Data visualization management methods, systems, equipment and media for enterprise digital transformation

By employing adaptive interface generation, edge cloud collaborative computing, and hybrid recommendation models, the shortcomings of the data visualization management system in terms of terminal adaptation, data processing latency, and personalized recommendations have been addressed, enabling efficient and real-time data display and customized analysis.

CN123115261APending Publication Date: 2026-11-17GUANGDONG ZHONGTE GRP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611200636.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing data visualization management systems lack flexibility in interface interaction and terminal adaptation, resulting in low user operation efficiency; they also suffer from high network bandwidth pressure and transmission delays in data processing timeliness, affecting the timeliness of decision-making; and they fail to provide differentiated recommendations based on user behavior in terms of personalized services, making it difficult to meet customized analysis needs.

Method used

By acquiring and preprocessing multi-source heterogeneous data streams, monitoring traffic in real time and allocating computing tasks to the edge or cloud, collecting interaction operation logs to construct user feature vectors, combining terminal screen parameters to generate adaptive interfaces, establishing a multi-level caching architecture for the edge cloud, and adopting differential data compression and hybrid recommendation models, adaptive layout and real-time processing are achieved.

Benefits of technology

The system can automatically adapt to different terminal screens, reducing the user's operating threshold and improving cross-terminal operation efficiency and information acquisition speed; it reduces network latency through edge cloud collaborative computing to ensure the real-time display of data; and it provides personalized report recommendations to meet the customized analysis needs of users with different roles.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115261A_ABST
    Figure CN123115261A_ABST
Patent Text Reader

Abstract

This invention relates to the field of big data analytics and enterprise management technology, and discloses a method, system, device, and medium for data visualization management in enterprise digital transformation. The method includes: cleaning and filling missing values ​​in multi-source heterogeneous data streams using statistical features and Euclidean distance; calculating a load pressure index based on a comparison of traffic and processing thresholds, dynamically offloading tasks to edge nodes or the cloud accordingly, and performing differential dynamic compression on cloud-transmitted data; introducing a time decay factor to analyze interaction logs to construct user profiles and cluster them; calculating a recommendation score based on the fused content and popularity, determining the number of grid columns based on the terminal screen width, and prioritizing the mapping of high-scoring components to the core area of ​​the first screen to generate an adaptive interface; the system includes acquisition, edge, cloud, and display terminals. This invention solves the problems of high data processing latency and poor terminal adaptation through edge-cloud collaboration and adaptive layout, improving decision-making timeliness.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of big data analysis and enterprise management technology, and in particular to methods, systems, equipment and media for data visualization management of enterprise digital transformation. Background Technology

[0002] As enterprises advance their digital transformation, data visualization tools have become an important means of assisting management decision-making. Existing technologies use general-purpose business intelligence software to integrate and display enterprise data, presenting business status in chart form.

[0003] However, existing data visualization management technologies still have limitations in practical applications. First, in terms of interface interaction and terminal adaptation, existing systems use fixed layouts or simple responsive designs, lacking the ability to perceive terminal screen parameters and content priorities. When accessing data across devices, core data cannot be automatically adapted to key areas of the first screen, requiring non-professional users to perform cumbersome zooming or scrolling operations, reducing information acquisition efficiency.

[0004] Secondly, regarding the timeliness of data processing, the existing architecture relies on centralized cloud processing and lacks load assessment and dynamic traffic distribution mechanisms at the edge. When faced with massive real-time data streams, network bandwidth is under great pressure, which can easily cause data transmission delays and update lags. This makes it impossible for visualization results to reflect the current business dynamics in real time, affecting the timeliness of decision-making.

[0005] Furthermore, in terms of personalized services, traditional systems typically only provide standardized report views, failing to effectively mine users' historical interaction behavior and role characteristics. The system cannot dynamically recommend services based on users' changing business concerns over time. This lack of differentiation forces users to spend time searching through massive amounts of data, making it difficult to meet the customized data analysis needs of different roles. Summary of the Invention

[0006] The purpose of this invention is to provide a data visualization management method, system, device, and medium for enterprise digital transformation. It solves the problems of poor interface layout adaptability leading to low user operation efficiency, high data transmission and processing latency affecting decision-making timeliness, and lack of personalized recommendation capabilities based on user behavior, which makes it difficult to meet customized analysis needs.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] The first aspect of this invention provides a data visualization management method for enterprise digital transformation, comprising:

[0009] Acquire multi-source heterogeneous raw data streams and perform preprocessing including data cleaning, missing value imputation, and standardization to generate standard format data. The cleaning step includes identifying and correcting abnormal data points that deviate from the preset threshold of distribution characteristics, and the imputation step includes selecting nearby historical samples based on Euclidean distance and performing weighted calculation of imputation values.

[0010] Real-time monitoring of data traffic; generating offload instructions based on the comparison between traffic and edge computing node processing capacity thresholds; allocating tasks to the edge or cloud; calculating a load pressure index characterizing the load level; determining the cloud transmission rate when the index exceeds the safe range; extracting the data differential component; and dynamically compressing and transmitting the differential component based on the target compression ratio determined by the ratio of the required transmission bandwidth to the available bandwidth.

[0011] Collect interaction operation logs, introduce a time decay factor to weight behavioral events to construct user feature vectors, and use clustering algorithms to divide user role groups.

[0012] The system calculates a recommendation score between user feature vectors and report attribute vectors, and generates an adaptive interface by combining terminal screen parameters. The recommendation score integrates content matching degree and group popularity based on group access frequency. The interface generation steps include determining the number of grid columns based on the screen's visible width, sorting components according to the recommendation score, mapping high-scoring components to the core area of ​​the first screen, and mapping low-scoring components to secondary areas.

[0013] In addition, the method also includes establishing a multi-level caching architecture for edge clouds, calculating a popularity score based on access frequency and time, and preloading high-popularity data blocks into the first-level cache of edge nodes.

[0014] A second aspect of the present invention provides a data visualization management system for enterprise digital transformation, comprising:

[0015] A data acquisition terminal cluster is configured to acquire raw data streams from multiple heterogeneous sources.

[0016] Edge computing devices are configured to perform data preprocessing to generate standard format data; calculate load stress index, and based on this, generate computing offload instructions to allocate tasks to local or cloud environments, and perform bandwidth-aware differential data dynamic compression during cloud transmission.

[0017] A cloud server cluster is configured to receive unloading tasks and construct feature vectors and divide user role groups based on user interaction logs with a time decay factor.

[0018] The visualization display terminal is configured to calculate a recommendation score that integrates content and popularity. The grid layout is determined based on the score and screen width, and high-scoring components are mapped to the core area of ​​the first screen to generate an adaptive interface for data rendering.

[0019] A third aspect of the present invention provides an electronic device, comprising:

[0020] The system includes a processor, a memory, and a communication interface, which are connected via a bus. The memory stores computer-executable instructions. The processor executes the computer-executable instructions stored in the memory to implement the method described in the first aspect of the present invention.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium,

[0022] A computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in the first aspect of the present invention.

[0023] In summary, the present invention has at least one of the following beneficial technical effects:

[0024] 1. This invention utilizes an adaptive layout mapping mechanism based on recommendation ratings and terminal screen parameters. The system can automatically determine the number of grid columns according to the screen's visible width and prioritize mapping high-rated visual components to the core area of ​​the first screen. This intelligent interface generation method eliminates the tedious manual adaptation to different resolution terminals, lowers the barrier to entry for non-professional users, and improves operational efficiency and information acquisition speed across terminal scenarios.

[0025] 2. This invention adopts an edge-cloud collaborative computing architecture based on load pressure index, combined with differential data compression and multi-level caching mechanism, to realize dynamic distribution and real-time processing of computing tasks. This architecture responds to high-frequency data requests on the edge in real time, and only unloads tasks to the cloud when the load exceeds the safe range, reducing network transmission latency and bandwidth consumption, ensuring real-time updates of data display, thereby improving the timeliness of enterprise decision-making.

[0026] 3. This invention constructs a hybrid recommendation model that integrates content matching degree and group popularity by introducing user behavior modeling and cluster analysis algorithms with time decay factors. This model can capture the dynamic changes in users' recent business focus and generate personalized report recommendation lists accordingly, enhancing the system's ability to provide customized services for users with different roles and meeting the diverse data analysis needs of enterprises. Attached Figure Description

[0027] Figure 1 A flowchart illustrating a data visualization management method for enterprise digital transformation provided in an embodiment of the present invention;

[0028] Figure 2 This is a flowchart illustrating the logic of data preprocessing and missing value imputation in an embodiment of the present invention.

[0029] Figure 3This is a schematic diagram of the unloading decision logic based on the load pressure index in an embodiment of the present invention;

[0030] Figure 4 This is a schematic diagram of the adaptive visual interface layout mapping based on recommendation rating in an embodiment of the present invention;

[0031] Figure 5 This is a schematic diagram of the architecture of a data visualization management system for enterprise digital transformation, provided as an embodiment of the present invention. Detailed Implementation

[0032] The following is in conjunction with the appendix Figure 1 - Appendix Figure 5 The present invention will be further described in detail below.

[0033] See attached document Figure 5 This invention provides a data visualization management system for enterprise digital transformation, which includes: a data acquisition terminal cluster, an edge computing device, a cloud server cluster, and a visualization display terminal.

[0034] The data acquisition terminal cluster is configured to generate raw data streams and establish communication connections with edge computing devices through a first communication network. The data acquisition terminal cluster includes an enterprise resource planning system server, a customer relationship management system server, and industrial IoT sensors deployed on the production site. The raw data stream is denoted as... ,in , here Indicates the first One original data record, It is a positive integer.

[0035] Edge computing devices are deployed on the network access side of a data acquisition terminal cluster. Each edge computing device is equipped with a first processor and a first memory. The edge computing device is configured to receive raw data streams. and the raw data stream Perform preliminary data processing. Preliminary data processing includes data cleaning and traffic monitoring.

[0036] Edge computing devices are configured to monitor the current time. Data traffic Edge computing devices have their own pre-stored processing capacity thresholds. Edge computing devices are configured to adjust based on data traffic. and processing capacity threshold Generate traffic diversion decision instructions.

[0037] The cloud server cluster establishes a communication connection with the edge computing devices through a second communication network and is equipped with a second processor and a second memory. The cloud server cluster is configured to receive data sent by the edge computing devices and perform tasks such as building data analysis models, clustering user behavior profiles, and generating report recommendation strategies. The cloud server cluster provides higher computing resources than the edge computing devices.

[0038] The visualization display terminal connects to a cloud server cluster or edge computing devices via a network and includes a display screen and input interfaces. The visualization display terminal is configured to receive rendering commands and display the user interface. It is also configured to collect user interaction logs and send these logs to the cloud server cluster.

[0039] Figure 1 Edge computing devices, servers in cloud server clusters, or visualization display terminals can all be used. Figure 2 The hardware structure shown.

[0040] Electronic devices include a processor, memory, communication interface, and bus. The processor, memory, and communication interface communicate with each other via the bus.

[0041] Memory configurations are used to store instructions executed by the computer, including high-speed random access memory (RAM) and non-volatile memory. Non-volatile memory includes disk storage devices or flash memory devices.

[0042] The processor is configured to execute computer-executable instructions stored in memory to implement the steps of a data visualization management method.

[0043] The communication interface configuration is used to establish communication connections between electronic devices and other external devices. Communication connections can be wired or wireless. Wired connections include fiber optic communication or twisted-pair communication. Wireless connections include wireless LAN communication or cellular mobile communication.

[0044] The bus is configured as an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect Standard (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. The bus is divided into an address bus, a data bus, and a control bus.

[0045] The processor can be a central processing unit (CPU), a network processor (NP), or a combination thereof. The processor can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices.

[0046] During system operation, the data acquisition terminal cluster continuously outputs raw data streams. The edge computing device receives the raw data stream through the first communication network. Edge computing devices utilize anomaly detection algorithms to identify and process raw data streams. Outliers in the data.

[0047] When edge computing devices determine data traffic When the data load is less than or equal to a preset safe load value, the edge computing device completes data processing locally and sends the results directly to the visualization display terminal; when the edge computing device determines the data flow... When the load exceeds the safe load value, the edge computing device will offload some or all of the computing tasks and corresponding data to the cloud server cluster through the second communication network.

[0048] The cloud server cluster updates the data analysis model based on the received data and generates interface layout parameters based on the user feature vectors fed back from the visualization terminal. The visualization terminal then renders data charts on the display screen according to the interface layout parameters.

[0049] See attached document Figure 1 This invention provides a data visualization management method for enterprise digital transformation, the method comprising:

[0050] S1. Obtain the raw data stream from multiple heterogeneous sources and perform data preprocessing operations on the raw data stream to generate standard format data.

[0051] raw data stream The data originates from Enterprise Resource Planning (ERP) systems, Customer Relationship Management (CRM) systems, or Industrial Internet of Things (IIoT) sensors. Data preprocessing operations include data cleaning, missing value imputation, and data standardization. During data cleaning, the system calculates the statistical distribution characteristics of the numerical sequences in the original data stream, identifies outlier data points that deviate from the statistical distribution characteristics, and removes or corrects these outliers. During missing value imputation, the system calculates imputation values ​​based on the numerical characteristics of neighboring samples. During data standardization, the system maps different data to a unified numerical range and converts the data structure into a preset intermediate data format.

[0052] S2. Real-time monitoring of current data flow and according to Processing capacity threshold of edge computing nodes The comparison results are used to generate calculation unloading instructions.

[0053] The compute offload command is used to instruct the distribution of data processing tasks to edge computing nodes or cloud server clusters. This is relevant when data traffic... Less than or equal to the processing capacity threshold When multiplied by a preset safety factor, the system executes data processing tasks on edge computing nodes. This is based on data traffic... Greater than the processing capacity threshold When multiplied by a safety factor, the system transmits the excess computational tasks and corresponding data packets to the cloud server cluster. During data transmission, the system calculates the difference between the data from the previous moment and the data from the current moment, transmits only the difference, and dynamically adjusts the compression level of the data compression algorithm based on network bandwidth conditions.

[0054] S3. Collect the interaction operation logs of the target users and construct user feature vectors based on the interaction operation logs. .

[0055] The interaction log includes interface dwell time, functional component click frequency, and historical query categories. The system maps the unstructured data in the interaction log into numerical feature dimensions, combining them to form a user feature vector. The system uses clustering algorithms to calculate user feature vectors. The target users are grouped into specific user role groups based on the Euclidean distance between the target users and multiple preset reference centroid vectors, according to the principle of minimum distance. Each user role group is associated with an initialized interface layout template.

[0056] S4. Calculate user feature vectors Report attribute vectors of candidate report templates Recommended ratings between Based on the recommended ratings and the screen parameters of the visualization display terminal, an adaptive visualization interface is generated.

[0057] Recommended rating This characterizes the target user's preference for candidate report templates. The system calculates user feature vectors. With report attribute vector The system calculates a recommendation score based on the cosine similarity between the components and receives user feedback signals to update the weight parameters using gradient descent. When generating the interface, the system obtains the screen's visible width from the visualization display terminal. Calculate the maximum number of columns supported by the current screen. The system then maps the visualization components corresponding to the candidate report templates to the grid system. During the mapping process, the system prioritizes placing visualization components with high recommendation scores in the core visible area of ​​the first screen, and places visualization components with low recommendation scores in secondary areas or collapsed areas, ultimately rendering and outputting the user interface.

[0058] See attached document Figure 2 The data preprocessing operation in step S1 first performs abnormal data cleaning. The system starts from the raw data stream. The sequence of numerical attributes to be detected is parsed from the original data stream. Includes 1 data record, denoted as .in, This represents the total number of raw data records within the current statistical time window. The system starts from the [number]th [time window]. Original data record Extract the target numerical attribute This allows us to construct the sequence of values ​​to be detected.

[0059] ;

[0060] The system calculates the numerical sequence to be detected. Arithmetic mean within the current statistical time window The calculation formula is as follows: ;

[0061] in, Represents the numerical sequence to be detected The arithmetic mean within the current statistical time window; Indicates the first The target numerical attribute corresponding to each original data record; Indicates to From 1 to All Perform cumulative summation.

[0062] Next, the system calculates the sequence of values ​​to be detected. Standard deviation The calculation formula is as follows:

[0063] ;

[0064] in, Represents the numerical sequence to be detected The standard deviation is used to quantify the dispersion of data. Represents the numerical sequence to be detected The arithmetic mean within the current statistical time window; This represents the sum of squared deviations of all sample data, i.e., the sum of squared deviations. It represents the variance, which is the average of the sum of squared deviations.

[0065] In obtaining the arithmetic mean and standard deviation Then, the system analyzes the target data points to be detected. Perform anomaly detection. The system has pre-stored anomaly detection threshold coefficients. The system verifies the target data points. Does it satisfy the following anomaly detection inequality:

[0066] ;

[0067] in, This represents the target data point that needs to be validated at the current moment; Represents the numerical sequence to be detected The arithmetic mean within the current statistical time window; The absolute value operator is represented by the symbol. This represents the preset anomaly detection threshold coefficient, used to control the sensitivity of anomaly detection; this coefficient is a positive real number. Represents the numerical sequence to be detected Standard deviation; This indicates the boundary of the maximum permissible deviation range for data to be considered normal.

[0068] When the anomaly detection inequality holds, that is, when Greater than the boundary of the maximum permissible deviation range At that time, the system determines the target data point The data record in question is classified as an anomalous record. In response to this determination, the system performs data cleaning, which includes cleaning the original data stream. Remove the abnormal record from the list, or retain the abnormal record and extract its contents. The value is set to a null flag for subsequent filling steps.

[0069] In the data preprocessing operation of step S1, when the raw data stream is identified... There are missing data records. And the target data point in the missing data record When the data is empty or invalid, the system performs intelligent filling processing based on the K-Nearest Neighbor (KNN) algorithm. The system first starts from the original data stream... The candidate sample set is formed by filtering out historical data records with complete target attribute values. The system in the candidate sample set Search and missing data records Features most similar 10 neighboring samples, and based on this The attribute values ​​of the neighboring samples are used to estimate the fill value. .

[0070] The system calculates missing data records. With candidate sample set Each candidate sample record The distance between them is calculated using the Euclidean distance. When calculating the distance, the system selects feature attributes other than the target attribute value from the data records as the dimension. The distance calculation formula is as follows:

[0071] ;

[0072] in, Indicates missing data records Record with candidate samples The Euclidean distance between them; This represents the total dimension of the feature attributes used to calculate similarity; Indicates missing data records In the Specific numerical values ​​for each feature dimension; Represents candidate sample records In the Specific numerical values ​​for each feature dimension; Indicates all The summation is performed by accumulating the squares of the numerical differences of each feature dimension.

[0073] The system calculates the Euclidean distance. For candidate sample set Sort in ascending order and select the first few with the smallest distance value. The nearest neighbor set consists of 10 neighboring samples. ,in This is a preset positive integer parameter.

[0074] Next, the system sets the nearest neighbors. Each candidate sample record Calculate the corresponding weights Weighted weights The weights are inversely proportional to the square of the Euclidean distance, used to assign a greater filling influence to samples that are closer together. The weight calculation formula is as follows:

[0075] ;

[0076] in, Represents the nearest neighbor set The Middle The weighted weights corresponding to each sample record; This represents the calculated Euclidean distance between two records.

[0077] Finally, the system utilizes the nearest neighbor set. Calculate the final imputation value by taking the target attribute values ​​of all samples and their corresponding weighted weights. The calculation formula is as follows: ;

[0078] in, This represents the calculated fill value used to fill missing data records; This indicates that the set belongs to the nearest neighbor set. Traverse the sample index in the database; Represents the nearest neighbor set The Middle Each sample record contains the known numerical values ​​of the target attribute; Indicates the first The weighted weights corresponding to each sample record; This represents the sum of the weighted attribute contributions of all samples in the nearest neighbor set; This represents the sum of weighted weights of all samples in the nearest neighbor set, used for normalization.

[0079] The system will calculate the fill value Write missing data records In the target attribute field, complete the repair of missing values, and incorporate the repaired complete data record into the subsequent data standardization process.

[0080] After performing the abnormal data cleaning and missing value imputation operations in step S1, the system obtains a repaired complete numerical sequence. In order to eliminate the differences in units and orders of magnitude between different business indicators and ensure the stability of subsequent data mining model convergence, the system performs Z-Score standardization on the complete numerical sequence.

[0081] The system first determines the complete numerical sequence to be processed. .in, Indicates the first after cleaning and filling The original business values ​​of each raw data record. The system calculates the global arithmetic mean of the complete numerical sequence within the current time window. and global standard deviation .

[0082] Subsequently, the system uses the aforementioned statistics to analyze each original business value in the sequence. Perform a linear transformation and calculate the corresponding standardized value. The calculation formula is as follows:

[0083] ;

[0084] in, This represents the first number obtained after standardization. The standardized value of each original data record, which follows a standard normal distribution; This indicates that after data cleaning and filling are completed, the first... The original business values ​​in each data record; This represents the complete numerical sequence currently to be processed. The global arithmetic mean; This represents the complete numerical sequence currently to be processed. The global standard deviation.

[0085] Through the above calculations, the system maps the original data to standard distribution data with a mean of 0 and a standard deviation of 1.

[0086] After completing numerical standardization, the system performs a data format conversion operation. The system has a pre-defined intermediate data structure (IDS). The intermediate data structure includes a timestamp field, a device identifier field, a service type field, and a standardized numerical field. The system will then convert the calculated data... The system inputs standardized numerical fields and encapsulates the corresponding timestamps and device identifiers to generate a standard data packet. The system then outputs this standard data packet to the subsequent distributed computing or storage module.

[0087] See attached document Figure 3 During the real-time monitoring and scheduling process in step S2, the edge computing device periodically collects the data throughput status of the current network port and its own computing resource occupancy status. The system sets the monitoring time period. In each monitoring time period At the end, the system obtains the data traffic during the monitoring period and records it as... At the same time, the system reads the edge computing device's status at the current moment. The available computing power is denoted as . .

[0088] The system is equipped with a load balancing decision algorithm to determine whether the current data processing task should be executed locally or offloaded to a cloud server cluster. The system first calculates the current load pressure index. Load pressure index The formula used to quantify the load level of edge computing devices is as follows: ;

[0089] in, Indicates the current moment The monitored data flow rate per unit time, measured in megabits per second (Mbps). This indicates the maximum theoretical data processing rate of the edge computing device, measured in megabits per second (Mbps). This represents the preset safety load factor, which is a real number between 0 and 1, used to preserve the system's redundant processing capacity to cope with sudden traffic surges.

[0090] The system calculates the load pressure index. Then, it is compared with the preset decision threshold 1. When When this occurs, it indicates that the current data traffic has not exceeded the security processing capacity of the edge computing device. The system then generates a local processing instruction to handle all incoming data traffic. The data is routed to the local processor of the edge computing device for analysis.

[0091] when This indicates that the current data traffic has exceeded the safe carrying capacity of the edge computing device, and the system triggers the load balancing mechanism. The system's computation requires data traffic to be transmitted over the network to the cloud server cluster. The calculation formula is as follows:

[0092] ;

[0093] in, This indicates the data traffic rate that needs to be transmitted over the network to the cloud server cluster; This represents the data flow at the current moment; Indicates the load pressure index; This indicates the proportion of traffic that the edge computing device can currently handle.

[0094] The system is based on the calculations obtained A flow splitting control signal is generated. In response to the flow splitting control signal, the network scheduler splits the data stream into two parts: the first part of the data stream has a rate of... This part is retained for local processing on the edge computing device; the rate of the second part of the data stream is... This part is encapsulated and sent to the cloud server cluster via a communication interface. The cloud server cluster receives this data and allocates computing resources for parallel processing. The processing results are then returned to the edge computing device or directly sent to the visualization terminal.

[0095] To reduce data access latency in an edge-cloud collaborative computing environment, the system constructs a three-tier data caching architecture between edge computing devices and the cloud server cluster. The three-tier data caching architecture includes: a Level 1 cache layer (L1), a Level 2 cache layer (L2), and a Level 3 cache layer (L3). The Level 1 cache layer resides in the high-speed random access memory of the edge computing device. The Level 2 cache layer resides in the local persistent storage device of the edge computing device. The Level 3 cache layer resides in the database system of the cloud server cluster.

[0096] The system is configured with a cache controller to handle data read requests from applications or visualization terminals. When the system receives a read request for a specific data block, it searches in the order of L1, L2, and L3. The system establishes an average response latency model to evaluate and optimize the performance of the caching strategy. The average response latency model is used to calculate the expected time to complete one data read operation. The calculation formula is as follows:

[0097] ;

[0098] in, This represents the average response latency for the system to acquire data, in milliseconds (ms). This represents the average physical time required to read data from the first-level cache layer; this value depends on the read rate of the edge device's memory. This represents the hit rate of the L1 cache layer, which is the probability that a data request will successfully find the required data in L1. The value ranges from 0 to 1. This represents the average physical time required to read data from the second-level cache layer, and this value depends on the I / O rate of the edge device's local disk; This indicates the probability that a data request will not be hit in L1; This represents the hit rate of the second-level cache layer, which is the conditional probability that a data request will successfully find the required data in the L2 cache if the L1 cache is not hit. This represents the joint probability that a data request is hit only in L2; This represents the average physical time required to retrieve data from the three-level cache layer. This value includes network transmission latency and cloud database query latency. This represents the joint probability that a data request has failed to hit either L1 or L2 and must be retrieved from L3.

[0099] Real-time system monitoring and The system has a preset maximum allowable delay threshold. When the calculated average response delay time Greater than the maximum allowable delay threshold When the system determines that the current caching strategy is invalid, it triggers a cache warm-up and replacement mechanism in response. The system calculates the heat score for each data block. The data block with the highest popularity score is preloaded into the first-level cache layer. The formula for calculating the popularity score is as follows:

[0100] ;

[0101] in, Indicates the first The popularity score of each data block; Indicates the first The frequency of access to each data block in the most recent statistical period; This represents the frequency weighting coefficient, used to adjust the degree to which the frequency of access affects popularity. Indicates the current system time; Indicates the first The last time a data block was accessed; This represents the time decay weighting coefficient, used to adjust the degree of influence of time interval on heat.

[0102] The system is based on the calculated The data is sorted in descending order among all candidate cache blocks, and the top-ranked blocks are written to the first-level cache layer until the capacity limit of the first-level cache layer is reached, thereby improving efficiency. Value and decrease .

[0103] During data transmission in step S2, to minimize network bandwidth consumption and improve real-time performance, the system employs a dynamic compression transmission strategy based on differential data. The system maintains a historical data mirror synchronized with the cloud receiver at the edge computing device's sending end. When the system is at the current moment... When preparing to transmit a new set of data frames, first compare the data frame with the previous time frame. Successfully transmitted data frames are compared bit by bit to extract the differential data portions that have changed.

[0104] The system calculates the instantaneous bandwidth requirements for transmitting differential data. The instantaneous bandwidth requirement is determined by the size of the differential data packets and the current synchronization frequency. The calculation formula is as follows:

[0105] ;

[0106] in, This indicates the instantaneous bandwidth requirement for transmitting differential data at the current moment, expressed in megabits per second (Mbps). Indicates the current moment The physical size of the extracted differential data packets, in megabits (Mb). This indicates the system's preset data synchronization frequency, which is the number of data transmission operations performed per second, measured in Hertz (Hz).

[0107] At the same time, the system monitors the network link status between edge computing devices and cloud server clusters in real time to obtain the available network bandwidth at the current moment. The system calculates the target compression ratio based on available network bandwidth and instantaneous bandwidth requirements. The target compression ratio guides the selection of compression algorithms and parameter adjustments to ensure that the data stream can fit the current network pipeline. The calculation formula is as follows:

[0108] ;

[0109] in, This represents the calculated target compression ratio; This represents the preset network congestion control coefficient, with a value between 0 and 1. It is used to reserve a portion of bandwidth margin when calculating available transmission capacity to prevent packet loss, and is typically set to 0.8 to 0.9.

[0110] The system is based on the calculations obtained The compression strategy is dynamically adjusted.

[0111] when This indicates that the current network bandwidth is sufficient, and the system does not perform compression operations on the differential data packets, or only performs lossless compression algorithms and directly sends the original differential data.

[0112] when This indicates that the current network bandwidth is insufficient to support the real-time transmission of raw data, and the system must perform lossy compression on the data.

[0113] In this situation, the system adjusts the quantization parameters of the compression algorithm to ensure that the compressed data size meets transmission requirements. The system then adjusts the quantization parameters according to the target compression ratio. Determine the actual compressive strength. The calculation formula is as follows:

[0114] ;

[0115] in, This indicates the actual compression intensity performed. The larger the value, the higher the compression ratio, but the greater the loss of data accuracy. This indicates the preset base quantization step size; This indicates the factor by which the basic quantization step size is amplified based on the bandwidth gap.

[0116] The system uses the calculated compressive strength The differential data packets are encoded and compressed to generate compressed data packets, which are then sent to the cloud server cluster via a communication interface. Upon receiving the data, the cloud server cluster uses the corresponding decoding algorithm to restore the data and merges it with the local historical image to recover the complete data frame for the current moment.

[0117] In the user feature construction process of step S3, the system aims to transform discrete, unstructured user interaction logs into a computer-processable mathematical vector representation. The system collects target user data in real time. The operational behavior data on the visual management platform includes page browsing history, component click history, report export history, and query condition setting history. The system predefines... Each business focus dimension corresponds to a key area in enterprise management, such as financial status, inventory levels, production efficiency, or sales performance. The system is designed for target users. Build a 3D user feature vector .

[0118] The system calculates each standardized user feature vector component in the user feature vector. Corresponding original interest value To accurately reflect users' current focus, the system introduces a time decay mechanism, giving more weight to recent interactions in the interest score calculation, while the weight of older interactions decays exponentially over time. The calculation formula is as follows:

[0119] ;

[0120] in, Indicates target user For the The original cumulative value of interest for each business focus dimension, among which The value range is from 1 to ; This indicates that the target user has a history with the first... The total number of interaction events related to each business dimension; The index variable represents the interactive event, with a value range from 1 to... ; Indicates to From 1 to All Perform cumulative summation; Indicates the first The weight of the behavior type of each interactive event is determined by the depth of the behavior. For example, the weight of the data export behavior is greater than the weight of the page browsing behavior. The base of the natural logarithm; This represents the preset time decay coefficient, used to control the rate at which the influence of historical behavior decreases over time. This coefficient is a positive real number. The timestamp represents the current system calculation time. Indicates the first The timestamp of when each interactive event occurred; This represents the time decay factor, which decreases monotonically as the time interval increases.

[0121] After calculating all After accumulating the original interest values ​​for each business focus dimension, the system performs L2 normalization on the vector composed of these values ​​to eliminate the impact of user activity differences on the feature vector magnitude, generating the final standardized user feature vector components. The calculation formula is as follows:

[0122] ;

[0123] in, Represents the normalized user feature vector In the The standardized user feature vector components in each business dimension have values ​​ranging from 0 to 1. Indicates the first The original cumulative interest value for each business dimension; Indicates the first The square of the original cumulative interest value for each business dimension; This indicates the target user The sum of squares of the original interest values ​​across all business dimensions; This represents the Euclidean norm of the original interest vector.

[0124] Through the above calculations, the system generates user feature vectors with clear directionality. This user feature vector represents the target user in a multidimensional space. The system will generate the business preference direction. The feature vectors are stored in the user profile database and periodically updated as new interaction logs are generated, ensuring that the feature vectors always reflect the user's latest attention status.

[0125] In the intelligent recommendation process of step S3, after obtaining each target user User feature vector Subsequently, the system performs unsupervised learning-based clustering analysis to identify user groups with similar business interest patterns. The system collects feature vectors from all active users on the platform to form a user vector set. ,in This represents the total number of users participating in the clustering.

[0126] The system uses an improved K-Means++ clustering algorithm to divide the user vector set into... Clusters, of which This represents the preset total number of clusters. The system defines the objective function for cluster optimization. The goal is to minimize the sum of squared distances between all users and the center of their respective clusters. The objective function is formulated as follows:

[0127] ;

[0128] in, This represents the preset total number of clusters; The index variable represents the cluster, with values ​​ranging from 1 to... ; This represents the total number of users participating in the clustering. This indicates that the summation is performed by iterating through all clusters in the outer layer; This indicates that the summation is performed by iterating over all target users in the inner layer. Indicates the attribution indicator variable, when the first The target user was assigned to the first The value is 1 when there are multiple clusters; otherwise, it is 0. Indicates the first The centroid vector of each cluster, whose dimension is... same; This represents the square of the Euclidean distance between the user's feature vector and the centroid vector.

[0129] The system updates the centroid vector of each cluster through iterative calculation. In each iteration cycle, the system first fixes the centroid vector and determines the attribution indicator variable for each user based on the principle of minimizing distance. Subsequently, the system recalculates the centroid vector based on the current affiliation. The centroid update formula is as follows:

[0130] ;

[0131] in, Indicates the updated number The new centroid vectors of each cluster; Indicates that it currently belongs to the first The vector sum of all user feature vectors in each cluster; Indicates that it currently belongs to the first Total number of users in each cluster.

[0132] The system repeats the above allocation and update steps until the objective function is reached. The change value is less than the preset convergence threshold, or the preset maximum number of iterations is reached.

[0133] After clustering converges, the system assigns a cluster name to each cluster. Generate corresponding group profile tags. The system extracts the centroid vector. The largest median value Each dimension is mapped to specific business keywords (such as inventory-sensitive or financial statement-related), and this mapping relationship is stored in the recommendation rule base. When a new user logs into the system, the system calculates their feature vector and each centroid vector. Based on the distance, it is assigned to the nearest cluster and inherits the group profile tags and interface recommendation strategy corresponding to that cluster.

[0134] In the intelligent recommendation operation of step S3, the system uses the user feature vector constructed in the preceding steps. Based on the identified user clusters, a hybrid recommendation algorithm is executed to filter the most suitable target users from the report database. A list of reports required by business needs. Assume the system report library contains... There are 1 candidate report, denoted as .in, A positive integer, representing the total number of candidate reports; Indicates candidate reports, This represents the index identifier for the candidate report, with a value ranging from 1 to... .

[0135] The system first calculates a content-matching score. The system pre-defines each candidate report. The report attribute vector was constructed. Business focus dimensions of report attribute vectors With user feature vector The dimensions remain consistent, corresponding to the same business focus areas. Indicates the first The candidate report in the first Content weights across each business dimension. The system calculates user feature vectors. With report attribute vector The cosine similarity between them. The calculation formula is as follows:

[0136] ;

[0137] in, Indicates target user With the The content matching score between the candidate reports is rated from 0 to 1. Indicates target user The eigenvectors in the th Values ​​in each dimension; Indicates the first The attribute vector of the candidate report in the th... Content weighting across multiple dimensions; Represents the dot product of two vectors; This represents the product of the magnitudes of two vectors, used to normalize the calculation result.

[0138] Meanwhile, the system calculates a group popularity score based on collaborative filtering. Target users identified using cluster analysis To which cluster The system counts all other users within this cluster regarding the first... The historical access frequency of each candidate report. The calculation formula is as follows:

[0139] ;

[0140] in, This represents the group popularity score of the l-th candidate report in the cluster c to which the target user belongs, with a value ranging from 0 to 1; This represents the total number of times all users belonging to cluster c access the l-th candidate report within the statistical time window; This represents the number of visits to the report with the highest number of visits in the same statistical time window, which serves as the normalization benchmark.

[0141] Finally, the system uses a dynamic weighted fusion strategy to calculate the final recommendation score for the l-th candidate report. The system introduces a user maturity coefficient λ to balance content preferences and group influence. The calculation formula is as follows:

[0142] ;

[0143] in, λ represents the final recommendation score of the l-th candidate report; λ represents the user maturity coefficient, which is positively correlated with the richness of the target user u's historical interaction logs, and ranges from 0 to 1; when the user is a new user, λ takes a smaller value, and the system is more inclined to recommend popular reports of the group; when the user has rich historical behavior, λ takes a larger value, and the system is more inclined to recommend reports that match their personalized characteristics. This indicates the content relevance score; This indicates the popularity score of the aforementioned group.

[0144] The system calculates all The system calculates the final recommendation score for each candidate report and sorts them in descending order of score. The system selects the top-ranked reports. The report identifier generates a recommendation list and pushes it to the "You May Like" or "Smart Recommendation" function area of ​​the visualization terminal. This is the preset length of the recommended list.

[0145] See attached document Figure 4 In step S3, a generator containing After generating a recommended list of candidate reports, the system activates its adaptive rendering engine. To ensure the visualization content adapts to different display sizes, the system first obtains the screen width of the current visualization display terminal via the front-end display driver interface. The system presets the minimum display width for a single standard visual component. and the horizontal spacing between components .

[0146] Based on the obtained screen viewing width, the system calculates the maximum number of columns that the current screen can support in a horizontal arrangement. This calculation step aims to determine the basic architecture of the grid system, ensuring that components are neither too dense, leading to information overlap, nor too sparse, wasting screen space. The calculation formula is as follows:

[0147] ;

[0148] in, This indicates the maximum number of columns that the current screen can support in a horizontal orientation; Indicates the visible width of the screen of the visualization display terminal; This indicates the preset horizontal spacing between components, used to maintain visual separation; Indicates the minimum display width of a single standard visual component; This indicates a floor function, ensuring that the number of columns is an integer and that the component does not exceed the screen boundaries.

[0149] Determining the maximum number of columns Then, the system constructs a virtual grid coordinate system. The system then updates the recommendation list according to the final recommendation score. Arrange them in descending order to generate an ordered sequence of components. .in, This indicates the position in an ordered sequence. The visualization component ranked [number] has a recommendation score that is better than or equal to that ranked [number]. A visual component for bits.

[0150] The system executes priority mapping logic, placing visual components with high recommendation ratings in the core visible area of ​​the first screen. The system calculates the... Row coordinates of the bit visualization component in the grid system Column coordinates The calculation formula is as follows:

[0151] ;

[0152] ;

[0153] in, Indicates the first The row coordinates of the visual component in the grid system. The smaller the row coordinate, the closer it is to the top of the screen, i.e., the visual core area. Indicates the first The column coordinates of the bit visualization component in the grid system follow a left-to-right reading order; This represents the sort index of the visualization component in the ordered sequence, with a value ranging from 1 to... ; This indicates the maximum number of columns that the current screen can support in a horizontal orientation; This indicates a modulo operation, used to determine the horizontal position of a visual component within the current row.

[0154] The system calculates the row coordinates. The visualization components are categorized by region. The system has a preset maximum number of rows displayed on the first screen. .when When the system determines that the component belongs to the core visible area of ​​the first screen, it generates an instant rendering instruction and loads the complete interactive chart of the visual component. When this happens, the system determines that the visualization component belongs to a secondary region or a collapsed region. For such visualization components, the system generates a lazy loading instruction or a collapsed summary card. To further distinguish between secondary regions and collapsed regions, the system calculates the collapse probability of the visualization component. The collapsing probability determines whether low-scoring visualization components are collapsed by default. The calculation formula is as follows:

[0155] ;

[0156] in, Indicates the first The probability of folding of the bit visualization component; This indicates the row coordinates of the visualization component; This indicates the threshold for the number of visible rows on the first screen. This indicates that the visualization component extends beyond the number of rows on the first screen; This represents the preset collapse sensitivity coefficient. This value is a positive real number and is used to control the rate at which the probability of the visualization component being collapsed increases as the number of rows increases.

[0157] The system uses the calculated grid coordinates The system generates a final page layout description file (DOM Structure) based on the state determination result (full rendering or collapsed), and outputs this page layout description file to the rendering engine for drawing the user interface. Through the above process, the system achieves the technical effect of automatically focusing high-value information on the golden position of the screen, while dynamically adjusting the layout density according to the screen width.

[0158] The data visualization management system in this embodiment is based on a layered architecture, comprising, from bottom to top, a data access and preprocessing layer, a core computing and scheduling layer, an intelligent analysis and recommendation layer, and an interactive display layer. Data interaction and command transmission between each layer are achieved through standardized API interfaces or an internal message bus.

[0159] At the data access and preprocessing layer, the system is configured with a multi-source heterogeneous data access subsystem. This subsystem integrates a protocol adaptation engine, which stores driver libraries for different industrial protocols (such as Modbus and OPC UA) and database protocols (such as JDBC and ODBC). When the system connects to an external data source, the protocol adaptation engine automatically identifies the data source type and loads the corresponding communication driver, establishing a stable data transmission channel. This multi-source heterogeneous data access subsystem also includes a metadata management unit, used to extract structural description information from the source data, including field names, data types, and data precision, and map this metadata to a unified, general-purpose data model within the system, thereby shielding the heterogeneity of the underlying data sources.

[0160] The core computing and scheduling layer is the system's operational hub, containing an edge-cloud co-controller. This edge-cloud co-controller integrates load balancing decision-making algorithm logic. Internally, the edge-cloud co-controller has a status monitoring probe that periodically reads the CPU utilization, memory usage, and network I / O throughput of hardware devices at millisecond-level granularity, calculating the load pressure index in real time. The edge-cloud co-controller also includes a traffic shaping unit, which dynamically adjusts the priority of the data packet sending queue based on the calculated overflow traffic value and controls the data compression module to initiate a differential compression encoding process.

[0161] The core computing and scheduling layer further includes a multi-level cache management module. This module maintains index tables for the first-level memory cache, the second-level disk cache, and the third-level cloud cache. The preheating scheduler within the multi-level cache management module generates background I / O tasks during system idle cycles based on the data block popularity score ranking results. These tasks asynchronously load high-frequency data blocks from low-speed storage media to high-speed storage media, ensuring zero-wait response for data read requests.

[0162] The intelligent analysis and recommendation layer mainly consists of a user behavior insight engine and recommendation algorithm service components. The user behavior insight engine deploys a front-end event tracking agent, which is embedded in the runtime environment of the visualization terminal to capture atomic interaction events such as user mouse clicks, page scrolling, and dwell time. The vector computation unit inside the insight engine receives these atomic event streams, performs weighted calculations and time decay operations in real time, and dynamically updates the user feature vector. .

[0163] The recommendation algorithm service component includes a clustering analyzer and a hybrid recommendation engine. The clustering analyzer periodically runs the K-Means++ algorithm to iteratively calculate the centroid positions of user groups based on the user's current feature vector library. The hybrid recommendation engine, based on the user's current feature vector and their group affiliation, performs content-based filtering and collaborative filtering-based retrieval in parallel, calculates the comprehensive recommendation score for each candidate report template, and generates an ordered sequence of recommendation list objects.

[0164] The interactive presentation layer consists of an adaptive layout renderer and a component lifecycle manager. The adaptive layout renderer is initialized when the visualization terminal starts up, and reads the visible screen width through the display driver interface. The renderer's internal mesh calculation unit dynamically divides the virtual mesh system based on the screen width and a preset baseline width, and establishes a mapping relationship between the screen coordinate system and the mesh index.

[0165] The component lifecycle manager is responsible for instantiating visual component objects based on the sorting and layout description files in the recommendation list. For components determined to be in the core visible area of ​​the first screen, the manager immediately allocates video memory resources and establishes data binding connections, triggering graphics drawing instructions. For components determined to be in the collapsed area, the manager only creates a lightweight placeholder object, without actually loading data or rendering graphics, until it detects a user's expand or scroll operation, thereby minimizing the initial loading time and memory usage of the front-end page.

[0166] The system is also equipped with a security authentication and access control module that spans all the above levels. Based on role-based access control (RBAC) policies, it intercepts illegal connections at the data access layer and filters sensitive report data that users do not have permission to access at the intelligent recommendation layer, ensuring the security and compliance of the entire data flow process.

[0167] This embodiment provides an electronic device that can be configured as an edge computing node, a cloud server, or a user computer hosting a visualization terminal. The electronic device physically includes at least one processor, a memory, a communication interface, and a system bus connecting the various components.

[0168] The processor is the core of an electronic device's computation and control center. It is equipped with an arithmetic logic unit (ALU) and a floating-point unit (FPU) for performing mathematical model calculations. Specifically, when the processor executes a computer program stored in memory, it performs the following operations: real-time acquisition of hardware monitoring data and inputting it into formulas to calculate load stress indices. Based on the calculation results, determine whether to trigger the edge cloud collaborative traffic offloading strategy; and calculate the user feature vector. The distance between the reference centroid vector of the standard user role and the reference centroid vector.

[0169] The memory includes high-speed random access memory (RAM) and non-volatile memory. The high-speed RAM is configured to physically carry the Level 1 cache (L1) for storing heat scores. The highest data block size, providing nanosecond-level data read response.

[0170] Non-volatile memory (such as solid-state drives (SSDs) or hard disk drives (HDDs) is configured to physically carry the second-level cache layer (L2) for persistent storage of data blocks that are accessed less frequently but may be accessed again in the near future.

[0171] In addition, the memory stores an operating system, a database management system, and application program instructions specifically for executing the distributed real-time data processing and visualization method of the present invention.

[0172] The communication interface is configured as a physical port supporting multiple network communication protocols, including but not limited to Ethernet interfaces, fiber optic interfaces, or wireless network adapters. The communication interface is used to establish a data transmission channel between edge computing devices and cloud server clusters.

[0173] During data transmission, the communication interface, in conjunction with the processor, monitors the network link status in real time to obtain available network bandwidth. The processor sends differential data packets processed by a dynamic compression algorithm through the communication interface, and receives the layout description file of the visualization component at the receiving end through the communication interface.

[0174] The system bus is used to transmit signals between the processor, memory, and communication interfaces. This bus architecture employs a high-bandwidth design to ensure the flow of massive amounts of industrial monitoring data between components, especially during large-scale matrix operations, meeting the processor's throughput requirements for memory data.

[0175] This embodiment also provides a computer-readable storage medium. The computer-readable storage medium may be a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. A computer program is stored on the computer-readable storage medium.

[0176] When a computer program is read and executed by the processor of a computer or electronic device, it causes the computer or electronic device to implement the method steps described in any embodiment of the present invention, specifically including:

[0177] Step S1: Perform data cleaning and standardization at the edge, and calculate the load stress index based on hardware monitoring data. Based on the comparison between the load pressure index and the preset threshold, the data flow direction is dynamically adjusted, and the overflow traffic is dispatched to the cloud for processing.

[0178] Step S2: Calculate the instantaneous bandwidth requirement during data transmission. and target compression ratio Dynamically adjust compression intensity according to the target compression ratio Compress the differential data and optimize data read and write through a multi-level caching strategy;

[0179] Step S3: Construct user feature vectors The distance between the user's feature vector and the baseline centroid vector of the standard user role is calculated to determine the user's role and cluster. A personalized recommendation list is generated by combining the content-based matching score and the collaborative filtering-based group popularity score.

[0180] Step S4: Obtain the visible width of the screen And calculate the maximum number of columns. Calculate the grid coordinates of the component based on the recommendation score. High-resolution components are mapped to the core visible area of ​​the first screen, while low-resolution components are mapped to secondary or collapsed areas for rendering.

[0181] By combining hardware and software, the electronic device and storage medium in this embodiment can solve the problems of high processing latency, large bandwidth consumption and interface display in the face of massive high-concurrency data, and achieve high real-time performance of data processing, high efficiency of transmission and intelligent user experience.

Claims

1. A data visualization management method for enterprise digital transformation, characterized in that: Includes the following steps: S1. Obtain the multi-source heterogeneous raw data stream, and perform data preprocessing operations including data cleaning, missing value filling and data standardization on the raw data stream to generate standard format data; S2. Monitor the data flow at the current moment in real time, generate a computation unloading instruction based on the comparison result between the data flow and the processing capacity threshold of the edge computing node, and allocate the data processing task for the standard format data to the edge computing node or cloud server cluster according to the computation unloading instruction. S3. Collect the interaction operation logs of the target user, construct a user feature vector based on the interaction operation logs, and use a clustering algorithm to classify the target user into user role groups; S4. Calculate the recommendation score between the user feature vector and the report attribute vector of the candidate report template, generate an adaptive visualization interface based on the recommendation score and the screen parameters of the visualization display terminal, and use the adaptive visualization interface to render and display the standard format data processed by the data processing task.

2. The data visualization management method for enterprise digital transformation according to claim 1, characterized in that, Step S1, specifically the step of performing data cleaning on the original data stream, includes: The distribution characteristics of the numerical sequence in the original data stream are statistically analyzed, and target data points that deviate from the distribution characteristics by more than a preset threshold are identified. The target data point is determined to be an abnormal record and is then removed or corrected. Step S1, specifically the step of performing missing value imputation on the original data stream, includes: Based on Euclidean distance, the nearest neighbor sample that is closest to the missing data record is selected from historical data records; The neighboring samples are weighted according to the Euclidean distance to calculate the fill value for the missing data record.

3. The data visualization management method for enterprise digital transformation according to claim 1, characterized in that, Step S2, the step of generating a computation offloading instruction based on the comparison result between the data traffic and the processing capacity threshold of the edge computing node, specifically includes: The load pressure index is obtained by calculating the ratio of the data traffic to the product of the processing capacity threshold and the preset safe load factor. The load pressure index characterizes the load level of the data traffic relative to the product of the processing capacity threshold and the safe load factor. When the load pressure index exceeds a preset safety range, the transmission traffic rate that needs to be transmitted to the cloud server cluster is determined, and the transmission traffic rate is positively correlated with the excess ratio of the load pressure index.

4. The data visualization management method for enterprise digital transformation according to claim 1, characterized in that, Step S2, which involves assigning data processing tasks for the standard format data to the cloud server cluster, specifically includes: Extract the difference between the data from the previous time step and the data from the current time step; The target compression ratio is determined based on the ratio of the bandwidth required to transmit the differential components to the currently available network bandwidth. The compression intensity of the data compression algorithm is dynamically adjusted according to the target compression ratio, and the differential components are compressed and then transmitted.

5. The data visualization management method for enterprise digital transformation according to claim 1, characterized in that, Step S3, the step of constructing a user feature vector based on the interaction operation log, specifically includes: A time decay factor is introduced to weight the behavioral events in the interaction log to generate a business dimension interest value that reflects the user's recent attention. Step S4, specifically the step of calculating the recommendation score, includes: The fusion calculation is based on content matching score and group popularity score based on collaborative filtering, wherein the content matching score is based on the similarity between the user feature vector and the report attribute vector, and the group popularity score is based on the access frequency of the user role group.

6. The data visualization management method for enterprise digital transformation according to claim 1, characterized in that, Step S4, the step of generating an adaptive visualization interface based on the recommended score and the screen parameters of the visualization display terminal, specifically includes: The number of columns in the grid system is determined based on the screen viewing width of the visualization display terminal; The visualization components are sorted according to the recommended scores; The visualization components with high recommendation ratings are mapped to the core visible area of ​​the first screen in the grid system, while the visualization components with low recommendation ratings are mapped to secondary areas or collapsed areas.

7. The data visualization management method for enterprise digital transformation according to claim 1, characterized in that, The enterprise digital transformation data visualization management method also includes: A multi-level data caching architecture is established between the edge computing node and the cloud server cluster; A popularity score is calculated based on the access frequency and last access time of the data block. The data block with the highest popularity score is preloaded into the first-level cache layer of the edge computing node.

8. A data visualization management system for enterprise digital transformation, characterized in that: The data visualization management method for enterprise digital transformation as described in claims 1-7 includes: A cluster of data acquisition terminals is configured to acquire raw data streams from multiple heterogeneous sources. An edge computing device is configured to perform data preprocessing operations on the raw data stream to generate standard format data, monitor the data flow at the current moment in real time, generate a computing offload instruction based on the comparison result of the data flow and the processing capacity threshold, and allocate the data processing task for the standard format data to a local or cloud server cluster according to the computing offload instruction. A cloud server cluster is configured to receive unloaded data processing tasks, collect interaction operation logs of target users, construct user feature vectors based on the interaction operation logs, and use clustering algorithms to classify the target users into user role groups. A visualization terminal is configured to calculate a recommendation score between the user feature vector and the report attribute vector of the candidate report template, generate an adaptive visualization interface based on the recommendation score and screen parameters, and use the adaptive visualization interface to render and display the standard format data processed by the data processing task.

9. An electronic device, characterized in that, The data visualization management method for enterprise digital transformation applied to any one of claims 1 to 7 includes: The processor, memory, and communication interface are connected via a bus. The memory stores computer-executed instructions; The processor executes the computer execution instructions stored in the memory.

10. A computer-readable storage medium, characterized in that, The data visualization management method for enterprise digital transformation according to any one of claims 1 to 7 is provided, wherein a computer program is stored in the computer-readable storage medium and the computer program is executed by a processor.