to obtain an optimized global communication service evaluation model; Input the user data to be processed into the global communication service evaluation model to obtain the communication service evaluation result; wherein, A neural network algorithm is used to establish a sub-communication service evaluation model corresponding to each user group. The sub-communication service evaluation model is expressed by the following expression: Y= [A1 +A2 +...+An ], wherein Y represents the evaluation score output by the sub-communication service evaluation model; , ,..., are multiple feature data selected from user data, including service time, cost, call quality, network coverage, network access time, and place of residence information. A1, A2, ..., An represent the weight parameters corresponding to the first feature data, the second feature data, and the nth feature data, respectively, representing the importance of different feature data in the group. The larger the An, the more important it is. It will be continuously updated during the model training process. The sum of each item is multiplied by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n represents the number of feature data, and n is a positive integer; By adjusting and calculating characteristic parameters corresponding to different user groups, the sub-communication service evaluation model is trained for multiple rounds so that the difference between the model calculation score and the actual score is minimized, so as to determine to stop training the sub-communication service evaluation model and obtain a trained sub-communication service evaluation model; The following expression is used to represent the global communication service evaluation model: ； in, represents the model calculation score of the global communication service evaluation model, t represents the t-th user group, and t is a positive integer; c1, c2, ... c m They represent the number of data sets of m sub-communication service evaluation models when they are uploaded to the central server in each round to train the global communication service evaluation model. They are used to adjust the weight ratios of sub-communication service evaluation models of different sizes. Each sub-communication service evaluation model corresponds to a different service provider. m is a positive integer. Specifically, c1 represents the number of data sets of the first sub-communication service evaluation model uploaded to the central server for the first round of training. c2 represents the number of data sets of the second sub-communication service evaluation model uploaded to the central server for the first round of training. m represents the number of data sets uploaded to the central server for the first round of training of the mth sub-communication service evaluation model; , ,..., They represent the first feature data, the second feature data, and the nth feature data, respectively, where n is a positive integer; a1 represents the weight parameter for the first feature data X1 under the global communication service evaluation model, a2 represents the weight parameter for the second feature data X2 under the global communication service evaluation model, and a m Represents the nth feature data X under the global communication service evaluation model n The weight parameter is to add each item and multiply it by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n represents the number of feature data; The updated global communication service evaluation model then sends the relevant parameters to each sub-communication service evaluation model for training again, and uploads the parameters obtained from this training after the model training is completed, repeating the above steps until the final global communication service evaluation model is obtained, and the number of parameter uploads between each sub-communication service evaluation model and the global communication service evaluation model is greater than or equal to 50 times, or the number of parameter sends between each sub-communication service evaluation model and the global communication service evaluation model is greater than or equal to 50 times; For the training of each sub-communication service evaluation model, in order to prevent the model from overfitting, the loss error is not higher than 0.05 and the number of training rounds is not higher than 2000 echos.

2. The communication service evaluation method according to claim 1, characterized in that: include: According to the different importance of different communication service dimensions for users, the users are grouped according to the specified ratio to obtain multiple user groups; According to different user groups, multiple sub-communication service evaluation models are established, and group training corresponding to the user groups is carried out, and the model parameters are updated once the training is carried out; among them, based on the following characteristic data formed by each communication service of each user group, a training data set is established for training the respective sub-communication service evaluation models: service time, cost, call quality, network coverage, network access time, and residence information.

3. The communication service evaluation method according to claim 2, characterized in that: Further including: Before grouping users according to the specified ratio, an initial classification is performed based on the user's gender, annual income, network connection length, and place of residence information to obtain the first user group, the second user group, and the third user group corresponding to high consumption, medium consumption, and low consumption.

4. The communication service evaluation method according to claim 1, characterized in that: include: The difference between the model-calculated score of the sub-communication service evaluation model and the real score given by the user is expressed as follows: = Where E represents the difference between the model calculation score Y of the sub-communication service evaluation model and the real score given by the user , Y represents the model calculation score of the sub-communication service evaluation model, that is, the score predicted by the model, Represents the real rating given by the user; , ,..., They represent the first feature data, the second feature data, and the nth feature data respectively. A1, A2, ..., An represent the weight parameters corresponding to the first feature data, the second feature data, and the nth feature data respectively, representing the importance of different feature data in the group. The larger the An, the more important it is. It will be continuously updated during the model training process. The sub-communication service evaluation model is trained in each group. By collecting multiple groups of user data, each group of data includes feature data. , ,..., , add up the items and multiply by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n is a positive integer, n represents the number of feature data, through the model training of the sub-communication service evaluation model, the value of the loss function E is close to 0, that is, the score Y predicted by the model is different from the actual score given by the user When E<0.05, the model converges and a set of trained weight parameters is obtained. , training completed.

5. The communication service evaluation method according to claim 1, characterized in that: include: For each service provider, based on the final global communication service evaluation model and cost model, we use linear programming to solve the optimal solution to seek a balance between high service score and low cost. We set the cost of feature data of each dimension to be in the interval [1,10], where 1 represents the lowest cost and 10 represents the highest cost. The total cost that each service provider can bear is denoted as u. Let C = u / n, which represents the arithmetic mean of the cost of each feature, so that it is consistent with Y t Aligned in terms of dimensions, the objective function is F = Y t -C, by solving the following system of equations to maximize F: ； Wherein, F represents the global communication service performance index under the t-th user group. The larger the global communication service performance index F is, the higher the user satisfaction is and the lower the cost is. t is a positive integer; Y t The model calculation score of the global communication service evaluation model is used to represent user satisfaction, t represents the t-th user group, c1, c2, ... c m They represent the number of data sets of m sub-communication service evaluation models when they are uploaded to the central server in each round to train the global communication service evaluation model. They are used to adjust the weight ratios of sub-communication service evaluation models of different sizes. Each sub-communication service evaluation model corresponds to a different service provider. m is a positive integer. Specifically, c1 represents the number of data sets of the first sub-communication service evaluation model uploaded to the central server for the first round of training. c2 represents the number of data sets of the second sub-communication service evaluation model uploaded to the central server for the first round of training. m represents the number of data sets uploaded to the central server for the first round of training of the mth sub-communication service evaluation model; , ,..., They represent the first feature data, the second feature data, and the nth feature data, respectively, where n is a positive integer; a1 represents the weight parameter for the first feature data X1 under the global communication service evaluation model, a2 represents the weight parameter for the second feature data X2 under the global communication service evaluation model, and a m Represents the nth feature data X under the global communication service evaluation model n The weight parameter is to add each item and multiply it by , used to obtain the arithmetic mean of the evaluation scores of each feature data, n represents the number of feature data; C represents the cost model of the service provider corresponding to the t-th user group, and u represents the total cost that each service provider can bear, that is, 0≤ + +...+ ≤u, i is a positive integer, specifically 1, 2, 3, ..., n, n represents the number of feature data, u i Indicates the cost value of the feature data of the first latitude, 1≤ ≤10, Represents the cost value of the feature data of the second dimension, 1≤ ≤10, Represents the cost value of the feature data at the nth latitude, 1≤ ≤10; Finally, we get a set of optimal solutions , ,..., , is used to represent the cost coefficient of the t-th user group.

6. The communication service evaluation method according to claim 2, characterized in that: Further including: Remove outliers, including: Use the following expression to normalize the numerical features and remove the sample data that is not within (μ-3σ, μ+3σ): ; in, It represents the feature data of the i-th feature data in the sample data of the same user after being normalized by numerical features; represents the i-th feature data value in the sample data of the same user without numerical feature standardization, i is a positive integer; μ represents the mean of the i-th feature data of all user sample data; σ represents the variance of the i-th feature data of all user sample data; When the frequency of the numerical feature corresponding to a sample data in the data set is less than 0.05, the sample data is removed; For discrete features, one-hot encoding is used for processing, and (1,0,0), (0,1,0), and (0,0,1) are used to represent first-tier cities, second-tier cities, and third-tier cities respectively; After processing different types of data, merge them into a new data set, and then use the K-Means clustering method to remove outliers.

7. A communication service evaluation system, characterized in that: According to any one of claims 1 to 6, the communication service evaluation system comprises: The data collection module is used to automatically collect user data from each service provider, encrypt the user data, and upload the initial model parameters of each service provider; Sub-model training module, each service provider performs sub-communication service evaluation model training based on each local user data and initial model parameters, specifically including establishing each local training data for training the corresponding sub-communication service evaluation model based on the following feature data: user number, communication service time, cost, call quality, network coverage, and evaluation data corresponding to each feature data; group training corresponding to the user group, and obtaining updated model parameters each time training; The federated learning module performs federated learning based on the model parameters uploaded by each service provider to obtain an optimized global communication service evaluation model; The prediction module inputs the user data to be processed into the global communication service evaluation model to obtain the communication service evaluation result.

8. The communication service evaluation system according to claim 7, characterized in that: include: According to the different importance of different communication service dimensions for users, the users are grouped according to the specified ratio to obtain multiple user groups; According to different user groups, multiple sub-communication service evaluation models are established, and group training corresponding to the user groups is performed, and the model parameters are updated once the training is performed; wherein, based on the following characteristic data formed by each communication service of each user group, a training data set is established to train the respective sub-communication service evaluation models: service time, cost, call quality, network coverage, network access time, and residence information; Before grouping users according to the specified ratio, an initial classification is performed based on the user's gender, annual income, network connection length, and place of residence information to obtain the first user group, the second user group, and the third user group corresponding to high consumption, medium consumption, and low consumption.

Images