Home customer identification methods, devices, electronic devices, media, and products based on federated models

Abstract

This application discloses a method, apparatus, electronic device, medium, and product for identifying household customers based on a federated model, relating to the field of federated learning technology. The method for identifying household customers based on a federated model includes: collecting raw customer data, wherein the raw customer data includes at least one of the following: operator data, social data, e-commerce data, vehicle usage data, location data, or smart home data; inputting the raw customer data into a preset household group identification federated model to obtain multiple sets of household group data; inputting each household group data into a preset high-value customer identification federated model to obtain high-value customer identification results, wherein the high-value customer identification results include high-value tags corresponding to the household group data. The technical solution of this application solves the problems of low accuracy and data security risks in current identification schemes for high-value household customers.

Description

Technical Field

[0001] This application relates to the field of federated learning technology, and in particular to methods, devices, electronic devices, computer-readable storage media, and computer program products for identifying home customers based on federated models. Background Technology

[0002] As competition intensifies in the personal market, more and more companies are starting to focus on the family market, which means designing products or services with the family as the customer unit in mind and launching a series of family products.

[0003] When making marketing recommendations to households, it's essential to first uncover their value. However, this process involves analyzing diverse customer behavior data. Currently, individual companies have limited data on household customer behavior, and the data dimensions are often too narrow. Using only internal data from a single company for household customer identification fails to comprehensively and accurately identify high-value households. Conversely, aggregating and centrally analyzing internal data from multiple companies can lead to sensitive data leaks and compromise data security.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this application is to provide a method, device, electronic device, computer-readable storage medium, and computer program product for identifying household customers based on a federated model, aiming to solve the technical problems of low accuracy and data security risks in current identification schemes for high-value household customers.

[0006] To achieve the above objectives, this application proposes a household customer identification method based on a federated model, applicable to participating parties. The household customer identification method based on a federated model includes:

[0007] Collect raw customer data, wherein the raw customer data includes at least one of the following: operator data, social data, e-commerce data, vehicle data, location data, or smart home data;

[0008] The original customer data is input into a preset family group identification federated model to obtain multiple sets of family group data. The family group identification federated model is obtained by multiple participants using local customer data for federated learning.

[0009] The data of each family group is input into a preset high-value customer identification federated model to obtain high-value customer identification results. The high-value customer identification federated model is obtained by multiple participants using local customer data for federated learning. The high-value customer identification results include high-value tags corresponding to the family group data.

[0010] In one embodiment, after the step of inputting the data of each of the said family groups into a preset high-value customer identification federated model to obtain high-value customer identification results, the method further includes:

[0011] Based on the high-value customer identification results, evaluate the model quality of the high-value customer identification federated model;

[0012] If the quality of the model does not meet the preset requirements, the high-value customer identification federated model is optimized based on the original customer data to obtain an updated high-value customer identification federated model.

[0013] In one embodiment, after the step of inputting the data of each of the said family groups into a preset high-value customer identification federated model to obtain high-value customer identification results, the method further includes:

[0014] The family group data with high-value labels are input into a preset high-value customer classification federated model to determine the family type corresponding to each family group data. The high-value customer classification federated model is obtained by multiple participants using local customer data for federated learning.

[0015] Recommend the corresponding family type products to the family customers of each of the aforementioned family group data.

[0016] In one embodiment, before the step of inputting the original customer data into a preset family group identification federated model to obtain multiple sets of family group data, the method further includes:

[0017] Collect existing customer data, align the existing customer data with the existing customer data of other participants to obtain training customer data, wherein the existing customer data includes at least customer attribute data, social data, location data and household business data.

[0018] The training customer data is subjected to binning statistical analysis to calculate the information value and evidence weight of various features in the training customer data.

[0019] Based on the information value and evidence weight of various features, multiple feature data are selected from the training customer data;

[0020] The federated model is trained based on the feature data of each participant and the feature data of other participants to obtain the family group identification federated model.

[0021] In one embodiment, the step of training a federated model based on the feature data of each of the aforementioned feature data and the feature data corresponding to other participants to obtain a family group identification federated model includes:

[0022] The corresponding local model is obtained by training based on the local feature data.

[0023] The gradient corresponding to the local model is sent to the coordinator, so that the coordinator can calculate the weighted average of the gradients received from each participant to obtain the updated gradient.

[0024] Receive the update gradient sent by the coordinator, update the local model according to the update gradient, and obtain the updated local model;

[0025] After iteratively training the updated local model based on local feature data, the execution steps are returned: the gradient corresponding to the local model is sent to the coordinator until a preset stopping condition is reached, thus obtaining the family group identification federated model.

[0026] In one embodiment, the step of inputting the data of each of the said family groups into a preset high-value customer identification federated model includes:

[0027] The feature data in each of the family group data are sequentially subjected to feature scaling, data cleaning and normalization to obtain the target feature data;

[0028] The feature data of various feature types in the target feature data are weighted and calculated to obtain deep feature data, wherein the feature types include at least one of family structure features, online consumption features, offline consumption features, car use behavior features, residential behavior features, and home furnishing consumption features;

[0029] The deep feature data and the target feature data are fused by a preset gating signal to obtain multi-feature fused data, wherein the gating signal includes at least an activation function and a fully connected layer;

[0030] The multi-feature fusion data is input into the high-value customer identification federated model to predict the corresponding high-value customer identification result.

[0031] Furthermore, to achieve the above objectives, this application also proposes a home customer identification device based on a federated model, the home customer identification device based on a federated model comprising:

[0032] The data acquisition module is used to collect raw customer data, wherein the raw customer data includes at least one of the following: operator data, social data, e-commerce data, vehicle data, location data, or smart home data.

[0033] The family group identification module is used to input the original customer data into a preset family group identification federated model to obtain multiple sets of family group data. The family group identification federated model is obtained by multiple participants using local customer data for federated learning.

[0034] The high-value identification module is used to input the data of each family group into a preset high-value customer identification federated model to obtain high-value customer identification results. The high-value customer identification federated model is obtained by multiple participants using local customer data for federated learning. The high-value customer identification results include high-value tags corresponding to the family group data.

[0035] In addition, to achieve the above objectives, this application also proposes an electronic device, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the federated model-based home customer identification method as described above.

[0036] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the federated model-based home customer identification method described above.

[0037] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the federated model-based home customer identification method described above.

[0038] This application proposes a method for identifying household customers based on a federated model. First, raw customer data is collected, including at least one of the following: operator data, social data, e-commerce data, vehicle usage data, location data, or smart home data. Then, the raw customer data is input into a preset household group identification federated model to obtain multiple sets of household group data. This household group identification federated model is obtained by multiple participants using local customer data through federated learning. Finally, each household group data set is input into a preset high-value customer identification federated model to obtain high-value customer identification results. These results include high-value tags corresponding to the household group data. The high-value customer identification federated model is obtained by multiple participants using local customer data through federated learning. This technical solution enables sequential household group identification and high-value customer identification from the collected raw customer data without data leaving the domain. Furthermore, since both the household group identification federated model and the high-value customer identification federated model are obtained by multiple participants using local customer data through federated learning, the data dimensions of the models are enriched. This overcomes the low accuracy of models trained with internal data from a single participant or enterprise, achieving accurate identification of high-value household customers. In addition, the technical solution of this application also utilizes federated learning technology. The training and application of the federated model are both implemented locally by the participating parties, which ensures the privacy and security of customer data and does not pose a risk of sensitive data leakage. Attached Figure Description

[0039] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating an embodiment of the household customer identification method based on the federated model in this application.

[0042] Figure 2 This is a schematic diagram of a feasible high-value household customer identification system in the embodiments of this application;

[0043] Figure 3 This is a schematic diagram of a feasible high-value household customer identification scheme based on federated learning in the embodiments of this application;

[0044] Figure 4This is a schematic diagram illustrating a feasible high-value family group segmentation scheme based on federated clustering algorithm in an embodiment of this application.

[0045] Figure 5 This is a schematic diagram illustrating a feasible process for training a federated model for family group identification in an embodiment of this application;

[0046] Figure 6 This is a schematic diagram illustrating the process of feature data extraction and fusion for multiple feature types in an embodiment of this application;

[0047] Figure 7 This is a schematic diagram of the structural composition of the home customer identification device based on the federated model in the embodiments of this application;

[0048] Figure 8 This is a schematic diagram of the device structure of the hardware operating environment involved in the home customer identification method based on the federated model in this application embodiment.

[0049] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0050] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0051] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0052] The executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, mobile phone, server, etc., or an electronic device or control device capable of performing the above functions. The following description uses a server as the executing entity to illustrate this embodiment and the subsequent embodiments.

[0053] As competition intensifies in the personal market, more and more companies are focusing on the family market, designing products and services with families as the primary customer unit. This has led to the launch of various family-oriented products, such as family packages and supplementary family SIM cards offered by telecommunications operators, and family ticket packages offered by the entertainment industry. Therefore, identifying the value of family customers is crucial for businesses. Thus, identifying high-value family customers and targeting them with focused marketing is essential. Specifically, family customer value mining typically involves various aspects of customer behavior data. However, currently, individual companies have limited data on family customer behavior, and the data is often too narrowly defined. Using only internal data from a single company cannot comprehensively and accurately identify high-value family customers. Centralized analysis of data from multiple companies poses a risk of sensitive data leakage and fails to meet data security requirements. Furthermore, existing technical solutions often focus on analyzing individual customers, lacking a holistic analysis of the entire family, treating all family members as a unified entity.

[0054] To address the aforementioned issues, this application proposes a family customer identification method based on a federated model. This method enables the fusion and application of family customer-related data from multiple enterprises without requiring customer data to leave the domain. Furthermore, data storage and model building are completed locally by each participating party, with interactions between the participating parties and the coordinator limited to model training and updates. This approach not only protects customer privacy and security but also enriches the data dimensions, thereby achieving accurate identification of family customer value while ensuring customer data security.

[0055] This application provides a method for identifying family customers based on a federated model, applied to participating parties, with reference to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the household customer identification method based on the federated model of this application. The household customer identification method based on the federated model includes:

[0056] Step S10: Collect raw customer data, wherein the raw customer data includes at least one of the following: operator data, social data, e-commerce data, vehicle data, location data, or smart home data;

[0057] Among them, the participating parties refer to the enterprises that provide the original customer data. These enterprises can be telecommunications operators (corresponding to operator data and location data), social application providers (corresponding to social data), e-commerce application providers (corresponding to e-commerce data), car manufacturers (for vehicle data), and smart home providers (such as smart TVs, smart screens, etc.) (corresponding to smart home data).

[0058] For example, since the value of family customers involves multi-dimensional customer behavior, such as shopping, travel, and residential travel, the specific data sources and dimensions include, but are not limited to, the following: operator data includes basic information, communication consumption, location, call volume, DPI (Deep Packet Inspection) internet access (application / website category), family service subscriptions, etc.; e-commerce data includes order frequency, amount, and product categories from e-commerce platforms such as JD.com and Taobao; vehicle data includes customer vehicle model and price range; payment behavior in e-commerce data includes payment amount, merchant type, and frequency; travel behavior data; and overall family behavior (including but not limited to smart home usage behavior and large-screen viewing behavior). All of these raw customer data from various dimensions reflect the consumption capabilities and preferences of multiple family customers, and can be used for family group identification and high-value identification.

[0059] It should be noted that all the above original customer data is stored locally on the participating parties' premises to ensure the security of customers' privacy data.

[0060] Step S20: Input the original customer data into the preset family group identification federated model to obtain multiple sets of family group data. The family group identification federated model is obtained by multiple participants using local customer data for federated learning.

[0061] After collecting the original customer data, data that can represent the attributes of family groups can be selected from the original customer data and input into the pre-trained family group identification federated model to distinguish the original customer data by family unit, resulting in multiple sets of family group data, with each set of family group data corresponding to one family.

[0062] Specifically, the following data can be extracted from the original customer data as the basis for family group identification:

[0063] Basic attributes: Customer's basic information data such as age and gender, which can come from customer information data accumulated by the participating parties. Basic attributes = {age, gender, ...} etc.

[0064] Social Intensity: This quantifies the number of social interactions during weekday working hours, weekday commuting hours, weekday nighttime hours, and holiday daytime and holiday nighttime hours. It can be achieved through comprehensive analysis using call-based social data from telecommunications operators, and message and voice data from social applications. Social Intensity = Σ t∈T Social interaction frequency (t), where T is a set of time periods, including weekday work hours, weekday commuting hours, etc.

[0065] Location stability point similarity: This is calculated by comparing the distance similarity between stable points during weekday daytime and nighttime periods, as well as during holiday daytime and nighttime periods. Data can be sourced from location information provided by telecommunications operators or travel service applications. The formula for calculating location stability point similarity L is as follows:

[0066]

[0067] Among them, W A and J A For customer A, the latitude and longitude of the nighttime stable point, W B J B Here are the latitude and longitude coordinates of the nighttime stable point for customer B, and R is the Earth's radius.

[0068] Family service subscription or usage behavior: This includes information on primary and secondary SIM cards, family members, broadband connection devices, and smart home connection devices. Data is sourced from telecommunications operators, smart home applications, or other companies with family service subscription and member information. In other words, family service behavior = {primary and secondary SIM card information, family members, broadband connection devices, smart home devices, ...}.

[0069] It should also be noted that the family group identification federated model is pre-trained using locally sourced customer data from multiple participating parties through federated learning. During the training process, the raw data from each participating party does not leave its domain and does not interact with data from other participating parties. Data interaction is limited to the central server of the coordinating party, which handles model parameters. This ensures that user privacy data is not leaked and protects data security.

[0070] Step S30: Input the data of each family group into the preset high-value customer identification federated model to obtain the high-value customer identification result. The high-value customer identification federated model is obtained by multiple participants using local customer data for federated learning. The high-value customer identification result includes the high-value label corresponding to the family group data.

[0071] After determining the data of each family group, it is equivalent to distinguishing the data of each family. The next step is to identify the high-value family customers in these families. In this embodiment of the application, the family group data with the high-value tag refers to the high-value family customers.

[0072] Similarly, before inputting the data from each household group into the pre-defined high-value customer identification federated model, feature data closely related to high-value household customers can be filtered from the household group data. For example, when extracting feature data from household group data, the following features may be included:

[0073] Family structure characteristics: These refer to whether a family consists of one person, two people, whether they have children, or three generations living together. Based on data such as the types of applications accessed by family group members (from carriers, such as parenting apps), the types of goods purchased (from e-commerce platforms, such as parenting and K-12 online education), and the age and gender of family members (from basic customer information of each participating party), the expression for family structure characteristics is as follows:

[0074]

[0075] Among them, f 一人 f 二人 f 有孩子 The condition can be a binary flag representing whether these conditions are met; f app母婴 f appK12 It can be a continuous variable, representing the frequency or duration of corresponding app usage; f 年龄 f 性别 These can be categorical variables, representing the age and gender of family members, respectively.

[0076] Online consumption characteristics include: shopping frequency, amount, and product categories on e-commerce platforms; order frequency and amount on travel apps; and other consumption-related applications. Data from multiple family members needs to be weighted and accumulated to obtain the characteristic values ​​of the family group. Weights can be assigned based on family roles, which can be determined based on the family structure and age information mentioned above. Additionally, business-related consumption, such as large-volume purchases of goods or frequent travel, needs to be excluded.

[0077]

[0078] All of the above features can be continuous variables, reflecting the corresponding consumer behavior.

[0079] Offline consumption characteristics: This refers to offline consumption and online payment behavior, such as the frequency of use of payment apps, the amount spent, and the type of merchant. Data from multiple family members needs to be weighted and accumulated; the specific processing method is the same as above.

[0080]

[0081] The above characteristics are similar to those of online consumption; all of these characteristics can be continuous variables.

[0082] Vehicle usage behavior characteristics: Data such as the models, brands, and price ranges owned by car manufacturers.

[0083]

[0084] Residential Characteristics: Based on operator location data, identify stable points at night. Obtain information about the residential community where the family resides based on map positioning data. Use web crawling to obtain the corresponding housing prices for the community, and derive residential behavior data based on community housing price information and the location (core area, suburbs). Additionally, it is necessary to distinguish between owner-occupied and rented residences, and remove the influence of special data such as tenants; this can be identified through the stable point change cycle and frequency.

[0085]

[0086] Other potential consumption characteristics include subscription and viewing behavior for home large-screen services, such as the type of goods ordered, the amount spent, and the duration of viewing.

[0087]

[0088] The above-mentioned characteristics can reflect the spending power and preferences of family users, making it easier to identify high-value family customers.

[0089] The high-value customer identification federated model is pre-trained using locally sourced customer data from multiple participating parties through federated learning. During the training process, the raw data from each participating party remains within its domain and does not interact with other participating parties. Data interaction is limited to the central server of the coordinating party, which handles model parameters. This ensures that user privacy data is not leaked and protects data security.

[0090] After obtaining the high-value customer identification results, the high-value family customers corresponding to the family group data with high-value tags can be determined. The participating parties (enterprises) can then develop targeted marketing strategies for the identified high-value family customers to improve the conversion rate. The technical solution of this application embodiment can not only meet the data security management requirements, but also achieve accurate identification of high-value family customers, and can provide a reference for subsequent precise policy implementation for high-value family customers.

[0091] For ease of understanding, the household customer identification method of the federated model provided in this application embodiment can be further extended to a high-value household customer identification system, such as... Figure 2 As shown, the entire system includes a data acquisition and preprocessing module, a high-value household customer identification model construction module, and a high-value customer analysis and application module.

[0092] In the data collection module, each participant collects data related to the behavior of household customers, such as basic information, location data, social data lake, application usage or internet access data, e-commerce shopping data, and other data such as travel behavior. After preprocessing, the data is stored locally on each party's premises.

[0093] The high-value household customer identification model construction module includes a household group identification module and a high-value household customer identification module. The household group identification module uses federated learning technology to identify household groups based on multi-party social data, location data, and household service usage behavior data. The high-value household customer identification module (including a household group feature extraction module and a federated learning-based high-value household customer identification module) constructs household group features based on multi-party household group customer data, then performs federated learning modeling to identify high-value household customers.

[0094] In another feasible embodiment, the high-value household customer identification system may also include a high-value household customer analysis and application module based on federated learning. This module is mainly used to perform cluster analysis on high-value household customers based on federated clustering algorithms, thereby identifying segmented groups and conducting precise marketing and refined services for household products to different high-value groups.

[0095] Furthermore, in a feasible embodiment, after the step of inputting the data of each of the said family groups into a preset high-value customer identification federated model to obtain the high-value customer identification result, the method may further include:

[0096] Step S40: Based on the high-value customer identification results, evaluate the model quality of the high-value customer identification federated model;

[0097] In step S50, if the model quality does not meet the preset requirements, the high-value customer identification federated model is optimized based on the original customer data to obtain an updated high-value customer identification federated model.

[0098] After obtaining the high-value customer identification results, the family group data with high-value tags in these high-value customer identification results can be screened based on the preset high-value customer rules in order to evaluate the model quality of the high-value customer identification federated model applied in this case.

[0099] For example, participants can pre-set certain requirements for high-value customer family group data, such as the frequency and amount of online and offline consumption needing to reach preset thresholds (i.e., preset requirements). If the frequency or amount of online and offline consumption in the currently identified family group data with high-value tags does not meet the preset thresholds, it indicates that the accuracy of the high-value customer identification federated model is insufficient, and further optimization and updates are needed. The optimization and update process is similar to the training process of the high-value customer identification federated model, requiring federated learning by combining the original customer data from multiple participants to obtain a more accurate high-value customer identification federated model.

[0100] In addition, after obtaining the updated high-value customer federated identification model, return to step S30 to obtain the new high-value customer identification result, and then execute steps S40 to S50 until the model quality meets the preset requirements.

[0101] In this embodiment, the model quality of the high-value customer identification federated model is effectively evaluated based on the high-value customer identification results, ensuring the accuracy of the output high-value customer identification results and improving the accuracy of identifying high-value household customers.

[0102] For example, in conjunction with the content of the foregoing embodiments, the process of the high-value household customer identification scheme based on federated learning in steps S10 to S50 can be as follows: Figure 3 As shown, the process first collects data from telecommunications operators, social media apps, e-commerce platforms, car manufacturers, and other sources. Then, it performs family group identification, specifically including feature extraction of the customers to be identified. Next, it performs family group identification based on federated learning to obtain multiple sets of family group data. Then, it performs high-value family customer identification, performing family group feature extraction and pre-training a high-value family customer identification model based on federated learning to obtain a federated learning model. Finally, it performs prediction based on the federated learning model to determine the high-value customer group (i.e., the high-value customer identification result). The model's performance (quality) is evaluated to see if it meets the preset requirements. If it does, the process ends; otherwise, it returns to performing family group feature extraction and retraining the model.

[0103] In one feasible embodiment, after the step of inputting the data of each of the said family groups into a preset high-value customer identification federated model to obtain the high-value customer identification results, the method may further include:

[0104] Step S60: Input the family group data with high-value labels into the preset high-value customer classification federated model to determine the family type corresponding to each family group data. The high-value customer classification federated model is obtained by multiple participants using local customer data for federated learning.

[0105] Step S70: Recommend the corresponding family type products to the family customers in each family group data.

[0106] After obtaining the high-value customer identification results and determining the high-value family customer group, in order to further segment the high-value customer group, facilitate the implementation of more refined and targeted marketing strategies, and improve marketing efficiency and customer experience, a pre-trained high-value customer classification federated model can be used to classify the families corresponding to each group of family data.

[0107] For example, in the embodiments of this application, a federated model for classifying high-value customers can be trained based on a federated learning clustering algorithm. Taking K-means clustering as an example, a feasible high-value family group segmentation scheme based on the federated clustering algorithm is as follows: Figure 4 As shown. During model training, given a household, the household customers need to be divided into k clusters. Each cluster C... j Corresponding to a center point μ j Then, given the feature vector X′ of household customers, select k feature vectors of household customers as the initial cluster centers: (μ1, μ2, ..., μ3). k ).

[0108] The iterative training process of the model includes an allocation phase: for each household customer feature vector X i ′, calculate its distance from each center point μ j The value is determined by the distance between the elements, and the element is assigned to the nearest cluster. The expression is as follows:

[0109]

[0110] Update phase: Recalculate the centroids of each cluster.

[0111] Termination condition: The iteration process will continue until the center point μ. j When the change is less than a predetermined threshold, or when the preset maximum number of iterations is reached, the training process of the federated learning model is completed, and a high-value customer classification federated model is obtained.

[0112] In the process of classifying family group data using the high-value customer classification federated model, family group data with high-value labels are input into the high-value customer classification federated model. The high-value customer classification federated model determines the data to which each family group is assigned, analyzes the data points in each cluster, and extracts the statistical characteristics of the cluster, such as the mean and median, to distinguish different customer groups and thus determine which family type the family corresponding to each family group data belongs to.

[0113] In this embodiment of the application, for each cluster C j Design different marketing strategies to improve marketing efficiency and customer experience. (Marketing Strategy) j With clustering C j The mapping relationship can be represented as:

[0114] Strategy j =DesignStrategy(C j );

[0115] Here, DesignStrategy refers to the design strategy function. Examples of segmented customer groups and marketing strategies are as follows: family types include, but are not limited to: families interested in education, recommending educational products to their family members; families interested in travel, recommending travel products to their family members; and families interested in mothers and babies, recommending mother and baby products to their family members. In practice, the development of marketing strategies also requires consideration of other influencing factors such as the passage of time and cyclicality. It is also necessary to evaluate the effectiveness of different marketing strategies and optimize them based on the results, making the marketing strategies more aligned with the characteristics and needs of each segmented group. The mathematical expression can be represented as:

[0116] OptimizedStrategy j =Optimize(Strategy) j EvaluationMetrics);

[0117] Among them, Strategy j OptimizedStrategy represents the marketing strategy for the j-th cluster. j This represents the optimized marketing strategy, while EvaluationMetrics are the evaluation metrics for assessing the marketing strategy.

[0118] This application essentially provides a method for segmenting and applying high-value family groups based on federated learning. Specifically, it uses a federated clustering algorithm to perform cluster analysis on high-value family customers, classifies the family types of the segmented family customers, and implements precise policies for high-value segmented family customers of different family types.

[0119] Additionally, in one feasible embodiment, before the step of inputting the original customer data into a preset family group identification federated model to obtain multiple sets of family group data, the method may further include:

[0120] Step A10: Collect existing customer data and align the existing customer data with the existing customer data of other participants to obtain training customer data. The existing customer data includes at least customer attribute data, social data, location data, and household business data.

[0121] Step A20: Perform binning statistical analysis on the training customer data to calculate the information value and evidence weight of various features in the training customer data;

[0122] Step A30: Based on the information value and evidence weight of various features, select multiple feature data from the training customer data;

[0123] Step A40: Train the federated model based on the feature data of each feature data and the feature data of other participants to obtain the family group identification federated model.

[0124] This application provides a method for collaboratively training a family group identification model based on federated learning technology. Specifically, the family group identification federated model is a federated classification model, and the existing customer data is the customer data stored by each participating party in this instance.

[0125] Reference Figure 5 Taking the participation of telecom operators and e-commerce companies as an example, the existing customer data of telecom operators can include multiple features such as basic customer information, social intensity information, location information, and family service subscription / usage information. The existing customer data of e-commerce companies can include online shopping behavior features such as user device information, purchasing behavior, payment method, order information, and delivery address. The participating parties load and import their respective feature data into their local machines to complete the data preparation work.

[0126] The above feature data is represented as: F 运营商 ={Basic Information, Social Intensity, Location Information, Home Service Subscription / Usage}, F 电商 ={User device, purchase behavior, payment method, order information, shipping address}.

[0127] Since the users of telecom operators and e-commerce platforms do not completely overlap, shared user mobile phone numbers are selected as sample IDs (Identity Documents). Then, the sample IDs, encrypted using RSA (asymmetric encryption) and hash algorithms, are used to align the samples between telecom operators and e-commerce platforms. This process confirms shared users without disclosing their respective data. This process ensures that the difference between the data sets of the two parties remains unknown to each other.

[0128] Define the RSA / Hash encryption function as: E(x) = RSA / Hash(x), then the shared user mobile phone number is encrypted as: ID 加密 =E(ID) 共有 ).

[0129] Furthermore, operators and e-commerce companies need to conduct binned statistical analysis on features such as location characteristics, social characteristics, and broadband service subscription and usage behavior to calculate the IV (Information Value) / WOE (Weight of Evidence) value of the features, and filter out and lock valid features with IV and WOE values ​​higher than preset thresholds to participate in the subsequent model construction process.

[0130]

[0131]

[0132] Wherein, feature i represents the type of feature being calculated, such as location features, social features, or home broadband service subscription and usage behavior features; Pgood,i represents the proportion of good samples to the total sample, and Pbad,i represents the proportion of bad samples to the total sample. In the scenario corresponding to the technical solution of this application embodiment, good samples are those belonging to high-value customer groups, while bad samples are the opposite.

[0133] After both the operator and the e-commerce platform have completed the extraction of feature data, the family group identification federated model can be trained according to the training process of the federated learning model. Specifically, the steps of training the federated model based on the feature data of each participant and the feature data of other participants to obtain the family group identification federated model include:

[0134] Step A41: Train the corresponding local model based on the local feature data;

[0135] Step A42: Send the gradient corresponding to the local model to the coordinator so that the coordinator can perform a weighted average calculation on the gradients received from each participant to obtain the updated gradient.

[0136] Step A43: Receive the update gradient sent by the coordinator, update the local model according to the update gradient, and obtain the updated local model.

[0137] Step A44: After iteratively training the updated local model based on local feature data, return to the execution step: send the gradient corresponding to the local model to the coordinator until the preset stopping condition is reached, and obtain the family group identification federated model.

[0138] It should be noted that in the process of training a federated model for family group identification using federated learning technology, in addition to the participation of the participating parties, a coordinator is also needed to integrate and process the model parameters of each participating party. For example... Figure 5 As shown, the coordinator first creates a key pair and sends the public key to each participant (PK, SK) = GenerateKeys(), where PK is the public key, SK is the private key, and GenerateKeys is the key pair. Then, each participant trains its corresponding local model based on its local feature data, i.e., a one-sided training model. For example, a telecom operator trains its local model based on its local feature data to obtain a one-sided training model. Similarly, an e-commerce platform trains its local model based on user online shopping behavior data to obtain a one-sided training model. The training task for each local model is represented by M. 运 =Train(F 运 ,PK), M 电 =Train(F电商 ,PK), where M 运 For the training tasks on the operator's side, F 运 For the carrier's local model, M 电 For training tasks on the e-commerce side, F 电商 This is a local model for the e-commerce platform. Then, each participant encrypts its calculated gradient and loss using its public key and sends them to the coordinator. The encrypted gradient expression is: Gradient 加密 =Encrypt(Gradient,PK), where Gradient refers to the gradient, and the expression for the encrypted loss value is: Loss 加密 =Encrypt(Loss,PK), where Loss refers to the loss.

[0139] Subsequently, the coordinating party's central server performs a weighted average calculation on the encryption gradients uploaded by each participant to obtain the updated gradient. The calculation expression is:

[0140]

[0141] Among them, [[g k The encrypted gradient is represented by ]], k represents the k-th participant in the federated learning, n represents the total weight of all participants, and nk represents the weighted weight of the k-th participant. The calculated encrypted gradient is then distributed to each participant. Each participant decrypts the received gradient and uses the decrypted gradient to update its own model gradient, thus updating its local model. This process is repeated until a preset stopping condition is met, resulting in a federated model for family group recognition. The preset stopping condition is one of the following: loss convergence, the number of iterations reaching a specified iteration limit, or the iteration time reaching a specified time.

[0142] In the virtual fusion and alignment process described above, all data undergoes de-identification or anonymization processing, ensuring that none of the participating parties can deduce the original information of other sample data besides the shared samples from the exchanged data and intermediate results. During the joint modeling and inference phase, each participating party uses appropriate encryption methods to process its feature information. Each party's data is only used for local model parameter calculations; all interactive data can only be intermediate parameters of their respective calculation results, such as gradient values ​​and loss function values. The interactive data is appropriately encrypted to ensure secure transmission and prevent the deduction of original data information. The above-described scheme in this application effectively guarantees user data privacy and security.

[0143] In one feasible embodiment, the step of inputting the data of each of the said family groups into a preset high-value customer identification federated model may include:

[0144] Step S31: The feature data in each family group data is sequentially subjected to feature scaling, data cleaning and normalization to obtain the target feature data.

[0145] Step S32: Weight the feature data of various feature types in the target feature data to obtain deep feature data. The feature types include at least one of the following: family structure features, online consumption features, offline consumption features, car usage behavior features, residential behavior features, and home furnishing consumption features.

[0146] Step S33: The deep feature data and the target feature data are fused by a preset gating signal to obtain multi-feature fusion data. The gating signal includes at least an activation function and a fully connected layer.

[0147] Step S34: Input the multi-feature fusion data into the high-value customer identification federated model to predict the corresponding high-value customer identification result.

[0148] It should be noted that feature preprocessing is required before applying the feature data in the family group data, which includes feature scaling, data cleaning and normalization.

[0149] Feature scaling typically employs Min-Max scaling or Z-score normalization.

[0150] The Min-Max scaling formula is as follows:

[0151]

[0152] The Z-score standardization formula is as follows:

[0153]

[0154] Where, X′ i These are the scaled eigenvalues, where μ(X) and σ(X) are the mean and standard deviation of the feature F, respectively.

[0155] Data cleaning includes imputing missing values, which can be done using the mean, median, or mode.

[0156] The expression for the L1 normalization square is:

[0157]

[0158] Where X′ is the normalized eigenvector.

[0159] After preprocessing, feature extraction is required based on the pre-set feature types. Typically, the target feature data for each participant spans multiple feature data. For each participant, feature fusion is performed according to the following six feature types: family structure features, online consumption features, offline consumption features, car usage behavior features, residential behavior features, and home furnishing consumption features (which can be classified as other features).

[0160] Reference Figure 6 The multi-feature extraction and fusion process shown can be divided into six categories for the multiple features X included in the target feature data of the participants: family structure, online consumption, offline consumption, word usage behavior, residential behavior, and others. The corresponding weights are W1-W6. Specifically, the fusion of each feature can be calculated by the following expression:

[0161] s=W1X' 家庭结构 +W2X' 线上消费 +W3X' 线下消费 +W4X' 用车行为 +W5X' 居住行为 +W6X' 其他 ;

[0162] Where X' 家庭结构 、X' 线上消费 、X' 线下消费 、X' 用车行为 、X' 居住行为 、X' 其他 X 消费、 X 家庭 The normalization feature, W i Representing feature X' i The corresponding weight parameters (e.g., W1 corresponds to X') 家庭结构 (Weight parameters).

[0163] Furthermore, the deep feature data s is fused with the shallow features (referring to multiple features X, equivalent to preprocessed target feature data) through a gating signal g, resulting in a multi-feature fusion result F'. The gating signal g includes a sigmoid activation function and a fully connected layer FC, with the corresponding expressions as follows: g = sigmoid(FC(x)), F' = s × g + x × (1 - g). It should be noted that if any attribute of a certain type of feature is not present in the participating party, the corresponding value is set to 0. For example, if the operator data does not include offline consumption features, the corresponding feature data is set to 0.

[0164] Finally, the fused multi-feature data is input into a pre-trained high-value customer identification federated model. This model classifies the data, predicts high-value household customers, and assigns high-value labels to the family groups of these customers. The training process of the high-value customer identification federated model is similar to that of the family group identification federated model described in the previous embodiments, and will not be repeated here.

[0165] The final high-value customer identification federated model is designed Let be the target variable, representing the value of a household customer. The goal of this model is to make predictions by learning the mapping relationship between feature vectors and the target variable, and output the target variable y, which can be represented as:

[0166]

[0167] Among them, Model is the federated model for identifying high-value customers.

[0168] This application provides a feature fusion and extraction method for identifying high-value family customers, including family structure features, online consumption features, offline consumption features, vehicle usage behavior features, and residential behavior features. The various features are then fused to obtain multi-feature fusion data for high-value customer identification and prediction. Compared with single-dimensional feature data identification methods, this method has higher accuracy in identifying customer types. Furthermore, it utilizes federated learning technology to train the federated model for high-value customer identification, ensuring the security of customer privacy data.

[0169] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the household customer identification method based on the federal model of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0170] This application also provides a household customer identification device based on a federated model, referring to... Figure 7 The federalized model-based home customer identification device includes:

[0171] The data acquisition module 10 is used to collect raw customer data, wherein the raw customer data includes at least one of the following: operator data, social data, e-commerce data, vehicle data, location data, or smart home data.

[0172] The family group identification module 20 is used to input the original customer data into a preset family group identification federated model to obtain multiple sets of family group data. The family group identification federated model is obtained by multiple participants using local customer data for federated learning.

[0173] The high-value identification module 30 is used to input the data of each family group into a preset high-value customer identification federated model to obtain high-value customer identification results. The high-value customer identification federated model is obtained by multiple participants using local customer data for federated learning. The high-value customer identification results include high-value tags corresponding to the family group data.

[0174] In one embodiment, the household customer identification device based on the federated model further includes a model optimization module, which is used for:

[0175] Based on the high-value customer identification results, evaluate the model quality of the high-value customer identification federated model;

[0176] If the quality of the model does not meet the preset requirements, the high-value customer identification federated model is optimized based on the original customer data to obtain an updated high-value customer identification federated model.

[0177] In one embodiment, the household customer identification device based on the federated model further includes a customer classification module, which is used for:

[0178] The family group data with high-value labels are input into a preset high-value customer classification federated model to determine the family type corresponding to each family group data. The high-value customer classification federated model is obtained by multiple participants using local customer data for federated learning.

[0179] Recommend the corresponding family type products to the family customers of each of the aforementioned family group data.

[0180] In one embodiment, the household customer identification device based on the federated model further includes a model training module, which is used for:

[0181] Collect existing customer data, align the existing customer data with the existing customer data of other participants to obtain training customer data, wherein the existing customer data includes at least customer attribute data, social data, location data and household business data.

[0182] The training customer data is subjected to binning statistical analysis to calculate the information value and evidence weight of various features in the training customer data.

[0183] Based on the information value and evidence weight of various features, multiple feature data are selected from the training customer data;

[0184] The federated model is trained based on the feature data of each participant and the feature data of other participants to obtain the family group identification federated model.

[0185] In one embodiment, the model training module is further configured to:

[0186] The corresponding local model is obtained by training based on the local feature data.

[0187] The gradient corresponding to the local model is sent to the coordinator, so that the coordinator can calculate the weighted average of the gradients received from each participant to obtain the updated gradient.

[0188] Receive the update gradient sent by the coordinator, update the local model according to the update gradient, and obtain the updated local model;

[0189] After iteratively training the updated local model based on local feature data, the execution steps are returned: the gradient corresponding to the local model is sent to the coordinator until a preset stopping condition is reached, thus obtaining the family group identification federated model.

[0190] In one embodiment, the high-value identification module 30 is further configured to:

[0191] The feature data in each of the family group data are sequentially subjected to feature scaling, data cleaning and normalization to obtain the target feature data;

[0192] The feature data of various feature types in the target feature data are weighted and calculated to obtain deep feature data, wherein the feature types include at least one of family structure features, online consumption features, offline consumption features, car use behavior features, residential behavior features, and home furnishing consumption features;

[0193] The deep feature data and the target feature data are fused by a preset gating signal to obtain multi-feature fused data, wherein the gating signal includes at least an activation function and a fully connected layer;

[0194] The multi-feature fusion data is input into the high-value customer identification federated model to predict the corresponding high-value customer identification result.

[0195] The federated model-based home customer identification device provided in this application, employing the federated model-based home customer identification method described in the above embodiments, can solve the technical problems of low accuracy and data security risks in current identification schemes for high-value home customers. Compared with the prior art, the beneficial effects of the federated model-based home customer identification device provided in this application are the same as those of the federated model-based home customer identification method provided in the above embodiments, and other technical features in this federated model-based home customer identification device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0196] This application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the home customer identification method based on the federated model in Embodiment 1 described above.

[0197] The following is for reference. Figure 8 The diagram illustrates a structural schematic of an electronic device suitable for implementing embodiments of this application. The electronic devices in these embodiments may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 8 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0198] like Figure 8 As shown, the electronic device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the electronic device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. While electronic devices with various systems are shown in the figures, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0199] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0200] The electronic device provided in this application employs the federated model-based home customer identification method described in the above embodiments, which can solve the technical problems of low accuracy and data security risks in current identification schemes for high-value home customers. Compared with the prior art, the beneficial effects of the electronic device provided in this application are the same as those of the federated model-based home customer identification method provided in the above embodiments, and other technical features of this electronic device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0201] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0202] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0203] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the home customer identification method based on the federated model in the above embodiments.

[0204] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0205] The aforementioned computer-readable storage medium may be included in an electronic device or may exist independently without being assembled into an electronic device.

[0206] The aforementioned computer-readable storage medium carries one or more programs that, when executed by an electronic device, cause the electronic device to: collect raw customer data, wherein the raw customer data includes at least one of operator data, social data, e-commerce data, vehicle data, location data, or smart home data; input the raw customer data into a preset family group identification federated model to obtain multiple sets of family group data, wherein the family group identification federated model is obtained by multiple participants using local customer data through federated learning; and input each of the family group data into a preset high-value customer identification federated model to obtain high-value customer identification results, wherein the high-value customer identification federated model is obtained by multiple participants using local customer data through federated learning, and the high-value customer identification results include high-value tags corresponding to the family group data.

[0207] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0208] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0209] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0210] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described federated model-based home customer identification method. This addresses the technical problems of low accuracy and data security risks in current identification schemes for high-value home customers. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the federated model-based home customer identification method provided in the above embodiments, and will not be elaborated upon here.

[0211] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the federated model-based home customer identification method described above.

[0212] The computer program product provided in this application can solve the technical problems of low accuracy and data security risks in current identification schemes for high-value household customers. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the household customer identification method based on the federated model provided in the above embodiments, and will not be repeated here.

[0213] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for identifying household customers based on a federated model, characterized in that, Applied to participating parties, the federated model-based method for identifying household customers includes: Collect raw customer data, wherein the raw customer data includes at least one of the following: operator data, social data, e-commerce data, vehicle data, location data, or smart home data; The original customer data is input into a preset family group identification federated model to obtain multiple sets of family group data. The family group identification federated model is obtained by multiple participants using local customer data for federated learning. The data of each family group is input into a preset high-value customer identification federated model to obtain high-value customer identification results. The high-value customer identification federated model is obtained by multiple participants using local customer data for federated learning. The high-value customer identification results include high-value tags corresponding to the family group data.

2. The household customer identification method based on the federated model as described in claim 1, characterized in that, After the step of inputting the data of each of the aforementioned family groups into a preset high-value customer identification federated model to obtain the high-value customer identification results, the method further includes: Based on the high-value customer identification results, evaluate the model quality of the high-value customer identification federated model; If the quality of the model does not meet the preset requirements, the high-value customer identification federated model is optimized based on the original customer data to obtain an updated high-value customer identification federated model.

3. The household customer identification method based on the federated model as described in claim 1, characterized in that, After the step of inputting the data of each of the aforementioned family groups into a preset high-value customer identification federated model to obtain the high-value customer identification results, the method further includes: The family group data with high-value labels are input into a preset high-value customer classification federated model to determine the family type corresponding to each family group data. The high-value customer classification federated model is obtained by multiple participants using local customer data for federated learning. Recommend the corresponding family type products to the family customers of each of the aforementioned family group data.

4. The household customer identification method based on the federated model as described in claim 1, characterized in that, Before the step of inputting the original customer data into a preset family group identification federated model to obtain multiple sets of family group data, the method further includes: Collect existing customer data, align the existing customer data with the existing customer data of other participants to obtain training customer data, wherein the existing customer data includes at least customer attribute data, social data, location data and household business data. The training customer data is subjected to binning statistical analysis to calculate the information value and evidence weight of various features in the training customer data. Based on the information value and evidence weight of various features, multiple feature data are selected from the training customer data; The federated model is trained based on the feature data of each participant and the feature data of other participants to obtain the family group identification federated model.

5. The household customer identification method based on the federated model as described in claim 4, characterized in that, The step of training a federated model for family group identification based on the feature data of each of the aforementioned feature data and the feature data corresponding to other participants includes: The corresponding local model is obtained by training based on the local feature data. The gradient corresponding to the local model is sent to the coordinator, so that the coordinator can calculate the weighted average of the gradients received from each participant to obtain the updated gradient. Receive the update gradient sent by the coordinator, update the local model according to the update gradient, and obtain the updated local model; After iteratively training the updated local model based on local feature data, the execution steps are returned: the gradient corresponding to the local model is sent to the coordinator until a preset stopping condition is reached, thus obtaining the family group identification federated model.

6. The household customer identification method based on the federated model as described in claim 1, characterized in that, The step of inputting the data of each of the aforementioned family groups into a preset high-value customer identification federated model includes: The feature data in each of the family group data are sequentially subjected to feature scaling, data cleaning and normalization to obtain the target feature data; The feature data of various feature types in the target feature data are weighted and calculated to obtain deep feature data, wherein the feature types include at least one of family structure features, online consumption features, offline consumption features, car use behavior features, residential behavior features, and home furnishing consumption features; The deep feature data and the target feature data are fused by a preset gating signal to obtain multi-feature fused data, wherein the gating signal includes at least an activation function and a fully connected layer; The multi-feature fusion data is input into the high-value customer identification federated model to predict the corresponding high-value customer identification result.

7. A home customer identification device based on a federated model, characterized in that, The federalized model-based home customer identification device includes: The data acquisition module is used to collect raw customer data, wherein the raw customer data includes at least one of the following: operator data, social data, e-commerce data, vehicle data, location data, or smart home data. The family group identification module is used to input the original customer data into a preset family group identification federated model to obtain multiple sets of family group data. The family group identification federated model is obtained by multiple participants using local customer data for federated learning. The high-value identification module is used to input the data of each family group into a preset high-value customer identification federated model to obtain high-value customer identification results. The high-value customer identification federated model is obtained by multiple participants using local customer data for federated learning. The high-value customer identification results include high-value tags corresponding to the family group data.

8. An electronic device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the home customer identification method based on the federated model as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the home customer identification method based on the federated model as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the federalized model-based home customer identification method as described in any one of claims 1 to 6.

Images