Data platform content recommendation method and system based on intelligent analysis

Abstract

The invention discloses a data platform content recommendation method and system based on intelligent analysis, and relates to the technical field of data analysis. Comprising the steps of 1, collecting data and preprocessing the collected data; 2, constructing a user portrait and an interest preference model according to the collected data, and 3, constructing an intelligent analysis and question recommendation engine by using the interest preference model, analyzing a data set of a data platform through the intelligent analysis and question recommendation engine, identifying a question mode, and recommending a question to the user. The problem mode comprises data distribution abnormity, trend change and correlation analysis, a problem mode based on content recommendation and a collaborative filtering algorithm is adopted to match a user portrait and a data platform, multi-dimensional recommendation content is generated according to a matching result, and the recommendation content comprises data distribution abnormity detection, trend prediction, a problem analysis report and key data visualization. And the recommendation priority and the display mode are adjusted according to the preference of the user.

Description

Technical Field

[0001] The present invention discloses a method and system for recommending content based on intelligent analysis of a data platform, which relates to the technical field of data analysis. Background Art

[0002] With the rapid development of big data and artificial intelligence technologies, data platforms have become the core tools for various enterprises and organizations to manage data and support decision-making. On these platforms, users usually need to process a large amount of data information and conduct complex analyses to discover problems, predict trends, and support decision-making. However, existing data platforms have some significant deficiencies in problem analysis and content recommendation.

[0003] Firstly, existing data platforms usually rely on users to manually set and query analysis targets, which is not only time-consuming and laborious but also prone to analysis omissions or overlooking key problems. When processing data, users need to continuously browse and filter, making it difficult to quickly locate the most important analysis content. In addition, due to the complex data types and large data volumes on the data platform, users' understanding and analysis capabilities of data may be limited, making it difficult to efficiently utilize the data resources in the platform.

[0004] Secondly, existing data platform analysis tools are usually based on general models and lack a deep understanding of users' personalized needs, making it difficult to provide accurate content recommendations. With the diversification and refinement of users' needs, relying solely on standardized analysis models is difficult to meet the personalized analysis needs of different users. Summary of the Invention

[0005] Aiming at the problems of the existing technology, the present invention provides a method and system for recommending content based on intelligent analysis of a data platform, which combines federated learning, data mining, and intelligent recommendation technologies, aiming to intelligently process and recommend the data distribution, problem analysis, and user interest preferences in the user data platform.

[0006] The specific solution proposed by the present invention is as follows:

[0007] The present invention provides a method for recommending content based on intelligent analysis of a data platform, including:

[0008] Step 1: Collect data and preprocess the collected data:

[0009] Collect the interaction data of users on the data platform. The interaction data includes operation data, and record the interaction paths and behavior patterns of users according to the operation data.

[0010] Collect historical query data, which includes the data sets analyzed by users, the filtering conditions applied, the reports generated, and the visualization charts.

[0011] Collect and analyze parameter data, where the analyzed parameter data includes filtering conditions, grouping methods, and calculation method parameters.

[0012] Extract data distribution characteristics from the data set of the data platform, including the amount of data, data types, time span, and outlier distribution.

[0013] Preprocess the collected data.

[0014] Step 2: Construct a user profile and an interest preference model based on the collected data:

[0015] Extract the user's behavioral characteristics from the collected interaction data, combine them with historical query data and analyzed parameter data, and use machine learning algorithms to construct the user's interest preference model.

[0016] Step 3: Use the interest preference model to construct an intelligent analysis and problem recommendation engine, and analyze the data set of the data platform through the intelligent analysis and problem recommendation engine to identify problem patterns, where the problem patterns include data distribution anomalies, trend changes, and correlation analysis.

[0017] Adopt content-based recommendation and collaborative filtering algorithms to match the user profile with the problem patterns of the data platform.

[0018] Generate multi-dimensional recommended content according to the matching results. The recommended content includes data distribution anomaly detection, trend prediction, problem analysis reports, and visualization of key data, and adjust the recommendation priority and display method according to the user's preferences.

[0019] Furthermore, in step 1 of the method for recommending content for a data platform based on intelligent analysis, the preprocessing of the collected data includes:

[0020] Perform data cleaning, remove invalid data, handle missing values, and standardize the data to ensure that data from different data sources can be analyzed and processed under a unified framework. The cleaning rules are customized according to data types and business logics to ensure data quality and consistency.

[0021] Furthermore, in step 2 of the method for recommending content for a data platform based on intelligent analysis, according to the user's behavioral characteristics, historical query data, and analyzed parameter data, construct a user profile by using user clustering based on clustering, user behavior classification based on decision trees, and behavior pattern recognition based on association rules.

[0022] Adopt the long short-term memory network LSTM algorithm or the Transformer model to construct the user's interest preference model.

[0023] Further, in step 3 of the method for recommending content by a data platform based on intelligent analysis, a federated learning framework is introduced into the intelligent analysis and problem recommendation engine. A personalized interest preference model is trained locally at the user terminal, and the model parameters are uploaded to the data platform for aggregation to form a globally optimized model.

[0024] Further, in step 3 of the method for recommending content by a data platform based on intelligent analysis, it specifically includes:

[0025] Perform local model training: Use a lightweight deep learning model to perform local model training on each user terminal, and the model is updated according to the user's behavioral characteristics and historical records.

[0026] Perform global model aggregation: Use the federated averaging algorithm to aggregate the model parameters uploaded by each terminal to generate a globally optimized model. Use weighted averaging and knowledge distillation methods to process the model parameters of each terminal to ensure the accuracy and generalization ability of the global model.

[0027] Perform model distribution and update: After the globally optimized model is generated, distribute the globally optimized model to each user terminal to update the local recommendation model.

[0028] Maintain privacy protection: Use differential privacy methods to encrypt and perturb the gradients or weights of the local model to ensure the security of user data during transmission and processing, and prevent data leakage and unauthorized access.

[0029] The present invention also provides a system for recommending content by a data platform based on intelligent analysis, including a collection module, a model management module, and an engine analysis and recommendation module.

[0030] The collection module collects data and preprocesses the collected data:

[0031] Collect the interaction data of the user on the data platform. The interaction data includes operation data, and the interaction path and behavior pattern of the user are recorded according to the operation data.

[0032] Collect historical query data, which includes the data sets analyzed by the user, the filtering conditions applied, the reports generated, and the visualization charts.

[0033] Collect analysis parameter data, which includes filtering conditions, grouping methods, and calculation method parameters.

[0034] Extract the data distribution characteristics from the data sets of the data platform, including the data volume, data type, time span, and outlier distribution.

[0035] Preprocess the collected data;

[0036] The model management module constructs user portraits and interest preference models based on the collected data:

[0037] Extract the user's behavioral characteristics from the collected interaction data, combine them with historical query data and analysis parameter data, and use machine learning algorithms to construct the user's interest preference model.

[0038] The engine analysis and recommendation module uses the interest preference model to construct an intelligent analysis and problem recommendation engine, and analyzes the data sets of the data platform through the intelligent analysis and problem recommendation engine to identify problem patterns, including abnormal data distribution, trend changes, and correlation analysis.

[0039] Match the user portrait with the problem patterns of the data platform using content-based recommendation and collaborative filtering algorithms.

[0040] Generate multi-dimensional recommended content based on the matching results. The recommended content includes abnormal data distribution detection, trend prediction, problem analysis reports, and visualization of key data, and adjusts the recommendation priority and display method according to the user's preferences.

[0041] Furthermore, the acquisition module of the system for recommending content for a data platform based on intelligent analysis preprocesses the collected data, including:

[0042] Perform data cleaning, remove invalid data, handle missing values, and standardize the data to ensure that data from different data sources can be analyzed and processed under a unified framework. The cleaning rules are customized according to data types and business logics to ensure data quality and consistency.

[0043] Furthermore, the model management module of the system for recommending content for a data platform based on intelligent analysis constructs user portraits using clustering-based user grouping, decision tree-based user behavior classification, and association rule-based behavior pattern recognition according to the user's behavioral characteristics, historical query data, and analysis parameter data.

[0044] Use the long short-term memory network (LSTM) algorithm or the Transformer model to construct the user's interest preference model.

[0045] Furthermore, the engine analysis and recommendation module of the system for recommending content for a data platform based on intelligent analysis introduces a federated learning framework into the intelligent analysis and problem recommendation engine, trains a personalized interest preference model locally at the user terminal, and uploads the model parameters to the data platform for aggregation to form a globally optimized model.

[0046] Further, the engine analysis and recommendation module of the system for recommending content based on intelligent analysis performs local model training: lightweight deep learning models are used for local model training on each user terminal, and the models are updated according to the user's behavioral characteristics and historical records.

[0047] The engine analysis and recommendation module performs global model aggregation: the federated averaging algorithm is used to aggregate the model parameters uploaded by each terminal to generate a globally optimized model, and weighted averaging and knowledge distillation methods are used to process the model parameters of each terminal to ensure the accuracy and generalization ability of the global model.

[0048] The engine analysis and recommendation module performs model distribution and update: after the globally optimized model is generated, it is distributed to each user terminal to update the local recommendation model.

[0049] The engine analysis and recommendation module maintains privacy protection: differential privacy methods are used to encrypt and perturb the gradients or weights of the local model to ensure the security of user data during transmission and processing, preventing data leakage and unauthorized access.

[0050] The advantages of the present invention are as follows:

[0051] Combining intelligent recommendation algorithms and data analysis techniques, it solves problems such as low analysis efficiency, inaccurate content recommendation, and difficult data privacy protection existing in existing data platforms. For the needs of users when analyzing problems and processing data on the data platform, the system can automatically identify the problems and data anomaly points that users are interested in, and perform accurate content recommendation to help users quickly obtain key data insights. In addition, the present invention shares and optimizes models among different user terminals through a federated learning framework, which not only improves the accuracy of analysis and recommendation but also effectively protects user data privacy. The system is particularly suitable for scenarios such as leadership cockpits that require real-time data insights and personalized analysis, aiming to significantly improve the intelligent management level of the data platform and the decision-making efficiency of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0053] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, so that those skilled in the art can better understand the present invention and be able to implement it, but the illustrated embodiments do not limit the present invention.

[0055] Embodiment 1

[0056] The present invention provides a method for recommending content by a data platform based on intelligent analysis, including:

[0057] Step 1: Collect data and preprocess the collected data:

[0058] Collect the interaction data of the user on the data platform. The interaction data includes operation data, and record the user's interaction path and behavior pattern according to the operation data.

[0059] Collect historical query data, which includes the data sets analyzed by the user, the filtering conditions applied, the reports generated, and the visualization charts.

[0060] Collect analysis parameter data, which includes filtering conditions, grouping methods, and calculation method parameters.

[0061] Extract the data distribution characteristics from the data sets of the data platform, including the data volume, data type, time span, and outlier distribution.

[0062] Preprocess the collected data.

[0063] Among them, preprocessing the collected data may include:

[0064] Perform data cleaning, remove invalid data, handle missing values, and standardize the data to ensure that data from different data sources can be analyzed and processed under a unified framework. The cleaning rules are customized according to the data type and business logic to ensure the quality and consistency of the data.

[0065] Step 2: Construct a user portrait and an interest preference model based on the collected data:

[0066] Extract the user's behavior characteristics from the collected interaction data, combine them with the historical query data and analysis parameter data, and use machine learning algorithms to construct the user's interest preference model.

[0067] Among them, according to the user's behavior characteristics, historical query data, and analysis parameter data, user grouping based on clustering, user behavior classification based on decision trees, and behavior pattern recognition based on association rules can be used to construct the user portrait.

[0068] Use the long short-term memory network LSTM algorithm or the Transformer model to construct the user's interest preference model.

[0069] Step 3: Use the interest preference model to build an intelligent analysis and problem recommendation engine. Analyze the data sets in the data platform through the intelligent analysis and problem recommendation engine to identify problem patterns, which include abnormal data distribution, trend changes, and correlation analysis.

[0070] Adopt content-based recommendation and collaborative filtering algorithms to match the user profile with the problem patterns in the data platform.

[0071] Generate multi-dimensional recommended content based on the matching results. The recommended content includes abnormal data distribution detection, trend prediction, problem analysis reports, and visualization of key data, and adjust the recommendation priority and display method according to the user's preferences.

[0072] Among them, in Step 3, a federated learning framework is introduced into the intelligent analysis and problem recommendation engine. A personalized interest preference model is trained locally at the user terminal, and the model parameters are uploaded to the data platform for aggregation to form a globally optimized model. It can specifically include:

[0073] Perform local model training: Use a lightweight deep learning model to perform local model training on each user terminal, and the model is updated according to the user's behavioral characteristics and historical records.

[0074] Perform global model aggregation: Use the federated averaging algorithm to aggregate the model parameters uploaded by each terminal to generate a globally optimized model. Use weighted averaging and knowledge distillation methods to process the model parameters of each terminal to ensure the accuracy and generalization ability of the global model.

[0075] Perform model distribution and update: After the globally optimized model is generated, distribute the globally optimized model to each user terminal to update the local recommendation model.

[0076] Maintain privacy protection: Use differential privacy methods to encrypt and perturb the gradients or weights of the local model to ensure the security of user data during transmission and processing, and prevent data leakage and unauthorized access.

[0077] The method of the present invention can be applied to the leadership cockpit scenario. Through the intelligent analysis and recommendation engine, it can provide key data insights and problem analysis reports to decision-makers in real time, helping decision-makers make efficient and accurate decisions quickly. For example, it includes:

[0078] Perform real-time data monitoring and anomaly detection: Real-time monitor the key data indicators in the leadership cockpit. When data anomalies or trend changes are detected, automatically generate a problem analysis report and recommend relevant analysis content.

[0079] Perform multi-dimensional data visualization: Automatically generate multi-dimensional data visualization charts according to users' interest preferences, including data distribution charts, trend prediction charts, correlation analysis charts, etc., providing intuitive data insights for leaders.

[0080] Support intelligent decision-making: By analyzing historical decision data and current data status, automatically recommend possible decision-making plans and optimization strategies to assist leaders in making scientific decisions.

[0081] Further conduct data mining and predictive analysis: Through data mining techniques, identify potential patterns and trends in data, predict possible future changes, and provide forward-looking decision-making support, etc.

[0082] Embodiment 2

[0083] The present invention also provides a system for recommending content based on an intelligent analysis data platform, including a collection module, a model management module, and an engine analysis and recommendation module.

[0084] The collection module collects data and preprocesses the collected data:

[0085] Collect the interaction data of users on the data platform. The interaction data includes operation data, and record the interaction paths and behavior patterns of users according to the operation data.

[0086] Collect historical query data, which includes the data sets analyzed by users, the filtering conditions applied, the reports and visualization charts generated.

[0087] Collect analysis parameter data, which includes filtering conditions, grouping methods, calculation method parameters.

[0088] Extract data distribution characteristics from the data sets of the data platform, including the data volume, data type, time span, and outlier distribution.

[0089] Preprocess the collected data;

[0090] The model management module constructs a user portrait and an interest preference model according to the collected data:

[0091] Extract the behavior characteristics of users from the collected interaction data, combine them with historical query data and analysis parameter data, and use machine learning algorithms to construct a user interest preference model.

[0092] The engine analysis and recommendation module uses the interest preference model to construct an intelligent analysis and problem recommendation engine, and analyzes the data sets of the data platform through the intelligent analysis and problem recommendation engine to identify problem patterns, and the problem patterns include data distribution anomalies, trend changes, and correlation analysis.

[0093] Adopt content-based recommendation and collaborative filtering algorithms to match the user profile with the problem patterns of the data platform.

[0094] Generate multi-dimensional recommended content according to the matching results. The recommended content includes data distribution anomaly detection, trend prediction, problem analysis reports, and visualization of key data, and adjust the recommendation priority and display method according to the user's preferences.

[0095] Regarding the information interaction, execution process, etc. among the modules in the above system, since they are based on the same concept as the method embodiment of the present invention, the specific content can be referred to the description in the method embodiment of the present invention and will not be elaborated here.

[0096] Similarly, the advantages of the system of the present invention are as follows:

[0097] Through intelligent analysis and recommendation algorithms, automatically identify and recommend the key problems that users may encounter on the data platform, greatly improving the efficiency and accuracy of data analysis. Users can obtain personalized and accurate content recommendations without manually querying and filtering data, significantly improving work efficiency and user satisfaction. Secondly, the system realizes highly personalized recommendation services by constructing user profiles and interest preference models, ensuring that users can obtain analysis content and data displays that meet their needs in different scenarios. At the same time, the system adopts a federated learning framework to optimize the global model while protecting user data privacy, avoiding the risk of data leakage, and further enhancing the security of data processing through technologies such as differential privacy and homomorphic encryption. The system is particularly suitable for high-level decision-making scenarios such as leadership cockpits, and can provide real-time key data insights and forward-looking analysis, significantly improving decision-making efficiency and accuracy. In addition, the system has high flexibility and scalability, adopts a modular design, supports the access of multiple data sources and the application of multiple platforms, and can adapt to different business scenarios and user needs. The system's adaptive optimization function and online model update mechanism ensure that it still maintains efficient and stable operation in a diverse demand and large-scale data processing environment. At the same time, the system supports seamless applications on multiple terminals such as desktop, mobile, and cloud, providing consistent intelligent analysis and recommendation services, greatly expanding the application scenarios and commercial value of the system. In summary, the present invention shows significant technical advantages in multiple aspects such as data analysis, user experience, privacy protection, and security, and can provide a comprehensive solution for the intelligent management and optimization of user data platforms, with broad market application prospects and competitiveness.

[0098] It should be noted that not all steps and modules in the above-mentioned processes and system architectures are necessary, and some steps or modules can be ignored according to actual needs. The execution order of each step is not fixed and can be adjusted as needed. The system architectures described in the above-mentioned embodiments can be physical architectures or logical architectures, that is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities separately, or some components in multiple independent devices may jointly implement them.

[0099] The above-mentioned embodiments are only preferred embodiments cited to fully illustrate the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or transformations made by those skilled in the art on the basis of the present invention are within the protection scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. A method for a data platform to recommend content based on intelligent analysis, characterized in that Including: Step 1: Collect data and preprocess the collected data: Collect the interaction data of users on the data platform. The interaction data includes operation data, and record the interaction paths and behavior patterns of users according to the operation data. Collect historical query data, which includes the data sets analyzed by users, the filtering conditions applied, the reports generated, and the visualization charts. Collect analysis parameter data, which includes filtering conditions, grouping methods, and calculation method parameters. Extract the data distribution characteristics from the data sets of the data platform, including the data volume, data type, time span, and outlier distribution. Preprocess the collected data. Step 2: Construct user portraits and interest preference models based on the collected data: Extract the behavioral characteristics of users from the collected interaction data, combine them with the historical query data and analysis parameter data, and use machine learning algorithms to construct the interest preference models of users. Step 3: Use the interest preference models to construct an intelligent analysis and problem recommendation engine, and analyze the data sets of the data platform through the intelligent analysis and problem recommendation engine to identify problem patterns. The problem patterns include data distribution anomalies, trend changes, and correlation analysis. Adopt content-based recommendation and collaborative filtering algorithms to match user portraits with the problem patterns of the data platform. Generate multi-dimensional recommended content according to the matching results. The recommended content includes data distribution anomaly detection, trend prediction, problem analysis reports, and visualization of key data, and adjust the recommendation priority and display method according to the preferences of users.

2. The method for recommending content by a data platform based on intelligent analysis according to claim 1, characterized in that In Step 1, the preprocessing of the collected data includes: Perform data cleaning, remove invalid data, handle missing values, and standardize the data to ensure that data from different data sources can be analyzed and processed in a unified framework. The cleaning rules are customized according to the data type and business logic to ensure the quality and consistency of the data.

3. The method for recommending content by a data platform based on intelligent analysis according to claim 1, characterized in that In Step 2, construct user portraits according to the behavioral characteristics, historical query data, and analysis parameter data of users, using user clustering based on clustering, user behavior classification based on decision trees, and behavior pattern recognition based on association rules. Adopt the long short-term memory network LSTM algorithm or Transformer model to construct the interest preference models of users.

4. A method for recommending content for a data platform based on intelligent analysis according to claim 1, characterized in that in Step 3, a federated learning framework is introduced into the intelligent analysis and problem recommendation engine, and a personalized interest preference model is trained locally at the user terminal, and the model parameters are uploaded to the data platform for aggregation to form a globally optimized model.

5. A method for recommending content by a data platform based on intelligent analysis according to claim 1, characterized in that In Step 3, it specifically includes: Perform local model training: Use lightweight deep learning models to perform local model training on each user terminal, and the model is updated according to the behavioral characteristics and historical records of users. Perform global model aggregation: Adopt the federated average algorithm to aggregate the model parameters uploaded by each terminal to generate a globally optimized model, and use weighted average and knowledge distillation methods to process the model parameters of each terminal to ensure the accuracy and generalization ability of the global model. Perform model distribution and update: After the globally optimized model is generated, distribute the globally optimized model to each user terminal to update the local recommendation model. Maintain privacy protection: Use the differential privacy method to encrypt and perturb the gradients or weights of the local model to ensure the security of user data during transmission and processing, preventing data leakage and unauthorized access.

6. A system for recommending content by a data platform based on intelligent analysis, characterized in that It includes a data collection module, a model management module, and an engine analysis and recommendation module. The data collection module collects data and preprocesses the collected data: Collect the interaction data of users on the data platform. The interaction data includes operation data, and record the interaction paths and behavior patterns of users according to the operation data. Collect historical query data, which includes the data sets that users have analyzed, the filtering conditions applied, the reports generated, and the visualization charts. Collect analysis parameter data, which includes filtering conditions, grouping methods, and calculation method parameters. Extract the data distribution characteristics from the data sets on the data platform, including the data volume, data type, time span, and outlier distribution. Preprocess the collected data. The model management module constructs user portraits and interest preference models based on the collected data: Extract the behavioral characteristics of users from the collected interaction data, combine them with the historical query data and analysis parameter data, and use machine learning algorithms to construct the interest preference models of users. The engine analysis and recommendation module uses the interest preference model to construct an intelligent analysis and problem recommendation engine, and analyzes the data sets on the data platform through the intelligent analysis and problem recommendation engine to identify problem patterns. The problem patterns include abnormal data distribution, trend changes, and correlation analysis. Adopt content-based recommendation and collaborative filtering algorithms to match user portraits with the problem patterns on the data platform. Generate multi-dimensional recommendation content according to the matching results. The recommendation content includes abnormal data distribution detection, trend prediction, problem analysis reports, and key data visualization, and adjust the recommendation priority and display method according to the user's preferences.

7. The system for recommending content on a data platform based on intelligent analysis according to claim 6, characterized in that The data collection module preprocesses the collected data, including: Perform data cleaning, remove invalid data, handle missing values, and standardize the data to ensure that data from different data sources can be analyzed and processed under a unified framework. The cleaning rules are customized according to the data type and business logic to ensure the quality and consistency of the data.

8. The system for recommending content of a data platform based on intelligent analysis according to claim 6, characterized in that The model management module constructs user portraits based on the behavioral characteristics, historical query data, and analysis parameter data of users, using clustering-based user grouping, decision tree-based user behavior classification, and association rule-based behavior pattern recognition. Adopt the long short-term memory network LSTM algorithm or the Transformer model to construct the interest preference model of users.

9. A system for recommending content on a data platform based on intelligent analysis according to claim 6, characterized in that the engine The analysis and recommendation module introduces a federated learning framework into the intelligent analysis and problem recommendation engine, trains a personalized interest preference model locally at the user terminal, and uploads the model parameters to the data platform for aggregation to form a globally optimized model.

10. A method for recommending content by a data platform based on intelligent analysis according to claim 1, characterized in that The engine analysis and recommendation module conducts local model training: Use lightweight deep learning models to perform local model training on each user terminal, and the model is updated according to the behavioral characteristics and historical records of users. The engine analysis and recommendation module performs global model aggregation: The federated averaging algorithm is used to aggregate the model parameters uploaded by each terminal to generate a globally optimized model. The weighted average and knowledge distillation methods are used to process the model parameters of each terminal to ensure the accuracy and generalization ability of the global model. The engine analysis and recommendation module performs model distribution and update: After the globally optimized model is generated, the globally optimized model is distributed to each user terminal to update the local recommendation model. The engine analysis and recommendation module maintains privacy protection: The differential privacy method is used to encrypt and perturb the gradients or weights of the local model to ensure the security of user data during transmission and processing, preventing data leakage and unauthorized access.

Images