A Big Data Visualization Processing Method and System for a Smart Radio and Television Project

Abstract

The present invention discloses a big data visualization processing method and system for a smart radio and television project. The method includes: performing multi-source data collection based on user behavior data, content playback data, network traffic data, and device status data of the smart radio and television project to obtain a cleaned data set; inputting the cleaned data set into a feature extraction model to extract the correlation features of user behavior, content preferences, and device status, and obtaining a multi-dimensional feature matrix; inputting the multi-dimensional feature matrix into a visualization model to generate an initial visualization chart and obtaining a preliminary optimized visualization design scheme; deploying the visualization design scheme to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations, and obtaining a final smart radio and television big data visualization display scheme. By using the embodiments of the present invention, the processing efficiency and display effect of radio and television big data can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data, and particularly relates to a big data visualization processing method and system for a smart radio and television project. Background Art

[0002] With the rapid development of information technology and the advent of the big data era, the radio and television industry is facing unprecedented changes and opportunities. Through smart radio and television projects, radio and television institutions can collect and analyze user behavior data, content playback data, network traffic data, and device status data more efficiently, so as to provide more personalized and intelligent services for users. However, traditional data processing and display methods are unable to cope with massive, multi-source, and high-dimensional data, and cannot fully explore the potential value of data, thus affecting the accuracy of decision-making and the quality of services. Summary of the Invention

[0003] The purpose of the present invention is to provide a big data visualization processing method and system for a smart radio and television project to solve the deficiencies in the prior art and improve the processing efficiency and display effect of radio and television big data.

[0004] An embodiment of the present application provides a big data visualization processing method for a smart radio and television project, and the method includes:

[0005] Perform multi-source data collection based on user behavior data, content playback data, network traffic data, and device status data of the smart radio and television project, and use a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm to identify and eliminate noise data and redundant information in real time, and obtain a cleaned data set;

[0006] Input the cleaned data set into a feature extraction model based on a graph neural network to extract the correlation features of user behavior, content preference, and device status. Among them, the feature extraction model captures the implicit correlation relationships between data through dynamic graph construction and multi-hop relationship reasoning technologies, and obtains a multi-dimensional feature matrix;

[0007] Input the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart by combining attention technology and an adaptive layout algorithm, and obtains a preliminary optimized visualization design scheme;

[0008] Deploy the visualization design scheme to an interactive visualization platform to support users to explore data through multi-dimensional filtering and dynamic interaction operations. Among them, the interactive visualization platform updates the visualization content in real time through streaming computing and incremental update technologies to ensure the timeliness and operability of the data, and obtains a final big data visualization display scheme for smart radio and television.

[0009] Optionally, based on the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project, multi-source data collection is performed. A hybrid technology based on an adaptive rule engine and an anomaly detection algorithm is used to identify and eliminate noise data and redundant information in real time, resulting in a cleaned data set, including:

[0010] Based on the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project, a data collection framework based on edge computing is used to obtain multi-source data in real time. Through lightweight data caching technology, the real-time and continuity of data collection are ensured;

[0011] For the collected multi-source data, a cleaning method based on an adaptive rule engine is adopted. Combining preset business rules, data that does not conform to the rules is dynamically identified and eliminated. Through rule dynamic update technology, the timeliness and accuracy of the cleaning rules are ensured, and a preliminary cleaned data set is generated;

[0012] For the preliminary cleaned data set, a noise elimination method based on an anomaly detection algorithm is adopted. Combining the historical data distribution and real-time data stream, noise data and redundant information are identified and eliminated. Through dynamic threshold adjustment technology, the accuracy and robustness of anomaly detection are ensured, and a preliminary denoised data set is generated;

[0013] For the preliminary denoised data set, a method based on a data integration algorithm is adopted to map the multi-source data into a unified data structure. Through data consistency verification technology, the integrity and consistency of the data are ensured, and a final cleaned data set is generated.

[0014] Optionally, the cleaned data set is input into a feature extraction model based on a graph neural network to extract the correlation features of user behavior, content preference, and device status. Among them, the feature extraction model captures the implicit correlation relationships between data through dynamic graph construction and multi-hop relationship reasoning technology, resulting in a multi-dimensional feature matrix, including:

[0015] For the cleaned data set, a feature extraction model based on a graph neural network is adopted. Users, content, network traffic, and devices are abstracted as nodes in the graph structure, and the correlation relationships between the nodes are abstracted as edges. Through dynamic graph construction technology, a preliminary graph structure is generated;

[0016] For the preliminary graph structure, a feature extraction method based on multi-hop relationship reasoning technology is adopted. Combining the correlation relationships of user behavior, content preference, and device status, implicit features are extracted. Through attention technology, the selection of the reasoning path is optimized, and a preliminary feature representation is generated;

[0017] For the preliminary feature representation, a feature fusion method based on matrix factorization is adopted to perform weighted fusion on user behavior features, content preference features, and device status features. Through dynamic weight adjustment technology, the timeliness and consistency of the feature vectors are ensured, and a preliminary multi-dimensional feature matrix is generated;

[0018] For the preliminary multi-dimensional feature matrix, an optimization method based on error feedback technology is adopted. Combining real-time data streams and historical data distributions, the feature weights are dynamically adjusted. Through regularization constraints, overfitting is prevented, and a final multi-dimensional feature matrix is generated.

[0019] Optionally, the multi-dimensional feature matrix is input into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart by combining attention technology and an adaptive layout algorithm, and obtains a preliminary optimized visualization design scheme, including:

[0020] For the multi-dimensional feature matrix, a visualization model based on a generative adversarial network is adopted. Combining the features of user behavior, content preference, and device status, an initial visualization chart is generated. Through conditional constraint technology, the accuracy and readability of the chart are ensured, and a preliminary visualization chart is generated;

[0021] For the preliminary visualization chart, an optimization method based on attention technology is adopted. Combining the key information that users are concerned about, the visual expression of the chart is dynamically adjusted. Through multi-scale attention technology, information densities at different levels are captured, and a preliminary optimized visualization chart is generated;

[0022] For the preliminary optimized visualization chart, an adjustment method based on an adaptive layout algorithm is adopted. Combining the visual expression and information density of the chart, the layout of the chart is dynamically adjusted. Through dynamic weight allocation technology, the visual effect and information transmission efficiency of the chart are optimized, and a preliminary optimized visualization design scheme is generated;

[0023] For the preliminary optimized visualization design scheme, a verification method based on user feedback is adopted. Combining user interaction data and real-time performance monitoring, the design scheme is dynamically adjusted. Through feedback correction technology, a final preliminary optimized visualization design scheme is generated.

[0024] Optionally, the visualization design scheme is deployed to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations. Among them, the interactive visualization platform updates the visualization content in real time through streaming computing and incremental update technology to ensure the timeliness and operability of the data, and obtains a final intelligent radio and television big data visualization display scheme, including:

[0025] For the preliminarily optimized visual design scheme, adopt a deployment method based on an interactive visualization platform. Combine user requirements and technical architecture, deploy the design scheme to the platform, and ensure the efficiency and stability of the deployment through dynamic loading technology to generate a preliminary visual display scheme;

[0026] For the preliminary visual display scheme, adopt an interactive operation support method based on multi-dimensional screening. Combine user behavior data and content preferences, dynamically adjust the screening conditions, and support users to explore data through click and drag operations via an interactive operation interface to generate a preliminary interactive visualization scheme;

[0027] For the preliminary interactive visualization scheme, adopt a real-time update method based on stream computing and incremental update technology. Combine real-time data streams and historical data distributions, dynamically update the visual content, and ensure the timeliness and operability of the data through incremental update technology to generate a preliminary real-time update visualization scheme;

[0028] For the preliminary real-time update visualization scheme, adopt an optimization method based on user feedback. Combine user interaction data and real-time performance monitoring, dynamically adjust the visual content, and generate a final big data visualization display scheme for intelligent radio and television through feedback correction technology.

[0029] Another embodiment of the present application provides a big data visualization processing system for an intelligent radio and television project. The system includes:

[0030] A collection module for performing multi-source data collection based on user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project. Adopt a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm to identify and eliminate noise data and redundant information in real time to obtain a cleaned data set;

[0031] An extraction module for inputting the cleaned data set into a feature extraction model based on a graph neural network to extract the associated features of user behavior, content preferences, and device status. Among them, the feature extraction model captures the implicit association relationships between data through dynamic graph construction and multi-hop relationship reasoning technology to obtain a multi-dimensional feature matrix;

[0032] A visualization module for inputting the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart by combining attention technology and an adaptive layout algorithm to obtain a preliminarily optimized visual design scheme;

[0033] A deployment module for deploying the visual design solution to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations. The interactive visualization platform uses streaming computing and incremental update technologies to update the visualization content in real time, ensuring the timeliness and operability of the data, and obtaining the final intelligent radio and television big data visualization display solution.

[0034] Another embodiment of the present application provides a storage medium in which a computer program is stored. The computer program is configured to execute the method described in any one of the above when running.

[0035] Another embodiment of the present application provides an electronic device including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the method described in any one of the above.

[0036] Compared with the prior art, a big data visualization processing method for an intelligent radio and television project provided by the present invention performs multi-source data collection based on user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project to obtain a cleaned data set; inputs the cleaned data set into a feature extraction model to extract the correlation features of user behavior, content preferences, and device status, obtaining a multi-dimensional feature matrix; inputs the multi-dimensional feature matrix into a visualization model to generate an initial visualization chart, obtaining a preliminary optimized visual design solution; deploys the visual design solution to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations, obtaining the final intelligent radio and television big data visualization display solution, thereby being able to improve the processing efficiency and display effect of radio and television big data. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a hardware structure block diagram of a computer terminal for a big data visualization processing method of an intelligent radio and television project provided by an embodiment of the present invention;

[0038] Figure 2 It is a flow schematic diagram of a big data visualization processing method of an intelligent radio and television project provided by an embodiment of the present invention;

[0039] Figure 3 It is a structure schematic diagram of a big data visualization processing system of an intelligent radio and television project provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0041] An embodiment of the present invention first provides a big data visualization processing method for a smart radio and television project. This method can be applied to electronic devices, such as computer terminals, specifically ordinary computers, etc.

[0042] The following takes running on a computer terminal as an example to describe it in detail. Figure 1 It is a hardware structure block diagram of a computer terminal for the big data visualization processing method of a smart radio and television project provided by an embodiment of the present invention. As Figure 1 shown, the computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the memory can include a non-volatile storage medium and an internal memory.

[0043] The non-volatile storage medium can store an operating system and computer programs. The computer programs include program instructions. When the program instructions are executed, the processor can execute any big data visualization processing method for a smart radio and television project.

[0044] The processor is used to provide computing and control capabilities to support the operation of the entire computer device.

[0045] The internal memory provides an environment for the operation of the computer programs in the non-volatile storage medium. When the computer programs are executed by the processor, the processor can execute any big data visualization processing method for a smart radio and television project.

[0046] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that Figure 1 the structure shown in

[0047] is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0048] See Figure 2, embodiments of the present invention provide a big data visualization processing method for smart radio and television projects, which may include the following steps:

[0049] S201, perform multi-source data collection based on the user behavior data, content playback data, network traffic data, and device status data of the smart radio and television project, and use a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm to identify and eliminate noise data and redundant information in real time, obtaining a cleaned data set;

[0050] This step aims to efficiently and accurately collect and clean multi-source data in radio and television projects. Obtain data in real time through an edge computing framework, and use lightweight data caching technology to ensure data continuity. By combining an adaptive rule engine and an anomaly detection algorithm, dynamically identify and eliminate data that does not conform to the rules and noise data, ensuring that the generated cleaned data set has high quality and high reliability.

[0051] Through this step, the efficiency and accuracy of data collection and cleaning can be significantly improved, making subsequent data processing and analysis more accurate and highly credible. By using a hybrid technology that combines an adaptive rule engine and an anomaly detection algorithm, the cleaning rules can be dynamically updated and the anomaly detection threshold can be adjusted, improving the robustness and timeliness of data cleaning, providing a reliable data foundation for smart radio and television projects, and ensuring that the data of user behavior, content preferences, network traffic, and device status truly reflects the actual situation.

[0052] Specifically, based on the user behavior data, content playback data, network traffic data, and device status data of the smart radio and television project, a data collection framework based on edge computing can be used to obtain multi-source data in real time, and through lightweight data caching technology, ensure the real-time and continuous nature of data collection;

[0053] This step obtains multi-source data in the radio and television project in real time through an edge computing framework, including user behavior data, content playback data, network traffic data, and device status data. Use lightweight data caching technology to temporarily store the collected data to ensure the real-time and continuous nature of data transmission and avoid data loss. The significance of this step is that through edge computing and lightweight caching technology, data can be collected efficiently and securely, ensuring the real-time and continuous nature of data, thereby providing a reliable and complete data foundation for subsequent data processing and analysis. This is the first step in the data processing of smart radio and television projects and a key link to ensure the integrity of the data chain.

[0054] To implement an edge-computing-based data acquisition framework, edge devices are first installed at the core positions of radio and television projects and user terminals. These devices can be smart routers, set-top boxes, or dedicated acquisition devices, and lightweight data acquisition software runs on these devices. This software needs to be able to monitor and collect users' behavior data, content playback records, network traffic monitoring data, and device operating status in real time. For example, when a user changes channels, requests videos on demand, or performs other interactive operations, this data will be recorded by the edge devices.

[0055] Through the edge computing framework, data can not only be initially processed and cached on edge devices, but also relieve the computing and storage pressure on the central server. In the data acquisition process, lightweight data caching technology plays a key role, which ensures that data can be temporarily stored before being transmitted to the central server. For example, when a user is watching a certain program, the edge device continuously collects data such as the user's viewing duration and switching times, and temporarily stores these data in the device's cache. When the network connection is stable, the cached data is then batch-transmitted to the central server. This method not only improves the real-time performance of data acquisition, but also avoids data loss caused by network fluctuations.

[0056] In addition, lightweight data caching technology usually combines efficient data transmission protocols, such as MQTT (Message Queuing Telemetry Transport) or CoAP (Constrained Application Protocol), which can ensure the reliable transmission of data in low-bandwidth and high-latency network environments. For example, if the network is congested at a certain moment, the data caching technology can temporarily store the data and quickly transmit the data after the network returns to normal, ensuring the continuity and real-time performance of the data.

[0057] For the multi-source data collected, a cleaning method based on an adaptive rule engine is adopted. Combining with preset business rules, it dynamically identifies and eliminates data that does not conform to the rules. Through the rule dynamic update technology, the timeliness and accuracy of the cleaning rules are ensured, and a preliminary cleaned data set is generated;

[0058] In this step, the adaptive rule engine is applied to clean the multi-source data that has been collected. The adaptive rule engine dynamically identifies and eliminates data that does not conform to the rules according to the preset business rules, and through the rule dynamic update technology, ensures that the cleaning rules can timely reflect the latest business requirements and data characteristics. Using the adaptive rule engine to clean data can greatly improve the efficiency and accuracy of data cleaning, ensuring that the cleaned data set can meet the actual business needs and has high quality. Through the dynamic update of the rules, it can quickly adapt to business changes, avoid data mis-cleaning or omission caused by rule aging, and ensure the timeliness and reliability of the data cleaning process.

[0059] When cleaning the collected multi-source data, it is first necessary to establish an adaptive rule engine. This engine presets business rules related to radio and television projects, including legal patterns of user behavior, normal processes of content playback, reasonable ranges of network traffic, and stable states of device operation. For example, if a user frequently switches channels within a short period, this behavior may be regarded as abnormal, and the rule engine will mark it as data that does not conform to the rules.

[0060] The rule engine will automatically update the cleaning rules according to the changes in data characteristics to ensure that it can always adapt to the latest data situation. For example, a certain specific program may trigger concentrated viewing by users. In this case, the rule engine will automatically adjust the normal range of user behavior to avoid misdeleting normal data. The dynamically updated rules should not only be based on the statistical results of historical data but also combine the characteristics of real-time data and be optimized through adaptive algorithms to ensure that the rules are always timely and accurate.

[0061] To further illustrate, an actual scenario can be given: Suppose that during a certain period, there is a sudden abnormal peak in network traffic, which may be caused by a certain popular program. Traditional static rules may misjudge this data as abnormal, while the adaptive rule engine will analyze the distribution of historical data and real-time data, dynamically adjust the threshold, retain this data, and at the same time update the rules to cope with similar situations in the future. The preliminary cleaned data set generated in this way can accurately reflect the real data.

[0062] For the preliminary cleaned data set, a noise elimination method based on anomaly detection algorithms is adopted. Combining the historical data distribution and real-time data stream, noise data and redundant information are identified and eliminated. Through dynamic threshold adjustment technology, the accuracy and robustness of anomaly detection are ensured, and a preliminary denoised data set is generated;

[0063] In this step, noise elimination is carried out on the preliminary cleaned data set based on anomaly detection algorithms. By combining the historical data distribution and real-time data stream, noise and redundant information in the data are identified, and through dynamic threshold adjustment technology, the accuracy and robustness of the anomaly detection process are ensured, and finally a clean denoised data set is generated. The role of this step is to further improve the quality of the data set. By eliminating noise and redundant information, it is ensured that subsequent analysis and processing are based on high-quality data that truly reflects the actual situation. The dynamic threshold adjustment technology ensures the flexibility and adaptability of anomaly detection, and can accurately identify and eliminate noise even when the data characteristics change, enhancing the robustness and stability of the data processing system.

[0064] Noise elimination is performed on the preliminary cleaned dataset. First, an anomaly detection algorithm model needs to be established. This model utilizes the characteristics of historical data distribution and real-time data streams to identify outliers and redundant information in the data. For example, by analyzing historical data, a normal distribution pattern of user behavior can be constructed. When data is detected to deviate from this pattern, the system will mark it as noise data.

[0065] To achieve this process, various anomaly detection algorithms can be adopted, such as the statistical Z-score method, the machine learning-based Isolation Forest algorithm, etc. These algorithms will set initial thresholds by combining the historical distribution and real-time changes of the data. For example, if the status data of a certain device suddenly shows extreme changes and such scenarios have never occurred in historical data, the anomaly detection algorithm will identify this data as noise and eliminate it.

[0066] The dynamic threshold adjustment technique is the key to ensuring the accuracy and robustness of anomaly detection. As real-time data is continuously input and historical data is updated, the threshold needs to be adjusted continuously to adapt to the new data characteristics. For example, during a certain period, network traffic may generate new peak values due to program changes, and static thresholds may not be able to correctly judge the anomalies of this data. The dynamic threshold adjustment technique will continuously optimize the threshold setting according to the real-time changes of the data, so as to accurately identify and eliminate noise data. This method not only improves the accuracy of data cleaning but also enhances the system's robustness to various data anomaly situations.

[0067] For the preliminary denoised dataset, a method based on data integration algorithms is adopted to map multi-source data into a unified data structure. Through data consistency verification technology, the integrity and consistency of the data are ensured, and the final cleaned dataset is generated.

[0068] This step aims to further integrate the preliminary dataset that has been denoised. The data integration algorithm is used to map multi-source data into a unified data structure. Through data consistency verification technology, it is ensured that the merged data maintains integrity and consistency in terms of structure and content, and the final cleaned dataset is generated. Through data integration and consistency verification, it is ensured that the final dataset is consistent and of high quality in form and content, providing a standardized data basis for subsequent analysis. Integrating data from different sources makes the data more complete, eliminates the problems of information duplication or contradiction between different data sources, improves the utilization efficiency and accuracy of the data, and ensures the reliability of the analysis results.

[0069] When processing the preliminary denoised dataset, a data integration algorithm is adopted to map multi-source data into a unified data structure. This process requires standardizing data from different sources to ensure their correct alignment in the unified data structure. For example, after standardizing user behavior data, content playback data, network traffic data, and device status data, they are mapped into a unified table structure, and each data record contains information such as user ID, behavior type, played content, traffic value, and device status.

[0070] During the data integration process, to ensure data consistency, strict data consistency verification is required. This process includes data format verification, data redundancy verification, and logical verification. For example, multiple data from different sources may contain duplicate records, and through verification, it can be ensured that these duplicate records are correctly merged. Logical verification ensures that the relationships between data are reasonable. For example, the timestamp in user behavior data should match the timestamp in content playback data, and network traffic data should correspond to device status data.

[0071] For example, suppose there is a data record of a user watching a video at a certain time point. This record should contain information such as user ID, ID of the watched content, start playback time, end playback time, consumed traffic data, and device status. The integration process extracts this information from each data source and maps it into a unified record structure. If it is found that the time range or content ID of some records is inconsistent, the system will correct or eliminate them, and finally generate a cleaned dataset that passes the consistency verification. This processing method not only ensures data integrity and consistency but also provides a standardized data basis for subsequent complex data analysis.

[0072] S202, input the cleaned dataset into a feature extraction model based on a graph neural network to extract the associated features of user behavior, content preference, and device status. Among them, the feature extraction model captures the implicit association relationships between data through dynamic graph construction and multi-hop relationship reasoning techniques, and obtains a multi-dimensional feature matrix;

[0073] The dynamic graph construction technique can dynamically generate a multi-level association graph among users, content, networks, and devices. Through the multi-hop relationship reasoning technique, path reasoning and feature propagation are carried out between different data points to further reveal the deep dependence relationships between data in different dimensions. Through this process, the graph neural network realizes the deep association mining of different data sources in the radio and television project, ensuring the comprehensiveness and accuracy of feature extraction. The multi-dimensional feature matrix reflects the complex associations of user behavior, content preference, and device status, capturing the important relationships and dynamic changes between data. This deep association is difficult to achieve by traditional methods, providing high-quality and deep-information input data for the subsequent visualization model, and further improving the effect of big data visualization and decision support capabilities.

[0074] Specifically, we can use a feature extraction model based on a graph neural network to abstract users, content, network traffic, and devices into nodes in a graph structure, and abstract the relationships between nodes into edges. Then, we can generate a preliminary graph structure through dynamic graph construction technology.

[0075] In this step, the cleaned data set is input into the feature extraction model based on the graph neural network. Data sources such as users, content, network traffic and devices are abstracted into nodes in the graph structure, and the associations between these nodes are abstracted into edges. This method allows us to store and process data in a more intuitive and structured way. Dynamic graph construction technology dynamically generates and updates the structure of the graph through the input of real-time data streams, so that the graph can reflect the latest associations between data in a timely manner.

[0076] By abstracting data into a graph structure, the complex associations between different data sources can be intuitively displayed, which helps to mine deeper data features. Dynamic graph construction technology ensures that the graph structure can dynamically adapt to changes in real-time data and provides an efficient and flexible way to process data. This method not only improves the accuracy of data association analysis, but also provides a solid foundation for subsequent multi-hop relationship reasoning and feature extraction, ensuring that the model can capture the latest and most comprehensive data relationships.

[0077] First, the cleaned data needs to be preprocessed, and user behavior data, content playback data, network traffic data, and device status data need to be marked as different types of nodes. Each node will have its own specific attributes. For example, the user node may contain information such as user ID, viewing history, browsing habits, etc., the content node may contain attributes such as video ID, playback time, content type, etc., the network traffic node includes bandwidth usage, packet loss rate, etc., and the device status node contains device ID, running time, fault records, etc. These attribute information will be uniformly mapped to the graph structure.

[0078] Next, define the relationship between nodes and generate edges in the graph. For example, when a user watches a video, this constitutes an edge between the user node and the content node; when the user consumes network traffic during the viewing process, this constitutes an edge between the user node and the network traffic node; when the device's state changes while playing content, it constitutes an edge between the device node and the content node. Each relationship is represented by a type of edge and is assigned a corresponding weight. Once all nodes and edges are defined, the dynamic graph construction technology dynamically organizes this information into a graph structure.

[0079] During the dynamic graph construction process, the real-time nature and dynamic change characteristics of the data need to be considered. When new user behaviors occur, new video content is added, the network traffic situation changes, or the device status is updated, the graph structure needs to reflect these changes in a timely manner. For example, when a new user starts watching a certain content, the system will add an edge between the new user node and the content node on the basis of the existing graph; when the network traffic surges, the edge weight between the network traffic node and the relevant user node will be dynamically adjusted. This dynamic adjustment ensures the timeliness of the graph structure and its ability to reflect real-time situations.

[0080] For the preliminary graph structure, a feature extraction method based on multi-hop relationship reasoning technology is adopted. Combining the correlation relationships of user behaviors, content preferences, and device status, implicit features are extracted. Through the attention technology, the selection of the reasoning path is optimized to generate a preliminary feature representation.

[0081] In this step, the already generated preliminary graph structure will be further processed, and feature extraction is carried out through multi-hop relationship reasoning technology. The multi-hop relationship reasoning technology can mine multi-level correlation relationships between nodes. For example, by jumping from the user behavior node to the content node it watches, and then to other user behavior nodes, possible correlation paths can be inferred. The attention technology is used to optimize these reasoning paths to ensure that the most important correlation relationships are identified and extracted.

[0082] The multi-hop relationship reasoning technology can deeply explore the hidden multi-level correlation features in the data, and the introduction of the attention technology ensures the prominence of the reasoning process and the improvement of accuracy. This method can extract richer and more valuable implicit features, providing high-quality input for subsequent data analysis and visualization. The accuracy of the feature representation directly affects the effect of subsequent analysis and decision-making, so this step is crucial.

[0083] In this step, based on the preliminarily constructed graph structure, feature extraction is carried out through multi-hop relationship reasoning technology. The multi-hop relationship reasoning technology aims to mine multi-level correlation information from the complex network of nodes and edges. For example, a user watches multiple videos, and each video is watched by other users. Such relationships can form an implicit association network among users through the multi-hop reasoning technology starting from the user node, passing through the video node, and then to other user nodes.

[0084] By traversing the preliminary graph structure, the multi-hop relationship reasoning technology calculates the feature representation for each node. During the reasoning process, multi-level connections between multiple nodes are considered. For example, user A watches video X, and user B also watches video X. The system identifies the potential interest relationship between user A and user B through multi-hop reasoning, so that videos liked by user A can be recommended to user B. Each step of reasoning is based on the edges and weights in the graph structure to ensure the accuracy and rationality of the association relationship.

[0085] To further optimize the selection of the inference path, the attention technique is introduced. The attention technique weights each edge during the feature inference process and selects the most important associated path according to the weight of the edge. For example, the behavior of user A watching video X may obtain a higher weight due to a higher click-through rate, while relatively unimportant behaviors obtain lower weights. This weighting process effectively improves the accuracy of the inference path selection through the attention mechanism and ensures that the extracted feature representation reflects the most important associated information in the data.

[0086] For the preliminary feature representation, a feature fusion method based on matrix factorization is adopted to perform weighted fusion on user behavior features, content preference features, and device status features. Through the dynamic weight adjustment technique, the timeliness and consistency of the feature vector are ensured, and a preliminary multi-dimensional feature matrix is generated.

[0087] In this step, the features from different sources are weighted and fused through the matrix factorization method, integrating the features of user behavior, content preference, and device status. The dynamic weight adjustment technique is used to dynamically adjust the weight of each feature during the feature fusion process to ensure that the generated feature vector reflects the most timely and consistent data characteristics. The feature fusion method combines the features of different data dimensions to form a more comprehensive feature matrix, which can better capture the complex dependencies between data. The dynamic weight adjustment technique ensures the timeliness and consistency of the feature matrix, enabling the features to reflect data changes in a timely manner and avoiding problems such as information lag or inconsistency. The finally generated preliminary multi-dimensional feature matrix provides high-quality input for subsequent refined data analysis and visualization.

[0088] After completing the preliminary feature representation, matrix factorization technology is used to fuse these features. The matrix factorization method can decompose complex high-dimensional data into several low-dimensional matrices, enabling features from different sources to be fused in the same space. For example, the singular value decomposition (SVD) technique is used to decompose and reconstruct the user behavior feature matrix, content preference feature matrix, and device status feature matrix.

[0089] First, perform singular value decomposition on each type of feature matrix, decomposing it into several low-dimensional matrices. Then, different weights are assigned to each feature matrix according to the importance of the features. During the matrix fusion process, through weighted reconstruction, these low-dimensional matrices are recombined into a unified feature matrix. For example, during the reconstruction process, the weight of the user behavior feature matrix may be higher because user behavior has a greater impact on content recommendation; the weight of the device status feature matrix is the second because the device status affects the user experience but does not directly affect content preference.

[0090] To ensure the timeliness and consistency of the feature vectors, a dynamic weight adjustment technique is introduced. The dynamic weight adjustment technique dynamically updates the weights of the feature matrix according to the changes in real-time data. For example, when it is found that the user behavior characteristics change significantly during a certain period, the system will automatically increase the weight of the user behavior feature matrix to ensure that the result after feature fusion reflects the latest data changes. Through such dynamic adjustment, it is ensured that the generated multi-dimensional feature matrix can reflect both the stability of historical data and the dynamics of real-time data.

[0091] For the preliminary multi-dimensional feature matrix, an optimization method based on error feedback technology is adopted. Combining the real-time data stream and the historical data distribution, the feature weights are dynamically adjusted, and through regularization constraints, overfitting is prevented to generate the final multi-dimensional feature matrix.

[0092] After the multi-dimensional feature matrix is generated, it is further optimized through error feedback technology. This technology combines the real-time data stream and the historical data distribution to dynamically adjust the feature weights, thereby ensuring that the feature matrix is more accurate and robust. At the same time, to prevent the overfitting problem of the model, regularization constraints are introduced to maintain the robustness during the optimization process of the feature matrix. Through optimization with error feedback technology, the feature matrix can be continuously iterated to make it closer and closer to the distribution and characteristics of the actual data. The process of dynamically adjusting the feature weights ensures the flexibility and adaptability of the feature matrix in the face of real-time data changes. The introduction of regularization constraints effectively prevents overfitting, improves the generalization ability of the model when dealing with new data, and ensures the stable performance of the feature matrix in different data scenarios.

[0093] After the preliminary multi-dimensional feature matrix is generated, it is further optimized using error feedback technology. The error feedback technology continuously adjusts the feature weights to reduce the error by calculating the error between the model prediction value and the actual value. For example, if the error is large when the model makes a prediction using the current feature matrix, the system will adjust the feature weights according to the error feedback to make the feature matrix better adapt to the actual data.

[0094] The optimization process combines the real-time data stream and the historical data distribution to dynamically adjust the feature matrix. The real-time data stream provides the latest data change information, and the historical data distribution provides the long-term trend and stable characteristics of the data. By combining these two aspects of information, the system can ensure that the dynamic adjustment of the feature weights can not only reflect the latest data changes in a timely manner but also not deviate from the laws of the long-term data distribution. For example, if the real-time data stream shows that there are obvious changes in the user viewing behavior and there is a similar trend in the historical data, the system will adjust the weights of the user behavior characteristics to ensure the accuracy of the prediction model.

[0095] In order to prevent the model from overfitting during the optimization process, regularization constraints are introduced. For example, through the L2 regularization technology, the feature weights are constrained to prevent the model from overfitting due to excessive or too small feature weights. Regularization technology suppresses extreme changes in feature weights by adding penalty terms, making the optimized feature matrix more robust. For example, when certain features appear frequently in historical data but do not appear in real-time data, regularization constraints can prevent these features from having too large weights in the model, resulting in overfitting. Finally, after error feedback technology optimization and regularization constraints, the generated multidimensional feature matrix can accurately reflect the timeliness and consistency of the data, while having good generalization capabilities.

[0096] S203, inputting the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart, wherein the visualization model optimizes the visual expression and information density of the chart by combining attention technology and an adaptive layout algorithm to obtain a preliminary optimized visualization design scheme;

[0097] The Generative Adversarial Network (GAN) consists of two parts: the generator and the discriminator. The generator generates visual charts, and the discriminator evaluates them and provides feedback for optimization. Attention technology plays a key role in this process, highlighting the data points and key information that users are concerned about. The adaptive layout algorithm automatically adjusts the layout according to the content of the chart to achieve the best information density and visual effect. Finally, through continuous iteration and optimization, the best preliminary visualization design solution is generated.

[0098] Using generative adversarial networks for data visualization can not only generate high-quality visualization charts, but also continuously optimize the generated results through the feedback of the discriminator. Combined with attention technology, it can effectively highlight key information and data points, and improve the readability and practicality of charts. The adaptive layout algorithm ensures the reasonable layout of chart content and the optimization of information density, allowing users to obtain more valuable information in a limited space. This method not only improves the presentation of global data, but also enhances the user experience and efficiency in data analysis, and ultimately improves the ability of decision support.

[0099] Specifically, a visualization model based on a generative adversarial network can be used for a multi-dimensional feature matrix, combining the characteristics of user behavior, content preference, and device status to generate an initial visualization chart. The accuracy and readability of the chart can be ensured through conditional constraint technology to generate a preliminary visualization chart.

[0100] In this step, the multi-dimensional feature matrix is input into the visualization model of the generative adversarial network (GAN). The GAN model consists of a generator and a discriminator. The generator generates an initial visualization chart based on the input features, and the discriminator evaluates the generated chart to ensure its accuracy and readability. Here, conditional constraint technology is introduced. According to user behavior, content preferences, and device status features, specific constraints are imposed to ensure that the generated chart meets the actual business requirements and display standards.

[0101] Visualization chart generation through the GAN model can give full play to the synergistic effect of the generator and the discriminator to generate high-quality charts. Conditional constraint technology ensures that the charts are not only visually appealing but also accurately reflect the actual situation in terms of data. This is extremely important for the smart radio and television project because it can ensure that the information presented to users is reliable and accurate, thereby enhancing users' trust in the system and satisfaction with its use.

[0102] In this step, first, the data to be input into the visualization model of the generative adversarial network (GAN) needs to be prepared. These data come from the multi-dimensional feature matrix extracted and fused in the previous step. The multi-dimensional feature matrix contains information in different dimensions such as user behavior features, content preference features, and device status features. This information will be input into the GAN model, where the generator is responsible for generating the initial visualization chart, and the discriminator is used to evaluate the chart quality. To ensure the reliability and practicality of the generated chart, conditional constraint technology is introduced during the generation process, and these constraints are set according to the specific requirements of user behavior data, content preferences, and device status features.

[0103] Based on the input multi-dimensional feature matrix, the generator generates an initial visualization chart through a series of convolutional and deconvolutional operations. For example, for user behavior data, a heatmap can be generated to show the distribution of user behavior; for content preference data, a bar chart can be generated to show the popularity of different contents; for device status data, a line chart can be generated to show the trend of device status changes. These charts initially show the relationships among user behavior, content preferences, and device status.

[0104] While generating the initial chart, conditional constraint technology ensures the accuracy and readability of the chart. For example, for user behavior data, set constraint conditions to ensure that high-frequency behaviors and key events are prominently displayed in the chart; for content preference data, ensure that popular contents are placed in prominent positions; for device status data, ensure that abnormal states are prominently marked. The discriminator evaluates the generated chart, checks whether it meets the requirements according to the preset conditional constraints, and continuously provides feedback and adjusts the generator until a high-quality chart that meets the conditions is generated. Finally, the generated initial visualization chart not only accurately reflects the data but also has good readability.

[0105] For the preliminary visualization chart, an optimization method based on attention technology is adopted. Combining the key information that users are concerned about, the visual expression of the chart is dynamically adjusted. Through multi-scale attention technology, information density at different levels is captured to generate a preliminarily optimized visualization chart.

[0106] In this step, the preliminarily generated visualization chart will be further optimized by attention technology. Attention technology is used to identify and highlight the key information that users are concerned about, which can dynamically adjust the visual expression of the chart, making the chart not only beautiful but also clearer in information transmission. Through multi-scale attention technology, the density of information at different levels can be captured. Different levels of information include global trends and detailed data, ensuring that while the overall trend is shown in the chart, important detailed information is not overlooked.

[0107] The application of attention technology greatly improves the practicality and operability of the chart, enabling users to quickly obtain the most critical information. The multi-scale attention mechanism ensures clear levels of information expression, comprehensively and accurately displaying data characteristics. Through this optimization method, users can more easily focus on important content when viewing the chart, avoiding information redundancy and over-concentration, thus improving the efficiency of data analysis and decision-making.

[0108] After obtaining the preliminary visualization chart, attention technology is used for further optimization. Attention technology aims to highlight the key information that users care about. By dynamically adjusting the visual elements in the chart, users can quickly identify and obtain key information. In this step, first, the information in the preliminary chart is analyzed to identify the key data points and information levels that users may care about.

[0109] Specifically, the attention mechanism can be used to perform multi-scale processing on the chart. For example, for a user behavior heat map, through attention technology, the areas where users are concentrated in activities are identified and highlighted, while other less important areas are relatively faded. For a bar chart of content preferences, through attention technology, the categories of content that users most often view are shown in brighter colors, and other content is distinguished with relatively dull colors. For a line chart of device status, through attention technology, the parts where the device has abnormal fluctuations are highlighted and marked with different symbols.

[0110] The multi-scale attention technology plays an important role here, as it can capture information densities at different levels. For example, globally, the user behavior heatmap shows the overall behavior distribution trend; while from a local perspective, the attention mechanism can magnify the specific behavior details within a certain time period. For instance, during a peak period, the user's behavior characteristics change significantly, and the attention technology will focus on showing the detailed information of this time period to help users better understand the data changes. After being optimized by the attention technology, the finally generated chart is more visually appealing, and at the same time, the information expression is clearer and more hierarchical.

[0111] For the preliminarily optimized visualization chart, an adjustment method based on the adaptive layout algorithm is adopted. Combining the visual expression and information density of the chart, the layout of the chart is dynamically adjusted. Through the dynamic weight allocation technology, the visual effect and information transmission efficiency of the chart are optimized to generate a preliminarily optimized visualization design scheme;

[0112] After generating the optimized visualization chart, its layout is adjusted through the adaptive layout algorithm. The adaptive layout algorithm dynamically adjusts the layout according to the information density and visual expression in the chart to ensure that the chart content is reasonably distributed in space, avoiding information being too dense or sparse. The dynamic weight allocation technology plays a key role in this process. It allocates visual resources according to the importance and density of different information, optimizes the visual effect and information transmission efficiency of the chart, and generates a preliminarily optimized visualization design scheme.

[0113] The application of the adaptive layout algorithm can effectively improve the readability and information transmission effect of the chart. By dynamically adjusting the layout, it ensures that the important information in the chart can be reasonably displayed, enabling users to obtain key information more quickly and avoiding information overload or omission. The dynamic weight allocation technology further optimizes the resource allocation of the chart, enabling important data to obtain more display space and higher readability, thereby improving the user experience and usage efficiency.

[0114] In this step, the preliminarily optimized visualization chart will be further adjusted through the adaptive layout algorithm. The adaptive layout algorithm dynamically adjusts the layout according to the visual expression and information density of each information unit in the chart to ensure that the chart content is reasonably distributed, neither too dense nor too sparse. First, each information unit of the chart is analyzed to determine the priority and importance of each unit.

[0115] Dynamic weight allocation technology is crucial in this process. According to the weights of each information unit, the adaptive layout algorithm will automatically adjust the layout of the chart. For example, in a content playback distribution chart, if the playback frequencies of certain types of content are very high, these categories will be allocated larger display areas and more prominent positions. In the user behavior heat map, the areas with frequent user activities will be enlarged, while the areas with less activities will be compressed. In the device status line chart, the time periods with large fluctuations will be highlighted, while the time periods with stable operation will have reduced display areas.

[0116] During the adaptive layout process, it is also necessary to consider the overall visual effect and information transmission efficiency of the chart. For example, in a bar chart, the width and color of the bars need to be adjusted according to the data volume to ensure that users can identify important data at a glance. In a line chart, visual elements such as the thickness, color, and marker points of the lines also need to be adjusted according to the data fluctuations. Through the adaptive layout algorithm, the overall layout of the chart is optimized, and the information density is reasonably distributed, enabling users to obtain the most important information in the shortest time. The final generated visualization design plan reaches the best state in terms of visual effect and information transmission efficiency.

[0117] For the preliminarily optimized visualization design plan, a verification method based on user feedback is adopted. Combining user interaction data and real-time performance monitoring, the design plan is dynamically adjusted, and through feedback correction technology, the final preliminarily optimized visualization design plan is generated.

[0118] After generating the preliminarily optimized visualization design plan, it is verified through user feedback. The input of user feedback is crucial. By analyzing the data and opinions left by users during the interaction process and combining real-time performance monitoring, the design plan is dynamically adjusted. Feedback correction technology is used to optimize according to user feedback to ensure that the design plan of the chart can meet user needs and improve user satisfaction.

[0119] The application of user feedback in visualization design ensures that the design plan can conform to actual needs and usage habits, further enhancing the practicality and intelligence of the chart. Through real-time performance monitoring, problems or deficiencies in the design can be discovered and corrected in a timely manner to ensure the stability and efficiency of the system. The application of feedback correction technology enables the optimization plan to be rapidly iterated, forming a continuous improvement closed-loop process, greatly enhancing the user experience and data analysis effect.

[0120] In this step, the preliminarily optimized visualization design plan needs to be verified through user feedback. The collection and analysis of user feedback are key steps. By combining the operation data and opinions left by users during the interaction process and real-time performance monitoring, the design plan is dynamically adjusted. First, a user feedback collection system is established to obtain the real experiences and opinions of users during the operation process by recording user interaction behaviors such as clicks, swipes, and dwell times.

[0121] Specifically, when users use visual charts, they may have different interaction requirements and preferences. For example, when viewing a device status chart, users may be more interested in the time periods of abnormal fluctuations in the device; when viewing a user behavior heatmap, they may be more concerned about the behavior distribution during peak hours. By analyzing this interaction data, the data areas that users are most concerned about and usage habits can be identified. Combining with real-time performance monitoring, such as chart loading speed, interaction response time, etc., to understand the performance of the design solution in actual use.

[0122] The feedback correction technology is used to dynamically adjust the design solution according to the collected user feedback and performance monitoring data. For example, if users feedback that some information display is not intuitive enough, the system will adjust the information display method, such as improving intuitiveness by changing colors, chart types or re-layout. If performance monitoring finds that some charts load slowly, the system will optimize the chart loading logic or reduce the complexity of the chart to improve the response speed. Through multiple feedbacks and corrections, the design solution is continuously iterated to ensure that the finally generated visual design solution can meet user needs to the greatest extent, providing the best user experience and data analysis effect.

[0123] S204, deploy the visual design solution to an interactive visualization platform to support users to explore data through multi-dimensional filtering and dynamic interaction operations, where the interactive visualization platform uses streaming computing and incremental update technologies to update the visual content in real time, ensuring the timeliness and operability of the data, and obtaining the final intelligent radio and television big data visual display solution.

[0124] During the deployment process, it is necessary to fully consider the user's needs and technical architecture to ensure that the design solution can run efficiently and stably on the platform. Through the real-time update mechanism and dynamic interaction function, users can flexibly operate and deeply explore the data. Deploying the visual design solution to an interactive visualization platform can greatly improve the user's operation and analysis efficiency of the data. The multi-dimensional filtering and dynamic interaction operation functions enable users to explore the data according to their own needs and discover potential patterns and trends. At the same time, through streaming computing and incremental update technologies, it is ensured that the visual content always reflects the latest data changes and analysis results. This real-time nature and interactivity provide great convenience and value for users, improving the accuracy and timeliness of data decision-making, and ultimately contributing to the successful implementation and continuous optimization of the intelligent radio and television project.

[0125] Specifically, for the preliminarily optimized visual design solution, a deployment method based on the interactive visualization platform can be adopted. Combining the user's needs and technical architecture, the design solution is deployed to the platform. Through dynamic loading technology, the efficiency and stability of the deployment are ensured, and a preliminary visual display solution is generated;

[0126] In this step, the preliminarily optimized visual design solution is deployed to the interactive visualization platform. During the deployment process, it is necessary to combine user requirements and the technical architecture of the system to ensure the efficient and stable operation of the design solution on the platform. The dynamic loading technology plays a key role here, enabling the visual content to be loaded on demand, reducing the system loading time and resource consumption, thereby generating a preliminary visual display solution. By adopting the dynamic loading technology, the deployment of the visual design solution not only improves the loading efficiency of the system but also enhances the user's operation experience. Deploying in combination with user requirements ensures that users can quickly access and operate the data views they care most about, while ensuring the stability and efficiency of the system. This flexible and efficient deployment method enables the visualization platform to quickly respond to user requirements and data changes, improving the practicality of the system and user satisfaction.

[0127] During the deployment process of the preliminarily optimized visual design solution, it is necessary to comprehensively evaluate user requirements and the system technical architecture first. The evaluation of user requirements includes understanding the specific requirements of users for data visualization, such as the focus of data display, preferences for interaction methods, and performance requirements, etc. The evaluation of the technical architecture involves the existing hardware and software environment to ensure that the deployment solution can operate efficiently on the existing platform. Through this comprehensive evaluation, it is ensured that the design of the deployment solution can not only meet user requirements but also be well compatible with the existing technical architecture.

[0128] Secondly, the dynamic loading technology is adopted for the deployment of the visual design solution. The core of the dynamic loading technology is to load data and view modules only when needed, rather than loading all content at once. For example, when a user first accesses the platform, only the main page and basic data views are loaded. When the user clicks on a specific chart or selects different data dimensions, the corresponding views and data will be dynamically loaded into the page. This on-demand loading method can significantly reduce the initial loading time of the system and improve the user's first-use experience.

[0129] Finally, ensure the efficiency and stability of the deployment. The dynamic loading technology not only reduces the initial load of the system but also helps to disperse the subsequent computing pressure, thereby improving the response speed and stability of the entire platform. Special attention needs to be paid to the data transmission efficiency between the server and the client. By optimizing the data interface and compressing the amount of transmitted data, the performance of the platform can be further improved. For example, play data, user behavior data, and device status data are transmitted through different interfaces respectively to ensure the load balance of each interface. The finally generated preliminary visual display solution achieves the expected results in terms of efficiency and stability, and users can quickly and smoothly browse and analyze data.

[0130] For the preliminary visualization display scheme, an interactive operation support method based on multi-dimensional screening is adopted. The screening conditions are adjusted dynamically by combining user behavior data and content preferences. Through the interactive operation interface, users are supported to explore data by clicking and dragging, and a preliminary interactive visualization scheme is generated.

[0131] In this step, the preliminary visualization display solution provides the function of dynamically adjusting the filtering conditions by supporting interactive operation methods of multi-dimensional filtering, combining user behavior data and content preferences. Through the interactive operation interface, users can filter and explore data by clicking and dragging, and deeply explore the potential relationships and trends behind the data to generate a preliminary interactive visualization solution. The interactive operation and multi-dimensional filtering functions greatly improve the flexibility and depth of user data analysis. Users can flexibly adjust the filtering conditions according to their own needs and interests, and quickly locate and analyze the data of interest. This helps users to explore data and make decisions more comprehensively and efficiently. Such an operation experience not only improves user satisfaction, but also enhances the practicality and user stickiness of the platform.

[0132] In implementing the interactive operation method of multi-dimensional screening, it is first necessary to design and implement a friendly and feature-rich interactive operation interface. This interface allows users to filter and explore data through interactive operations such as clicking, dragging, and sliding. For example, various screening and filtering options can be provided on the interface, such as time range selectors, content type filters, user behavior type selectors, etc. Users can flexibly adjust the screening conditions through these controls. This design not only improves the convenience of operation, but also enhances the user's interactive experience.

[0133] Secondly, dynamically adjust the filtering conditions based on user behavior data and content preferences. The interactive operation interface needs to provide intelligent filtering suggestions based on the user's known behaviors and preferences. For example, when users frequently view certain types of content, the system can automatically put these content types at the top of the filter to facilitate users to quickly select. At the same time, through data analysis and machine learning algorithms, the system can dynamically recommend filtering conditions that may be of interest. For example, if a user frequently views data within a certain time period, the system will automatically set that time period as the default filtering condition.

[0134] Finally, optimize the data display and update process after filtering. When users filter data, the system needs to respond and update the display content in real time, which is the key to achieving a smooth interaction experience. For example, when users adjust the time range, the platform should immediately reload and display the play data, user behavior data, and device status data within the corresponding time period. Through incremental update technology, only the data within the filtered range is loaded and displayed, avoiding the reloading of all data and improving the update speed and efficiency. For example, when users drag the time slider, the system dynamically updates the chart data displayed on the right, thus quickly reflecting user operations. Through these optimizations, the initial interactive visualization solution ensures that users can efficiently and conveniently filter and explore data.

[0135] For the initial interactive visualization solution, adopt a real-time update method based on streaming computing and incremental update technology. Combine real-time data streams and historical data distributions to dynamically update the visualization content. Through incremental update technology, ensure the timeliness and operability of the data, and generate a preliminary real-time update visualization solution;

[0136] In this step, the initial interactive visualization solution is updated in real time through streaming computing and incremental update technology. Streaming computing technology can process real-time data streams and dynamically update the visualization content to ensure the timeliness of the displayed data. Incremental update technology, on the other hand, performs partial updates by comparing new and old data, reducing system resource consumption and improving update efficiency, thus generating a preliminary real-time update visualization solution.

[0137] The adoption of streaming computing and incremental update technology ensures the real-time and accuracy of data visualization content, improving the efficiency and experience of users in data analysis. The dynamic update of real-time data keeps the visualization content always up-to-date, enabling users to obtain the latest data analysis results and trends in a timely manner. This technology guarantees the efficiency and operability of the system, enhancing the timeliness and accuracy of data decision-making.

[0138] In this process, first, a streaming computing framework needs to be built to process real-time data streams. The streaming computing framework can receive and process in real time the data streams transmitted from various data sources, such as user behavior data, play data, and device status data. Streaming computing technology can ensure that data is processed and displayed immediately when it enters the system, thus ensuring the real-time nature of the visualization content. For example, the user behavior heat map is refreshed every few seconds to show the latest user activities.

[0139] Secondly, incremental update technology is applied to optimize the efficiency of data update. The core of incremental update technology lies in only processing and updating the changed parts of the data, rather than reloading and calculating the entire data volume. For example, when a new batch of playback data flows in, the system only needs to update the relevant data modules and display parts, while the other irrelevant modules and display parts remain unchanged. Such a design can significantly reduce the computational burden of the system and improve the efficiency of data update. Specifically, if a user behavior heatmap shows user behaviors within a certain time period, when new user data flows in, the system only needs to incrementally update the new data points in the map instead of redrawing the entire map.

[0140] Finally, by combining real-time data streams and historical data distributions, the visualization content is dynamically adjusted. By analyzing real-time data streams and historical data distributions, the system can intelligently predict and adjust the display method of the visualization content. For example, in the bar chart of playback data, if the system detects a sudden surge in the playback volume of a certain type of content, it can dynamically adjust the display ratio of the bar chart to highlight the changes. At the same time, in the line chart of device status, by combining historical data distributions, it can predict future device status changes and mark possible abnormal fluctuations in the chart in advance. Through these methods, the real-time updated visualization solution can not only display the latest data but also provide valuable predictive information to help users better understand and analyze the data.

[0141] For the preliminary real-time update visualization solution, an optimization method based on user feedback is adopted. By combining user interaction data and real-time performance monitoring, the visualization content is dynamically adjusted. Through feedback correction technology, the final intelligent radio and television big data visualization display solution is generated.

[0142] In this step, the preliminary real-time update visualization solution is further optimized through the optimization method based on user feedback. By combining user interaction data and real-time performance monitoring, the visualization content is dynamically adjusted. The feedback correction technology continuously adjusts and optimizes the display content according to the actual usage situation and feedback of users to ensure that the finally generated intelligent radio and television big data visualization display solution meets user needs and provides the best user experience.

[0143] The introduction of user feedback ensures that the visualization display solution can truly meet the needs and usage habits of users. Through dynamic adjustment and feedback correction technology, the quality and user experience of the visualization content can be continuously optimized and improved. This continuous improvement process not only improves user satisfaction but also enhances the intelligence and adaptability of the platform, enabling the system to respond to changes in user needs and dynamic changes in data at any time.

[0144] To further optimize the preliminary real-time update visualization solution, it is first necessary to establish a user feedback collection system to record all user interaction behaviors and feedback opinions. User interaction data includes information such as click frequency, dwell time, and operation path. This data can help analyze the real needs and pain points of users during use. For example, if a user stays on a certain part of the chart for a long time and frequently operates, it indicates that this part of the data is highly important to the user and needs to be optimized and improved with key emphasis.

[0145] Secondly, combined with real-time performance monitoring, analyze the actual performance and existing problems of the system. Performance monitoring includes indicators such as system response speed, data loading time, and interaction fluency. Through real-time monitoring, bottlenecks in the actual operation of the system can be discovered and solved. For example, if the loading time of a certain play data chart is too long and it is found through monitoring analysis that it is caused by a large amount of data, the data loading method and chart rendering algorithm can be optimized to improve the display efficiency and response speed.

[0146] Finally, based on user feedback and performance monitoring results, use feedback correction technology to dynamically adjust the visualization content. For example, by analyzing user feedback, it is found that users are more concerned about the play data of a certain type of content. The bar chart can be redesigned to highlight the data display of this type of content. At the same time, through performance monitoring data, it is found that the response speed of some charts is slow. It may be optimized through technical means such as data sharding loading and asynchronous processing to improve the overall system fluency and user experience. After multiple rounds of feedback, correction, and optimization, the finally generated intelligent radio and television big data visualization display solution can best meet user needs and reach the best state in terms of performance and operability. For example, users can smoothly perform various filtering, operations, and data exploration during use, the system quickly responds and updates data display in real time, providing the best data analysis and decision-making support experience.

[0147] It can be seen that according to the user behavior data, content play data, network traffic data, and device status data of the intelligent radio and television project, multi-source data collection is carried out to obtain a cleaned data set; the cleaned data set is input into a feature extraction model to extract the correlation features of user behavior, content preferences, and device status, obtaining a multi-dimensional feature matrix; the multi-dimensional feature matrix is input into a visualization model to generate an initial visualization chart, obtaining a preliminary optimized visualization design scheme; the visualization design scheme is deployed to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations, obtaining the final intelligent radio and television big data visualization display solution, thereby being able to improve the processing efficiency and display effect of radio and television big data.

[0148] Another embodiment of the present invention provides a big data visualization processing system for an intelligent radio and television project. Refer to Figure 3 and the system may include:

[0149] The acquisition module 301 is used to perform multi-source data acquisition based on the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project. It adopts a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm to identify and eliminate noise data and redundant information in real time, and obtains a cleaned data set.

[0150] The extraction module 302 is used to input the cleaned data set into a feature extraction model based on a graph neural network to extract the correlation features of user behavior, content preference, and device status. Among them, the feature extraction model captures the implicit correlation relationship between data through dynamic graph construction and multi-hop relationship reasoning technology, and obtains a multi-dimensional feature matrix.

[0151] The visualization module 303 is used to input the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart through a combination of attention technology and an adaptive layout algorithm, and obtains a preliminary optimized visualization design scheme.

[0152] The deployment module 304 is used to deploy the visualization design scheme to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations. Among them, the interactive visualization platform updates the visualization content in real time through streaming computing and incremental update technology to ensure the timeliness and operability of the data, and obtains a final intelligent radio and television big data visualization display scheme.

[0153] It can be seen that based on the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project, multi-source data acquisition is performed to obtain a cleaned data set; the cleaned data set is input into a feature extraction model to extract the correlation features of user behavior, content preference, and device status, and a multi-dimensional feature matrix is obtained; the multi-dimensional feature matrix is input into a visualization model to generate an initial visualization chart, and a preliminary optimized visualization design scheme is obtained; the visualization design scheme is deployed to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations, and a final intelligent radio and television big data visualization display scheme is obtained, thereby being able to improve the processing efficiency and display effect of radio and television big data.

[0154] An embodiment of the present invention further provides a storage medium, in which a computer program is stored, and the computer program is set to execute the steps in any one of the above method embodiments when running.

[0155] Specifically, in this embodiment, the above storage medium can be set to store a computer program for executing the following steps:

[0156] S201, Collect multi-source data based on the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project. Adopt a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm to identify and eliminate noise data and redundant information in real time, and obtain a cleaned data set.

[0157] S202, Input the cleaned data set into a feature extraction model based on a graph neural network to extract the associated features of user behavior, content preferences, and device status. Among them, the feature extraction model captures the implicit association relationships between data through dynamic graph construction and multi-hop relationship reasoning technology, and obtains a multi-dimensional feature matrix.

[0158] S203, Input the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart by combining attention technology and an adaptive layout algorithm, and obtains a preliminary optimized visualization design scheme.

[0159] S204, Deploy the visualization design scheme to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations. Among them, the interactive visualization platform updates the visualization content in real time through streaming computing and incremental update technology to ensure the timeliness and operability of the data, and obtains a final intelligent radio and television big data visualization display scheme.

[0160] It can be seen that, based on the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project, multi-source data is collected to obtain a cleaned data set; the cleaned data set is input into a feature extraction model to extract the associated features of user behavior, content preferences, and device status, and a multi-dimensional feature matrix is obtained; the multi-dimensional feature matrix is input into a visualization model to generate an initial visualization chart, and a preliminary optimized visualization design scheme is obtained; the visualization design scheme is deployed to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations, and a final intelligent radio and television big data visualization display scheme is obtained, thereby improving the processing efficiency and display effect of radio and television big data.

[0161] An embodiment of the present invention further provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0162] Specifically, the above electronic device may further include a transmission device and an input / output device. Among them, the transmission device is connected to the above processor, and the input / output device is connected to the above processor.

[0163] Specifically, in this embodiment, the above-mentioned processor can be set to execute the following steps through a computer program:

[0164] S201, perform multi-source data collection based on the user behavior data, content playback data, network traffic data, and device status data of the smart radio and television project, and use a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm to identify and eliminate noise data and redundant information in real time, obtaining a cleaned data set;

[0165] S202, input the cleaned data set into a feature extraction model based on a graph neural network to extract the associated features of user behavior, content preference, and device status. Among them, the feature extraction model captures the implicit association relationships between data through dynamic graph construction and multi-hop relationship reasoning technology, obtaining a multi-dimensional feature matrix;

[0166] S203, input the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart through a combination of attention technology and an adaptive layout algorithm, obtaining a preliminary optimized visualization design scheme;

[0167] S204, deploy the visualization design scheme to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations. Among them, the interactive visualization platform updates the visualization content in real time through streaming computing and incremental update technology to ensure the timeliness and operability of the data, obtaining a final smart radio and television big data visualization display scheme.

[0168] It can be seen that multi-source data collection is performed based on the user behavior data, content playback data, network traffic data, and device status data of the smart radio and television project to obtain a cleaned data set; the cleaned data set is input into a feature extraction model to extract the associated features of user behavior, content preference, and device status, obtaining a multi-dimensional feature matrix; the multi-dimensional feature matrix is input into a visualization model to generate an initial visualization chart, obtaining a preliminary optimized visualization design scheme; the visualization design scheme is deployed to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations, obtaining a final smart radio and television big data visualization display scheme, thereby being able to improve the processing efficiency and display effect of radio and television big data.

[0169] The above has detailed the structure, features, and function effects of the present invention according to the illustrated embodiments. The above is only the preferred embodiment of the present invention, but the present invention is not limited to the scope defined by the drawings. Any changes made according to the concept of the present invention, or equivalent embodiments modified into equivalent changes, still within the spirit covered by the specification and the drawings, shall be within the protection scope of the present invention.

Claims

1. A big data visualization processing method for a smart radio and television project, characterized in that The method includes: Performing multi-source data collection based on the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project, and using a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm to identify and eliminate noise data and redundant information in real time, obtaining a cleaned data set; Inputting the cleaned data set into a feature extraction model based on a graph neural network to extract the associated features of user behavior, content preference, and device status. Among them, the feature extraction model captures the implicit association relationships between data through dynamic graph construction and multi-hop relationship reasoning technologies, obtaining a multi-dimensional feature matrix. For the cleaned data set, a feature extraction model based on a graph neural network is used to abstract users, content, network traffic, and devices as nodes in a graph structure, and the association relationships between nodes are abstracted as edges. Through dynamic graph construction technology, a preliminary graph structure is generated. For the preliminary graph structure, a feature extraction method based on multi-hop relationship reasoning technology is used to combine the association relationships of user behavior, content preference, and device status to extract implicit features, and through attention technology, the selection of the reasoning path is optimized to generate a preliminary feature representation; For the preliminary feature representation, a feature fusion method based on matrix factorization is used to perform weighted fusion of user behavior features, content preference features, and device status features. Through dynamic weight adjustment technology, the timeliness and consistency of the feature vectors are ensured, generating a preliminary multi-dimensional feature matrix. For the preliminary multi-dimensional feature matrix, an optimization method based on error feedback technology is used to combine real-time data streams and historical data distributions to dynamically adjust feature weights. Through regularization constraints, overfitting is prevented, generating a final multi-dimensional feature matrix; Inputting the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart by combining attention technology and an adaptive layout algorithm, obtaining a preliminary optimized visualization design scheme; Deploying the visualization design scheme to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations. Among them, the interactive visualization platform updates the visualization content in real time through streaming computing and incremental update technologies, ensuring the timeliness and operability of the data, obtaining a final intelligent radio and television big data visualization display scheme.

2. The method according to claim 1, characterized in that, The performing multi-source data collection based on the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project, and using a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm to identify and eliminate noise data and redundant information in real time, obtaining a cleaned data set, includes: According to the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project, using a data collection framework based on edge computing to obtain multi-source data in real time, and through lightweight data caching technology, ensuring the real-time and continuous data collection; For the multi-source data collected, an adaptive rule engine-based cleaning method is adopted. Combining with the preset business rules, it dynamically identifies and eliminates the data that does not conform to the rules. Through the rule dynamic update technology, it ensures the timeliness and accuracy of the cleaning rules and generates a preliminary cleaned dataset; For the preliminary cleaned dataset, a noise elimination method based on anomaly detection algorithms is adopted. Combining with the historical data distribution and real-time data stream, it identifies and eliminates noise data and redundant information. Through the dynamic threshold adjustment technology, it ensures the accuracy and robustness of anomaly detection and generates a preliminary denoised dataset; For the preliminary denoised dataset, a method based on data integration algorithms is adopted to map the multi-source data into a unified data structure. Through the data consistency verification technology, it ensures the integrity and consistency of the data and generates the final cleaned dataset.

3. The method according to claim 2, wherein Inputting the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart by combining attention technology and an adaptive layout algorithm, and obtains a preliminary optimized visualization design scheme, including: For the multi-dimensional feature matrix, a visualization model based on a generative adversarial network is adopted. Combining with the characteristics of user behavior, content preference, and device status, it generates an initial visualization chart. Through conditional constraint technology, it ensures the accuracy and readability of the chart and generates a preliminary visualization chart; For the preliminary visualization chart, an optimization method based on attention technology is adopted. Combining with the key information that users focus on, it dynamically adjusts the visual expression of the chart. Through multi-scale attention technology, it captures information densities at different levels and generates a preliminary optimized visualization chart; For the preliminary optimized visualization chart, an adjustment method based on an adaptive layout algorithm is adopted. Combining with the visual expression and information density of the chart, it dynamically adjusts the layout of the chart. Through the dynamic weight allocation technology, it optimizes the visual effect and information transmission efficiency of the chart and generates a preliminary optimized visualization design scheme; For the preliminary optimized visualization design scheme, a verification method based on user feedback is adopted. Combining with user interaction data and real-time performance monitoring, it dynamically adjusts the design scheme. Through the feedback correction technology, it generates the final preliminary optimized visualization design scheme.

4. The method according to claim 3, wherein Deploying the visualization design scheme to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations. Among them, the interactive visualization platform updates the visualization content in real time through streaming computing and incremental update technology to ensure the timeliness and operability of the data and obtains the final intelligent radio and television big data visualization display scheme, including: For the preliminary optimized visualization design scheme, a deployment method based on an interactive visualization platform is adopted. Combining with user requirements and technical architecture, it deploys the design scheme to the platform. Through the dynamic loading technology, it ensures the efficiency and stability of the deployment and generates a preliminary visualization display scheme; For the preliminary visual display solution, an interactive operation support method based on multi-dimensional screening is adopted. Combining user behavior data and content preferences, the screening conditions are dynamically adjusted. Through an interactive operation interface, users are supported to explore data through click and drag operations, and a preliminary interactive visualization solution is generated; For the preliminary interactive visualization solution, a real-time update method based on stream computing and incremental update technology is adopted. Combining real-time data streams and historical data distributions, the visualization content is dynamically updated. Through incremental update technology, the timeliness and operability of the data are ensured, and a preliminary real-time update visualization solution is generated; For the preliminary real-time update visualization solution, an optimization method based on user feedback is adopted. Combining user interaction data and real-time performance monitoring, the visualization content is dynamically adjusted. Through feedback correction technology, a final intelligent radio and television big data visualization display solution is generated.

5. A big data visualization processing system for a smart radio and television project, characterized in that, The system includes: A collection module for performing multi-source data collection according to the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project. Using a hybrid technology based on an adaptive rule engine and an anomaly detection algorithm, noise data and redundant information are identified and removed in real time to obtain a cleaned data set; An extraction module for inputting the cleaned data set into a feature extraction model based on a graph neural network to extract the correlation features of user behavior, content preference, and device status. Among them, the feature extraction model captures the implicit correlation relationships between data through dynamic graph construction and multi-hop relationship reasoning technologies to obtain a multi-dimensional feature matrix; for the cleaned data set, a feature extraction model based on a graph neural network is used to abstract users, content, network traffic, and devices as nodes in a graph structure, and the correlation relationships between nodes are abstracted as edges. Through dynamic graph construction technology, a preliminary graph structure is generated; for the preliminary graph structure, a feature extraction method based on multi-hop relationship reasoning technology is used to combine the correlation relationships of user behavior, content preference, and device status to extract implicit features. Through attention technology, the selection of the reasoning path is optimized to generate a preliminary feature representation; For the preliminary feature representation, a feature fusion method based on matrix decomposition is adopted to perform weighted fusion of user behavior features, content preference features, and device status features. Through dynamic weight adjustment technology, the timeliness and consistency of the feature vector are ensured to generate a preliminary multi-dimensional feature matrix; for the preliminary multi-dimensional feature matrix, an optimization method based on error feedback technology is used to combine real-time data streams and historical data distributions to dynamically adjust the feature weights. Through regularization constraints, overfitting is prevented to generate a final multi-dimensional feature matrix; A visualization module for inputting the multi-dimensional feature matrix into a visualization model based on a generative adversarial network to generate an initial visualization chart. Among them, the visualization model optimizes the visual expression and information density of the chart by combining attention technology and an adaptive layout algorithm to obtain a preliminary optimized visualization design solution; A deployment module for deploying the visual design solution to an interactive visualization platform to support users in exploring data through multi-dimensional filtering and dynamic interaction operations. The interactive visualization platform uses streaming computing and incremental update technologies to update the visualization content in real time, ensuring the timeliness and operability of the data, and obtaining the final intelligent radio and television big data visualization display solution.

6. The system according to claim 5, wherein The acquisition module is specifically configured to: According to the user behavior data, content playback data, network traffic data, and device status data of the intelligent radio and television project, use a data acquisition framework based on edge computing to obtain multi-source data in real time, and ensure the real-time and continuous data acquisition through lightweight data caching technology; For the acquired multi-source data, use a cleaning method based on an adaptive rule engine, combine preset business rules, dynamically identify and eliminate data that does not conform to the rules, and ensure the timeliness and accuracy of the cleaning rules through rule dynamic update technology to generate a preliminary cleaned data set; For the preliminary cleaned data set, use a noise elimination method based on an anomaly detection algorithm, combine the historical data distribution and real-time data stream, identify and eliminate noise data and redundant information, and ensure the accuracy and robustness of anomaly detection through dynamic threshold adjustment technology to generate a preliminary denoised data set; For the preliminary denoised data set, use a method based on a data integration algorithm to map the multi-source data into a unified data structure, and ensure the integrity and consistency of the data through data consistency verification technology to generate the final cleaned data set.

7. A storage medium, characterized in that, A computer program is stored in the storage medium, wherein the computer program is set to execute the method according to any one of claims 1-4 when running.

8. An electronic device, comprising a memory and a processor, characterized in that, A computer program is stored in the memory, and the processor is set to run the computer program to execute the method according to any one of claims 1-4.

Images