Enterprise information visual management system and method based on data sharing

Abstract

The invention discloses an enterprise information visual management system and method based on data sharing, and belongs to the technical field of enterprise information management.The method comprises the steps that by connecting business systems in an enterprise, user behavior data and business data are collected and preprocessed, and an internal data set is constructed; acquiring data from an external data source, and constructing an external data set through semantic analysis, screening and classification; internal and external data are integrated, a dynamic cleaning rule is formulated, multi-source fusion is realized through data association, data are stored in a classified manner according to business requirements, and a data warehouse with a layered architecture is constructed; meanwhile, a personalized data mart is generated based on user roles and permissions, user demands are automatically predicted, and analysis indexes and dimensions are pre-generated; and finally, constructing a visual analysis platform, generating a preliminary recommendation scheme in combination with the pre-generated content and the user operation track, fusing user-defined setting, and outputting a final visual result in real time, thereby realizing efficient management and visual presentation of enterprise information.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise information management, and particularly to an enterprise information visualization management system and method based on data sharing. Background Art

[0002] In today's data-driven era, the operation and competitiveness of enterprises increasingly rely on the effective management and utilization of information. Data has become an important asset of enterprises, and its value lies not only in the quantity stored, but more in how to efficiently share, analyze, and apply this data. However, despite the huge investment of enterprises in information technology, in actual operation, many enterprises still face difficulties in data management and application and cannot fully unleash the potential of data.

[0003] During the process of digital transformation, enterprises often face the problem of data islands. The data formats between different departments and systems are not unified, and the data integration efficiency is low, making it impossible for data to flow smoothly. For example, the sales department may use a CRM system to store data, while the finance department relies on an ERP system. The differences in format and standards between the two make cross-departmental data sharing and analysis extremely difficult. Such data islands not only increase the difficulty and cost of data integration, but also seriously affect the enterprise's ability to obtain a comprehensive and accurate data view, thereby having an adverse impact on the timeliness and accuracy of decision-making. In addition to the challenges in data integration, traditional enterprise information visualization tools are also difficult to meet the growing needs of modern enterprises. The functions of existing tools are limited to static report display, lacking dynamic data analysis capabilities. Users cannot flexibly perform data filtering, extraction, and linkage analysis, and it is difficult to deeply explore the potential patterns and trends behind the data. Moreover, existing visualization tools lack personalized customization and intelligent recommendation mechanisms and cannot provide targeted visualization solutions based on user roles, behaviors, and analysis requirements. Users need to spend a lot of time manually configuring analysis dimensions and display forms, resulting in valuable business insights being difficult to be discovered in a timely manner, further restricting the accuracy and scientific nature of enterprise decision-making. Summary of the Invention

[0004] The purpose of the present invention is to provide an enterprise information visualization management system and method based on data sharing to solve the problems raised in the above background art.

[0005] To solve the above technical problems, the present invention provides the following technical solution: An enterprise information visualization management method based on data sharing, including: Step S100: Construct a first processing unit, collect user behavior data and business data by connecting various business systems within the enterprise, and perform preprocessing to construct an internal data set; Step S200: Construct a second processing unit, obtain data from an external data source, perform semantic parsing and feature extraction, conduct intelligent screening and classification in combination with a preset business rule library, extract enterprise-related data through a similarity matching algorithm, and construct an external data set; Step S300: Integrate the internal data and external data of the first processing unit and the second processing unit, formulate dynamic data cleaning rules, and perform multi-source data fusion through data association; Step S400: Classify and store the integrated data according to different business requirements and analysis topics of the enterprise, form an enterprise special topic library, and adopt a hierarchical architecture design to construct a data warehouse to store data; Step S500: Extract the data in the data warehouse, dynamically generate a personalized data mart based on user roles and permissions, automatically predict user analysis requirements, and pre-generate analysis metrics and data dimensions; Step S600: Construct a visual analysis platform, generate a preliminary visual analysis recommendation plan according to the pre-generated analysis metrics and data dimensions in combination with the user operation track; on this basis, the user customizes and selects the analysis data range, metrics and methods according to the requirements, and the platform intelligently fuses the pre-generated content with the user's custom settings to generate the final visual result in real time.

[0006] Further, Step S100 includes: Step S101: The first processing unit constructs an enterprise-level data bus based on a message queue middleware, establishes two-way communication with the enterprise business system through a standardized RESTful API; adopts a combination of timed polling and event triggering to collect user data and business data; the user data includes the operation logs and page access records of the user in the system, records the page paths visited by the user, the stay time, and clickstream data; set each record in the user data to be U1, U2,..., Um respectively, then the user data set is represented as U={U1, U2,..., Um}, where Um represents the mth user record; the business data includes obtaining business types, business process node information, business transaction data, resource usage data, and business association data according to the enterprise business system; set each record in the business data to be B1, B2,..., Bn respectively; then the business data set is represented as B={B1, B2,..., Bn}, where Bn represents the nth business record; Step S102: Use the K-means clustering algorithm to perform clustering analysis on the user operation logs to automatically identify abnormal access patterns; for time series data, use the ARIMA model to fill in missing values, use a hash function to judge the uniqueness of data records, and remove duplicate data records; set a unified standard for enterprise data, perform format conversion on the collected data, and process numerical data through the Z-score normalization method; after preprocessing, construct an internal dataset D = U ∪ B.

[0007] Further, step S200 includes: Step S201: Use the HTTP request library to send requests to the APIs of external data sources to obtain data; at the same time, for data sources that provide file download services, regularly download and parse files; for the obtained text data, process it through natural language processing technology, extract key information and semantic features in the text, and convert the text data into vector representations; for image data, use a convolutional neural network to extract feature vectors of the images. Step S202: Combine the preset business rule library to perform intelligent screening and classification on the processed external data, calculate the matching degree between the data and the business rules using cosine similarity, set a matching degree threshold, extract the data with calculation results greater than the threshold and perform preprocessing, and set the data records related to the enterprise obtained from the external data source and after screening and classification as E1, E2,..., Ek; then the external dataset is represented as E = {E1, E2,..., Ek}, where Ek represents the k-th external data record; the preset business rule library includes keywords related to enterprise business, classification criteria, and data screening conditions.

[0008] Further, step S300 includes: Step S301: Perform field analysis on the internal dataset D = U ∪ B and the external dataset E, for each field, identify its corresponding relationship in different datasets, and establish a mapping relationship between the data through data dictionary matching; construct a unified metadata warehouse for recording various data sources, field definitions, and data quality. Step S302: Dynamically generate data cleaning rules according to the business type. First, construct a multi-layer rule template library, which is classified according to business type, data type, and cleaning target. Automatically identify the business type through the data source metadata, screen the applicable rules from the rule template library according to the business type, and generate a candidate rule set R = {R1, R2,..., Ri}, where Ri represents the i-th candidate rule. Assign priorities to each rule, combine the candidate rules with the business data, and generate the final cleaning rules. Use a fuzzy matching algorithm to perform cross-data-source association on relevant entities, set a similarity threshold, and when the matching degree of two entities exceeds this threshold, perform data merging. Use linear interpolation to align the time granularities of different data sources. Combine the business rule library, and when there are data conflicts between internal data and external data, give priority to the internal data status. Complete the fusion of multi-source data through the above steps to form a fusion data set F.

[0009] Further, step S400 includes: Step S401: Set the enterprise business requirements and analysis topics through historical query logs and business department settings. Combine the field characteristics of the fusion data set F to establish a mapping relationship table between fields and topics. Based on the field mapping relationship, extract relevant fields from the fusion data set F to generate an independent enterprise topic library T = {T1, T2,..., Tk}, where Tk represents the k-th enterprise topic. When generating the topic library, take each business topic as the core and aggregate all related fields.

[0010] Step S402: Use a hierarchical architecture design to construct a data warehouse, which is divided into an operational data layer, a data integration layer, a summary data layer, and an application layer. The operational data layer directly stores the original data obtained from the fusion data set F as the basic data source of the data warehouse. The data integration layer performs mild aggregation processing on the original data, identifies redundant fields, and at the same time performs standardization conversion on the data to unify the data format and coding rules. The summary data layer performs deep aggregation on the integrated data and conducts thematic modeling based on the business logic of the enterprise topic library T. The application layer is used to store the enterprise topic library T.

[0011] Further, step S500 includes: Step S501: Define a set of user roles R = {R1, R2,..., Rx}, where Rx represents the x-th user role. At the same time, determine the corresponding set of permissions H = {H1, H2,..., Hx} for each user role, where Hx is the permission owned by role Rx. Extract the data in the enterprise topic library T from the application layer of the data warehouse. For each user role, only select the data that its permissions allow to access, organize the filtered data, and dynamically generate a personalized data mart M = {M1, M2,..., Mx}, where Mx is the personalized data mart generated for role Rx. Step S502: Extract the user's historical query features, operation behavior features, business environment features, and user attribute features from the user dataset U; the historical query features represent the analysis metrics and data dimensions queried by the user in the past period of time, and are encoded as one-hot vectors; the operation behavior features obtain the page jump path, data filtering operation, and chart interaction behavior through the user operation log, and are transformed into sequence data; the business environment features represent the current business stage and industry dynamic data of the enterprise, and are normalized; the user attribute features represent the relevant attribute information of the user's department and position level, and are transformed into numerical features through an encoding method; all the extracted features are integrated to construct a model input vector; Step S503: Divide the user's historical behavior data into training samples according to a time window, and each sample contains an input feature vector and the corresponding true query metrics and dimension labels; construct an LSTM network layer to capture the time series features of the user's behavior, stack an attention mechanism layer, add a fully connected layer and a Softmax layer at the end of the network, and output the probability distributions of metrics and dimensions; adopt the Adam optimizer, use the cross-entropy loss function as the optimization objective, perform backpropagation on the training set, and continuously adjust the model parameters; use the validation set to evaluate the trained model, calculate the accuracy, recall rate, and F1 value until all evaluation metrics of the model reach the preset standard; deploy the trained model to the server of the visualization analysis platform, and based on the trained user analysis requirement prediction model, receive the user's current operation data and business environment data in real time as inputs, and the model outputs the probability distributions of the predicted analysis metrics and data dimensions. Set a probability threshold α, where 0 < α < 1, and filter out the analysis metrics and data dimensions whose predicted probability values exceed the threshold, and pre-generate an analysis metric set and a data dimension set.

[0012] Further, step S600 includes: Step S601: Build a visual analysis platform. The underlying layer of the platform is connected to the personalized data mart M generated in step S500, the pre-generated analysis metric set, and the data dimension set. At the same time, the operation track data of users on the platform is recorded through the user data set; the selection behavior of users for different analysis metrics and data dimensions is converted into a transaction data set, where each transaction represents a user operation and contains the analysis metrics and data dimensions selected by the user in this operation; set the support threshold and confidence threshold; use the Apriori algorithm to screen all frequent item sets that meet the support threshold from the transaction data set. The frequent item set represents that in the user operation process, the combination probability of the simultaneously selected analysis metrics and data dimensions is greater than the preset threshold; based on the screened frequent item sets, calculate the confidence of all possible association rules; for each association rule, the calculation method of its confidence is: the proportion of the transactions that contain the rule consequent among the transactions that contain the rule antecedent; screen out the association rules with a confidence higher than the set threshold; according to the screening results, automatically generate a preliminary visual analysis recommendation scheme, covering chart types, data display dimensions, and color coding rules; Step S602: Users can customize and select the analysis data range, metrics, and display methods on the visual analysis platform. The platform intelligently integrates the user's custom settings with the pre-generated content: at the data level, according to the data range selected by the user, re-extract data from the personalized data mart M to update the data foundation; at the metric level, if the user adds a new metric, call the corresponding data in the data warehouse for calculation and merge the new metric with the pre-generated metrics; at the visual level, combine the chart types and display methods customized by the user to render and generate the final visual result in real time.

[0013] An enterprise information visualization management system based on data sharing, the system includes: an internal data processing module, an external data processing module, a data integration module, a data storage module, a personalized data and prediction module, and a visual analysis module; The internal data processing module collects user behavior data and business data by connecting to each business system within the enterprise, and performs preprocessing to build a structured internal data set; The external data processing module obtains data from external data sources, performs semantic parsing and feature extraction on unstructured data, conducts intelligent screening and classification in combination with a preset business rule library, extracts data related to the enterprise through a similarity matching algorithm, and builds an external data set; The data integration module integrates internal data and external data, formulates dynamic data cleaning rules, and performs multi-source data fusion through data association; The data storage module classifies and stores the integrated data according to different business requirements and analysis themes of the enterprise, forms an enterprise special database, and constructs a data warehouse to store data using a hierarchical architecture design; The personalized data and prediction module extracts data from the data warehouse, dynamically generates a personalized data mart based on user roles and permissions, automatically predicts user analysis requirements, and pre-generates analysis metrics and data dimensions; The visual analysis module constructs a visual analysis platform, generates a preliminary visual analysis recommendation scheme according to the pre-generated analysis metrics and data dimensions combined with the user operation track; on this basis, the user customizes and selects the analysis data range, metrics and methods according to needs, and the platform intelligently integrates the pre-generated content with the user-defined settings to generate the final visual result in real time.

[0014] Compared with the prior art, the beneficial effects achieved by the present invention are: The present invention comprehensively integrates and efficiently processes internal and external data of the enterprise by constructing a dual processing unit, and by means of technical means such as semantic parsing, feature extraction, similarity matching and dynamic data cleaning rules, significantly improves the quality and efficiency of data collection, processing and fusion, effectively meets the diverse data needs of the enterprise, and greatly improves the comprehensiveness and accuracy of data integration; The present invention realizes the refined management and personalized service of data by designing a data warehouse with a hierarchical architecture, dynamically generating a personalized data mart, and automatically predicting analysis requirements based on user roles and permissions; the hierarchical design of the data warehouse not only optimizes the data storage structure, but also improves the efficiency of data query and analysis; while the functions of the personalized data mart and automatically predicting analysis requirements enable the enterprise to provide personalized data services for different user roles, further enhancing the pertinence and practicality of data analysis; The present invention constructs a visual analysis platform, combines the pre-generated analysis metrics, data dimensions and user operation track, can intelligently generate a preliminary visual analysis recommendation scheme, and supports the user to make custom settings according to actual needs, realizing the intelligent integration of the pre-generated content and user needs. This innovative design not only improves the intuitiveness and usability of enterprise information management, but also enables the enterprise to quickly respond to market changes, accurately grasp business trends, and provide strong support for the strategic decision-making of the enterprise. Brief Description of the Drawings

[0015] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the drawings: Figure 1 It is a method flow chart of an enterprise information visualization management method based on data sharing. Detailed Embodiments

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] Please refer to Figure 1 , the present invention provides a technical solution: an enterprise information visualization management method based on data sharing, and the method includes: Step S100: Construct a first processing unit, collect user behavior data and business data by connecting various business systems within the enterprise, and perform preprocessing to construct an internal data set; Step S200: Construct a second processing unit, obtain data from external data sources, perform semantic parsing and feature extraction, perform intelligent screening and classification in combination with a preset business rule library, and extract data related to the enterprise through a similarity matching algorithm to construct an external data set; Step S300: Integrate the internal data and external data of the first processing unit and the second processing unit, formulate dynamic data cleaning rules, and perform multi-source data fusion through data association; Step S400: Classify and store the integrated data according to different business requirements and analysis topics of the enterprise to form an enterprise special database, and use a hierarchical architecture design to construct a data warehouse to store data; Step S500: Extract the data in the data warehouse, dynamically generate personalized data marts based on user roles and permissions, automatically predict user analysis requirements, and pre-generate analysis indicators and data dimensions; Step S600: Construct a visual analysis platform, generate a preliminary visual analysis recommendation plan according to the pre-generated analysis indicators and data dimensions in combination with the user operation track; on this basis, the user customizes and selects the analysis data range, indicators and methods according to the requirements, and the platform intelligently integrates the pre-generated content with the user's custom settings to generate the final visual result in real time.

[0018] Step S100 includes: Step S101: The first processing unit constructs an enterprise-level data bus based on the message queue middleware, establishes two-way communication with the enterprise business system through standardized RESTful APIs; collects user data and business data by combining timed polling and event triggering; the user data includes the operation logs and page access records of users in the system, records the page paths and stay times accessed by users, and clickstream data; sets each record in the user data as U1, U2,..., Um respectively, then the user data set is represented as U = {U1, U2,..., Um}, where Um represents the mth user record; the business data includes the business type, business process node information, business transaction data, resource usage data, and business association data obtained according to the enterprise business system; sets each record in the business data as B1, B2,..., Bn respectively; then the business data set is represented as B = {B1, B2,..., Bn}, where Bn represents the nth business record. Step S102: Use the K-means clustering algorithm to perform clustering analysis on the user operation logs to automatically identify abnormal access patterns; for time series data, use the ARIMA model to fill in missing values, use a hash function to judge the uniqueness of data records, and remove duplicate data records; set the unified standard for enterprise data, perform format conversion on the collected data, and process the numerical data through the Z-score standardization method; after preprocessing, construct the internal data set D = U ∪ B.

[0019] Step S200 includes: Step S201: Use the HTTP request library to send requests to the APIs of external data sources to obtain data; at the same time, for data sources that provide file download services, regularly download and parse files; for the obtained text data, process it through natural language processing technology, extract the key information and semantic features in the text, and convert the text data into vector representations; for image data, use a convolutional neural network to extract the feature vectors of the images. Step S202: Combine the preset business rule library to perform intelligent screening and classification on the processed external data, calculate the matching degree between the data and the business rules using cosine similarity, set the matching degree threshold, extract the data with the calculation result greater than the threshold and perform preprocessing, set the data records related to the enterprise obtained from the external data source and after screening and classification as E1, E2,..., Ek respectively; then the external data set is represented as E = {E1, E2,..., Ek}, where Ek represents the kth external data record; the preset business rule library includes keywords, classification criteria, and data screening conditions related to enterprise business.

[0020] Take an e-commerce enterprise as an example: Use an HTTP request library to send requests to the API of an industry information website to obtain text data such as recent policy adjustments in the e-commerce industry and analysis of consumption trends. Apply natural language processing techniques to extract key information such as "New Regulations on Live Streaming with Goods" and "Holiday Promotion Strategies", and convert them into vector representations. At the same time, regularly download CSV files related to user portraits from a cooperative data platform and parse out structured data such as age, region, and consumption preferences. In addition, obtain product display pictures of competitors through the API, and use a convolutional neural network to extract feature vectors such as the appearance of the product and price tags in the pictures. The business rule library preset by the enterprise contains keywords such as "Promotion Activities" and "Competitor Dynamics". Use cosine similarity to calculate the matching degree between the obtained data and the business rules, and set the threshold to 0.6. For example, a piece of text data about a competitor's upcoming anniversary promotion activity has a calculated matching degree of 0.7, which exceeds the threshold, so it is extracted. Then, perform preprocessing on it. Finally, together with other similar qualified data records E1, E2, etc., they jointly form an external dataset E. Step S300 includes: Step S301: Conduct field analysis on the internal dataset D = U ∪ B and the external dataset E. For each field, identify its corresponding relationships in different datasets, and establish a mapping relationship between the data through data dictionary matching. Construct a unified metadata warehouse for recording each data source, field definitions, and data quality. Step S302: Dynamically generate data cleaning rules according to the business type. First, construct a multi-layer rule template library, which is classified according to business type, data type, and cleaning target. Automatically identify the business type through the data source metadata, and screen out the applicable rules from the rule template library according to the business type to generate a candidate rule set R = {R1, R2,..., Ri}. Assign priorities to each rule, combine the candidate rules with the business data to generate the final cleaning rules. Use a fuzzy matching algorithm to perform cross-data-source association on relevant entities, set a similarity threshold, and when the matching degree of two entities exceeds this threshold, perform data merging. Use linear interpolation to align the time granularity of different data sources. Combine with the business rule library. When there are data conflicts between internal data and external data, give priority to the internal data status. Through the above steps, complete the fusion of multi-source data to form a fused dataset F.

[0021] Taking an e-commerce enterprise as an example, an internal data set D is obtained, which includes a user data set U (such as user purchase records, browsing preferences, etc.) and a business data set B (such as order information, inventory data, etc.). The external data set E comes from industry data provided by a cooperative market research institution. Analyze the fields in the internal and external data sets. For example, the "product ID" in the internal order table and the "product number" in the external industry data. Establish a mapping relationship through data dictionary matching and record this information in the metadata warehouse. Build a rule template library according to the e-commerce business type. If the identified business type is "product sales", filter out adaptation rules such as "removing duplicate order records" from the rule template library to generate a candidate rule set. Assign a higher priority to the "removing duplicate order records" rule and generate the final cleaning rule in combination with the business data. Use a fuzzy matching algorithm to associate the product information in the internal and external data, set the similarity threshold to 0.8, and merge the data if it exceeds this threshold. Use linear interpolation to unify the time granularity of different data sources to "days". When there is a conflict between the internal product inventory data and the external market inventory data surveyed, give priority to the internal data, and finally form a fused data set F.

[0022] Step S400 includes: Step S401: Set the enterprise business requirements and analysis topics through historical query logs and the settings of the business department; combine the field characteristics of the fused data set F to establish a mapping relationship table between the fields and the topics; based on the field mapping relationship, extract relevant fields from the fused data set F to generate an independent enterprise topic library T = {T1, T2,..., Tk}, where Tk represents the kth enterprise topic. When generating the topic library, take each business topic as the core and aggregate all related fields; One-to-many mapping: When a field is related to multiple topics, record multiple corresponding topics for this field in the mapping relationship table; for example: the "order number" field is related to both the "order processing efficiency analysis" topic and the "customer consumption behavior analysis" topic, then record these two groups of mapping relationships in the table respectively; Many-to-one mapping: When multiple fields jointly correspond to a topic, combine these fields and establish a mapping with this topic; for example, fields such as "customer name", "customer contact information", and "customer address" jointly serve the "customer information management" topic, and record this combined mapping relationship in the mapping table.

[0023] Step S402: Build a data warehouse using a layered architecture design, and divide the data warehouse into an operational data layer, a data integration layer, a summary data layer, and an application layer; the operational data layer directly stores the raw data obtained from the fusion dataset F as the basic data source of the data warehouse; the data integration layer performs mild aggregation processing on the raw data, identifies redundant fields, and at the same time performs standardization conversion on the data to unify the data format and coding rules; the summary data layer performs deep aggregation on the integrated data and conducts thematic modeling based on the business logic of the enterprise thematic library T; the application layer is used to store the enterprise thematic library T.

[0024] Step S500 includes: Step S501: Define a set of user roles R = {R1, R2,..., Rx}, where Rx represents the xth user role; at the same time, determine the corresponding set of permissions H = {H1, H2,..., Hx} for each user role, where Hx is the permission owned by role Rx; extract the data in the enterprise thematic library T from the application layer of the data warehouse; for each user role, only select the data that its permissions allow to access, organize the filtered data, and dynamically generate a personalized data mart M = {M1, M2,..., Mx}, where Mx is the personalized data mart generated for role Rx; Step S502: Extract the historical query features, operation behavior features, business environment features, and user attribute features of the user from the user dataset U; the historical query features represent the analysis metrics and data dimensions queried by the user in the past period of time, and encode them as one-hot vectors; the operation behavior features obtain the page jump path, data filtering operation, and chart interaction behavior through the user operation log and convert them into sequence data; the business environment features represent the current business stage of the enterprise and industry dynamic data, and perform normalization processing; the user attribute features represent the relevant attribute information of the user's department and position level, and are converted into numerical features through encoding; integrate all the extracted features to construct a model input vector; Step S503: Divide the user's historical behavior data into training samples according to time windows. Each sample contains an input feature vector and corresponding true query metrics and dimension labels. Construct an LSTM network layer to capture the time series features of user behavior, stack an attention mechanism layer, add a fully connected layer and a Softmax layer at the end of the network to output the probability distributions of metrics and dimensions. Use the Adam optimizer, with the cross-entropy loss function as the optimization objective, perform backpropagation on the training set, and continuously adjust the model parameters. Use the validation set to evaluate the trained model, calculate the accuracy, recall rate, and F1 value until all evaluation metrics of the model reach the preset standards. Deploy the trained model to the server of the visual analysis platform. Based on the trained user analysis requirement prediction model, receive the user's current operation data and business environment data in real time as inputs. The model outputs the probability distributions of the predicted analysis metrics and data dimensions. Set a probability threshold α, where 0 < α < 1, and filter out the analysis metrics and data dimensions whose predicted probability values exceed the threshold, and pre-generate an analysis metric set and a data dimension set.

[0025] Step S600 includes: Step S601: Construct a visual analysis platform. The underlying layer of the platform is connected to the personalized data mart M generated in Step S500, the pre-generated analysis metric set, and the data dimension set. At the same time, record the user's operation trajectory data on the platform through the user dataset. Convert the user's selection behaviors of different analysis metrics and data dimensions into a transaction dataset, where each transaction represents a user operation and contains the analysis metrics and data dimensions selected by the user in this operation. Set a support threshold and a confidence threshold. The support threshold is used to measure the frequency of the item set appearing in the transaction dataset, and the confidence threshold is used to measure the reliability of the association rule. Use the Apriori algorithm to find all frequent item sets that meet the support threshold from the transaction dataset. The frequent item set represents that in the user operation process, the combined probability of the simultaneously selected analysis metrics and data dimensions is greater than the preset threshold. Based on the filtered frequent item sets, calculate the confidence of all possible association rules. For each association rule, the calculation method of its confidence is: the proportion of the transactions that contain the rule consequent among the transactions that contain the rule antecedent. Filter out the association rules whose confidence is higher than the set threshold. According to the filtering results, automatically generate a preliminary visual analysis recommendation scheme, covering chart types, data display dimensions, and color coding rules. For example, if the user selects analysis metric A and data dimension B, then the user will also select chart type C. Calculate its confidence, that is, the proportion of the transactions that also select chart type C among the transactions that select analysis metric A and data dimension B. Finally, filter out the association rules whose confidence is higher than the set threshold. Step S602: The user can custom-select the analysis data range, metrics, and display methods on the visual analysis platform, and the platform will intelligently integrate the user's custom settings with the pre-generated content: At the data level, according to the data range selected by the user, data is re-extracted from the personalized data mart M to update the data foundation; at the metric level, if the user adds new metrics, the corresponding data in the data warehouse is called for calculation, and the new metrics are merged with the pre-generated metrics; at the visualization level, combined with the chart type and display method custom-defined by the user, the final visualization result is rendered and generated in real time.

[0026] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced by the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. An enterprise information visualization management method based on data sharing, characterized in that: The method includes: Step S100: Construct a first processing unit, collect user behavior data and business data by connecting various business systems within the enterprise, perform preprocessing, and construct an internal data set; Step S200: Construct a second processing unit, obtain data from external data sources, perform semantic parsing and feature extraction, perform intelligent screening and classification in combination with a preset business rule library, extract enterprise-related data through a similarity matching algorithm, and construct an external data set; Step S300: Integrate the internal data and external data of the first processing unit and the second processing unit, formulate dynamic data cleaning rules, and perform multi-source data fusion through data association; Step S400: Classify and store the integrated data according to different business requirements and analysis topics of the enterprise, form an enterprise special topic library, and use a hierarchical architecture design to construct a data warehouse to store data; Step S500: Extract data from the data warehouse, dynamically generate personalized data marts based on user roles and permissions, automatically predict user analysis requirements, and pre-generate analysis metrics and data dimensions; Step S600: Construct a visual analysis platform, generate a preliminary visual analysis recommendation plan according to the pre-generated analysis metrics and data dimensions in combination with the user operation track; on this basis, the user customizes and selects the analysis data range, metrics and methods according to the requirements, and the platform will perform intelligent fusion of the pre-generated content and the user's custom settings to generate the final visual result in real time.

2. The enterprise information visualization management method based on data sharing according to claim 1, characterized in that: The step S100 includes: Step S101: The first processing unit constructs an enterprise-level data bus based on the message queue middleware, establishes two-way communication with the enterprise business system through the standardized RESTful API; adopts a combination of timed polling and event triggering to collect user data and business data; the user data includes the operation logs and page access records of the user within the system, records the page paths and residence times visited by the user, and clickstream data; set each record in the user data as U1, U2,..., Um respectively, then the user data set is expressed as U={U1, U2,..., Um}, where Um represents the mth user record; the business data includes obtaining business type, business process node information, business transaction data, resource usage data, and business association data according to the enterprise business system; set each record in the business data as B1, B2,..., Bn respectively; then the business data set is expressed as B={B1, B2,..., Bn}, where Bn represents the nth business record; Step S102: Use the K-means clustering algorithm to perform clustering analysis on the user operation logs, and automatically identify abnormal access patterns; for time series data, use the ARIMA model to fill in missing values, judge the uniqueness of data records through a hash function, and remove duplicate data records; set the unified standard for enterprise data, perform format conversion on the collected data, and process numerical data through the Z-score standardization method; after preprocessing, construct the internal data set D = U ∪ B.

3. A method for visual management of enterprise information based on data sharing according to claim 1, characterized in that: The step S200 includes: Step S201: Send a request to the API of the external data source using an HTTP request library to obtain data. At the same time, for data sources that provide file download services, regularly download and parse the files. For the obtained text data, process it through natural language processing techniques to extract key information and semantic features in the text, and convert the text data into a vector representation. For image data, use a convolutional neural network to extract the feature vectors of the images. Step S202: Combine the preset business rule library to perform intelligent screening and classification on the processed external data. Calculate the matching degree between the data and the business rules using cosine similarity, set a matching degree threshold, extract the data with a calculation result greater than the threshold and perform preprocessing. Set the data records related to the enterprise obtained from the external data source and after screening and classification as E1, E2,..., Ek. Then the external data set is represented as E = {E1, E2,..., Ek}, where Ek represents the kth external data record. The preset business rule library includes keywords related to enterprise business, classification criteria, and data screening conditions.

4. A method for visual management of enterprise information based on data sharing according to claim 1, characterized in that: The step S300 includes: Step S301: Perform field analysis on the internal data set D = U ∪ B and the external data set E. For each field, identify its corresponding relationship in different data sets, and establish a mapping relationship between the data through data dictionary matching. Build a unified metadata warehouse to record each data source, field definitions, and data quality. Step S302: Dynamically generate data cleaning rules according to the business type. First, build a multi-layer rule template library, classified by business type, data type, and cleaning target. Automatically identify the business type through the data source metadata, screen and adapt the rules from the rule template library according to the business type to generate a candidate rule set R = {R1, R2,..., Ri}, where Ri represents the ith candidate rule. Assign priorities to each rule, combine the candidate rules with the business data to generate the final cleaning rules. Use a fuzzy matching algorithm to perform cross-data source association on relevant entities, set a similarity threshold, and when the matching degree of two entities exceeds the threshold, perform data merging. Use linear interpolation to align the time granularity of different data sources. Combine the business rule library, and when there are data conflicts between internal data and external data, give priority to the internal data status. Complete the fusion of multi-source data through the above steps to form a fused data set F.

5. A method for visual management of enterprise information based on data sharing according to claim 1, characterized in that: The step S400 includes: Step S401: Set the enterprise business requirements and analysis topics through historical query logs and settings of business departments. Combine the field characteristics of the fused data set F to establish a mapping relationship table between fields and topics. Based on the field mapping relationship, extract relevant fields from the fused data set F to generate an independent enterprise topic library T = {T1, T2,..., Tk}, where Tk represents the kth enterprise topic. When generating the topic library, take each business topic as the core and aggregate all related fields. Step S402: Build a data warehouse using a layered architecture design, and divide the data warehouse into an operational data layer, a data integration layer, a summarized data layer, and an application layer; the operational data layer directly stores the original data obtained from the fusion dataset F as the basic data source of the data warehouse; the data integration layer performs mild aggregation processing on the original data, identifies redundant fields, and at the same time performs standardization conversion on the data to unify the data format and coding rules; the summarized data layer deeply aggregates the integrated data and performs thematic modeling based on the business logic of the enterprise thematic library T; the application layer is used to store the enterprise thematic library T.

6. A method for visual management of enterprise information based on data sharing according to claim 1, characterized in that: The step S500 includes: Step S501: Define a set of user roles R = {R1, R2,..., Rx}, where Rx represents the xth user role; at the same time, determine the corresponding set of permissions H = {H1, H2,..., Hx} for each user role, where Hx is the permission owned by role Rx; extract the data in the enterprise thematic library T from the application layer of the data warehouse; for each user role, only select the data that its permissions allow to access, organize the filtered data, and dynamically generate a personalized data mart M = {M1, M2,..., Mx}, where Mx is the personalized data mart generated for role Rx; Step S502: Extract the historical query features, operation behavior features, business environment features, and user attribute features of the user from the user dataset U; the historical query features represent the analysis metrics and data dimensions queried by the user in the past period of time, and encode them as one-hot vectors; the operation behavior features obtain the page jump path, data filtering operation, and chart interaction behavior through the user operation log and convert them into sequence data; the business environment features represent the current business stage of the enterprise and industry dynamic data and perform normalization processing; the user attribute features represent the relevant attribute information of the user's department and position level and are converted into numerical features through encoding; integrate all the extracted features to construct a model input vector; Step S503: Divide the user's historical behavior data into training samples according to time windows. Each sample contains an input feature vector and corresponding true query metrics and dimension labels. Construct an LSTM network layer to capture the time series features of user behavior, stack an attention mechanism layer, add a fully connected layer and a Softmax layer at the end of the network to output the probability distributions of metrics and dimensions. Use the Adam optimizer, with the cross-entropy loss function as the optimization objective, perform backpropagation on the training set, and continuously adjust the model parameters. Use the validation set to evaluate the trained model, calculate the accuracy, recall rate, and F1 value until all evaluation metrics of the model reach the preset standards. Deploy the trained model to the server of the visualization analysis platform. Based on the trained user analysis requirement prediction model, receive the user's current operation data and business environment data in real time as inputs. The model outputs the probability distributions of the predicted analysis metrics and data dimensions. Set a probability threshold α, where 0 < α < 1, and filter out the analysis metrics and data dimensions whose predicted probability values exceed the threshold to pre-generate an analysis metric set and a data dimension set.

7. A method for visual management of enterprise information based on data sharing according to claim 1, characterized in that: The said step S600 includes: Step S601: Construct a visualization analysis platform. The bottom layer of the platform is connected to the personalized data mart M generated in step S500, the pre-generated analysis metric set, and the data dimension set. At the same time, record the operation trajectory data of the user on the platform through the user dataset. Convert the user's selection behavior of different analysis metrics and data dimensions into a transaction dataset, where each transaction represents a user operation and contains the analysis metrics and data dimensions selected by the user in this operation. Set the support threshold and the confidence threshold. Use the Apriori algorithm to screen all frequent item sets that meet the support threshold from the transaction dataset. The frequent item set means that during the user's operation, the combined probability of the simultaneously selected analysis metrics and data dimensions is greater than the preset threshold. Based on the screened frequent item sets, calculate the confidence of all possible association rules. For each association rule, the calculation method of its confidence is: among the transactions containing the antecedent of the rule, the proportion of the transactions that also contain the consequent of the rule. Screen out the association rules whose confidence is higher than the set threshold. According to the screening results, automatically generate a preliminary visualization analysis recommendation plan, covering the chart type, data display dimension, and color coding rules. Step S602: The user can customize the selection of the analysis data range, metrics, and display methods on the visualization analysis platform. The platform intelligently integrates the user's customized settings with the pre-generated content: at the data level, according to the data range selected by the user, re-extract data from the personalized data mart M to update the data basis; at the metric level, if the user adds new metrics, call the corresponding data in the data warehouse for calculation and merge the new metrics with the pre-generated metrics; at the visualization level, combine the chart type and display method customized by the user to render and generate the final visualization result in real time.

8. An enterprise information visualization management system based on data sharing, characterized in that: The system includes: an internal data processing module, an external data processing module, a data integration module, a data storage module, a personalized data and prediction module, and a visualization analysis module; The internal data processing module collects user behavior data and business data by connecting to various business systems within the enterprise, performs preprocessing, and constructs a structured internal data set; The external data processing module obtains data from external data sources, performs semantic parsing and feature extraction on unstructured data, conducts intelligent screening and classification in combination with a preset business rule library, extracts data related to the enterprise through a similarity matching algorithm, and constructs an external data set; The data integration module integrates internal data and external data, formulates dynamic data cleaning rules, and performs multi-source data fusion through data association; The data storage module classifies and stores the integrated data according to different business requirements and analysis themes of the enterprise, forms an enterprise special topic library, and constructs a data warehouse to store data using a hierarchical architecture design; The personalized data and prediction module extracts data from the data warehouse, dynamically generates a personalized data mart based on user roles and permissions, automatically predicts user analysis requirements, and pre-generates analysis metrics and data dimensions; The visualization analysis module constructs a visualization analysis platform, generates a preliminary visualization analysis recommendation plan according to the pre-generated analysis metrics and data dimensions in combination with the user operation track; on this basis, the user customizes and selects the analysis data range, metrics, and methods according to the requirements, and the platform intelligently fuses the pre-generated content with the user-defined settings to generate the final visualization result in real time.

Images