Data visualization large screen automatic generation method and generation system

Abstract

The invention discloses a method and a system for automatically generating a data visualization large screen. The method comprises the following steps of: A, preparing configuration of a visualization large screen template; b, inputting a question by a user, selecting a template, and configuring the input question and the template; c, outputting a data table most relevant to the user problem after analysis; d, extracting related fields; e, in combination with the obtained field set most related to the input problem, breadth-first traversal is conducted on the visual large-screen template configuration selected by the user, target charts in the visual large-screen template configuration are sequentially processed, and in the process of traversal of the visual large-screen template configuration selected by the user, each target chart object enters a chart generation circulation internal process; f, starting a fault-tolerant mechanism in a chart generation process; and G, generating a visual large screen which is consistent with the selected template in configuration style and style and has data related to the user problem through a visual large screen renderer. The problems of low automation degree, weak analysis effectiveness and poor expansibility in a data visualization large screen generation process are solved.

Description

Technical Field

[0001] This invention relates to the field of browsing technology, and in particular to a method and system for automatically generating large data visualization screens. Background Technology

[0002] The automatic generation method of data visualization dashboards refers to a set of technical paths and solutions that utilize computer technology, artificial intelligence and software engineering principles to automatically integrate and configure data sources, visualization components and design specifications through preset rules, algorithms or intelligent models. This allows for the rapid generation of data visualization dashboards with good visual effects and interactive functions with little or no manual coding and design.

[0003] Existing technologies in this area focus on automating layout or relying on manual data configuration, failing to simultaneously address: 1. High visual design costs; 2. Lack of intelligent data association; 3. Difficulty in adapting to multiple data sources. For example, the invention patent with patent application number "CN111208998A" uses image recognition and segmentation technology to analyze design files (such as PSD / Sketch), extracting features (such as chart types) and location information of the main objects. The recognition results are then mapped to a standardized component library to generate visual components. The components are laid out according to the location information in the original design file, ultimately generating code. However, its drawback lies in relying on image recognition rather than semantic understanding, failing to dynamically select data based on user questions, merely copying the visual layout of the design draft, lacking "dimension-indicator" mapping logic, and lacking data association analysis capabilities.

[0004] Existing technologies, such as the invention patent with patent application number "CN114138222A", specifically involve dynamically adjusting the module size according to the large screen resolution (allocating more space to important modules) and adopting a layout strategy that expands from the center to both sides to support the addition of interactive functions after user verification. It can be seen that the above solution also has obvious defects. It mainly focuses on physical resolution adaptation rather than data analysis, relies on users to manually input visual content, and only supports planar layout templates.

[0005] Existing technologies, such as the invention patent with patent application number "CN116610404A", disclose two modes by matching database fields through a thesaurus (general / industry / self-learning): fully automatic generation or user-selected data combinations that allow for later fine-tuning of chart styles. This approach is more mature than the previous two, but it also has certain limitations. First, it requires user participation in field selection; second, it only controls the number / type templates of charts; and finally, it does not mention a degradation strategy, which leads to poor compatibility.

[0006] In summary, it can be seen that the existing technology is very immature. The core problems are low automation (relying on manual input), weak analytical effectiveness (mechanical matching of data and charts), and poor scalability (not supporting complex layouts and dynamic adjustments). Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for automatically generating data visualization dashboards, which solves the problems of low automation, weak analysis effectiveness, and poor scalability in the existing data visualization dashboard generation process, resulting in difficulties and time consumption in the design of visualization dashboards.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0009] A method for automatically generating data visualization dashboards includes the following steps: A. Preparing visualization dashboard template configurations; B. The user inputs a question and selects a preferred template from a set of visualization dashboard template configurations, using the user-input question and the selected visualization dashboard template configuration as input for subsequent processing; C. Understanding the user-input question using semantic understanding technology and analyzing it in conjunction with a set of data tables provided by the data management module, outputting the data table most relevant to the user's question; D. Obtaining the field set of the output data table, and again combining it with the user-input question to obtain the field set most relevant to the user's question; E. Using the obtained field set most relevant to the user's question, performing a breadth-first traversal of the user-selected visualization dashboard template configurations, processing the target charts sequentially, and entering the chart generation loop for each target chart object during the traversal of the user-selected visualization dashboard template configurations; F. Activating a fault tolerance mechanism during the chart generation process; G. Finally, generating a visualization dashboard with a style and format consistent with the selected template configuration and whose data is relevant to the user's question through a visualization dashboard renderer.

[0010] As a further preferred embodiment of the present invention, the preparation of the visualization large screen template configuration in step A specifically involves pre-constructing a set containing several visualization large screen template configurations. Each visualization large screen template configuration is a list consisting of several target chart objects. Each target chart object consists of an id and a type, where the id is associated with the rendering configuration of the visualization large screen renderer; and the type is the target chart type supported by the visualization large screen renderer.

[0011] The rendering configuration includes basic style parameters (CSS / JSON), layout constraints (grid coordinates / responsive rules), data rendering specifications (responsible for converting query data into corresponding charts by dimension and indicator, such as the x-axis as the dimension and the y-axis as the indicator), and other information. The type is the target chart type supported by the visualization dashboard renderer, such as data table, vertical bar chart, horizontal bar chart, indicator, pie chart, etc. It also includes a special group (chart group) type. The chart object of the group type contains a charts array to store the specific charts in the chart group, and there is no limit to the nesting depth of the group and the number of sub-charts, supporting complex dashboard nesting designs.

[0012] Below is an example of configuring a visual dashboard template:

[0013] [{

[0014] "id": "c45a5e1a",

[0015] "type": "Data Table"

[0016] }, {

[0017] "id": "e5262026",

[0018] "type": "Vertical bar chart"

[0019] }, {

[0020] "id": "b6d5ae2e",

[0021] "type": "Horizontal bar chart"

[0022] }, {

[0023] "id": "f1dd10f2-26a7-4f99-bea5-f4ee7fbbe593",

[0024] "type": "group",

[0025] "charts": [ {

[0026] "id": "a9aec4a0",

[0027] "type": "indicator"

[0028] }, {

[0029] "id": "a320807a",

[0030] "type": "indicator"

[0031] } ]

[0033] }, {

[0034] "id": "6f8c864d-2a27-4ad6-bc93-14d69c75309c",

[0035] "type": "group",

[0036] "charts": [{

[0037] "id": "7df96769",

[0038] "type": "indicator"

[0039] }, {

[0040] "id": "61e86f03",

[0041] "type": "pie chart"

[0042] } ]

[0044] } ]

[0046] As a further preferred embodiment of the present invention, the rendering configuration includes basic style parameters, layout constraints, and data rendering specifications. The type includes data tables, vertical bar charts, horizontal bar charts, indicators, pie charts, and also includes a special group type. The chart object of the group type contains a charts array to store the specific charts in the chart group, and there is no limit to the nesting depth of the group and the number of sub-charts, supporting complex dashboard nesting designs.

[0047] As a further preferred embodiment of the present invention, in step B, the user input question can be input in a manner that supports both natural language and structured input, and the intelligent semantic analysis module can select an appropriate algorithm based on the input.

[0048] Input methods support both natural language and other languages, such as "Analyze the network alarm situation in the last 5 minutes" and "Product sales in the last three months".

[0049] As a further preferred embodiment of the present invention, before selecting a preferred template from the set of visualization large screen template configurations, the visualization large screen template configuration is displayed as an image with actual effect.

[0050] Displaying the visual dashboard template configurations as images with actual effects makes it easier for users to select a preferred template from the collection of visual dashboard template configurations.

[0051] As a further preferred embodiment of the present invention, the data table set in step C is an accessible data table set provided by the data management module. The user-input question and the accessible data table set provided by the data management module are received by the intelligent semantic analysis module. The intelligent semantic analysis module is configured to extract key semantic features from the input data through semantic understanding technology. It includes a text vectorization unit based on a word embedding model, a semantic matching unit using an attention mechanism, and a feature enhancement unit containing a domain knowledge graph.

[0052] The data management module provides an accessible collection of data tables, which contains basic information about multiple tables, such as table names and descriptions. During implementation, the semantic analysis module can include a preprocessing layer (using an embedding model to generate question embedding vectors), a feature association layer (calculating the matching degree between field descriptions and questions using cosine similarity), a filtering layer (applying a knowledge graph-based rule engine to filter incompatible fields), and a decision layer (applying an attention mechanism model for semantic matching and returning the optimal decision solution), ultimately outputting the data table most relevant to the user's question.

[0053] As a further preferred embodiment of the present invention, in step D, obtaining the field set of the output data table is completed by the data management module, and the output field set of the data table contains basic information of several fields; then, in combination with the user input question, obtaining the field set most relevant to the user input question is completed by the intelligent semantic analysis module, and the obtained field set most relevant to the user question is a subset of the field set of the data table.

[0054] As a further preferred embodiment of the present invention, the internal process of the chart generation loop in step E is as follows: e1. Start traversing the template configuration, determine whether there are any unprocessed chart objects, and if so, read the current chart object; e2. Call the statistical method generation module, input the current target chart type and database type, and return a set of "dimension type - indicator type" objects compatible with the chart type and database type; call the chart generation module, receive the set of "dimension type - indicator type" objects and a subset of the field set of the data table, and output a set of "dimension type - dimension field - indicator type - indicator field" objects; e3. Call the intelligent semantic analysis module, receive the user question and the set of "dimension type - dimension field - indicator type - indicator field" objects, and output "dimension type - dimension field - indicator type - indicator field" objects closely related to the user's concerns, which serve as the query conditions for the target chart; e4. The memory module uses a Hash algorithm to process the generated "dimension type - dimension field - indicator type - indicator field" objects. e5. Store the "Indicator Field" object to remove duplicates and avoid chart repetition; e6. Pass the query conditions to the visualization dashboard renderer for chart rendering and display; e7. Repeat the above steps until all chart objects have been processed.

[0055] The data management module supports multiple data sources, including structured databases such as MySQL and H2, and unstructured databases such as Elasticsearch and Clickhouse. The differences between different database types are mainly handled by the statistical method generation module to ensure that the returned dimension and indicator types are supported by the database.

[0056] As a further preferred embodiment of the present invention, the fault tolerance mechanism is applied as follows: during the chart generation process, if a set of "dimension type - dimension field - indicator type - indicator field" objects compatible with the target chart type cannot be generated, the fault tolerance mechanism is activated. First, a substitute of the same type is searched. If the substitute fails, the display is downgraded to the basic chart.

[0057] A system for automatically generating data visualization dashboards according to any one of the above methods includes a template configuration module, a user input module, a user selection module, a data management module, an intelligent semantic analysis module, a processing module, a memory module, a fault-tolerant module, and a visualization dashboard renderer. The template configuration module is used to pre-build a set containing multiple visualization dashboard template configurations. The user input module receives user-input questions, which can be in natural language or structured input. The template selection module displays the visualization dashboard template configurations as images for user selection and receives the selected template configuration. The intelligent semantic analysis module is configured to extract key semantic features from the input using semantic understanding technology, including a preprocessing layer, a feature association layer, a filtering layer, and a decision layer. The data management module is used to... The system provides basic information and field sets from multiple data tables. The data management module supports various data sources, including structured and unstructured databases. The processing module performs the following processing based on the user-input question and the selected template configuration: a) Selecting the data table most relevant to the user's question from multiple data tables provided by the data management module through the intelligent semantic analysis module; b) Extracting the field set from the selected data table and selecting the subset of fields most relevant to the user's question from the field set through the intelligent semantic analysis module; c) Performing a breadth-first traversal of the target chart objects in the selected template configuration. For each target chart object, it calls the statistical method generation module and the chart generation module to generate query conditions and selects the most relevant object as the query condition through the intelligent semantic analysis module. The memory module stores and deduplicates the query conditions based on a hash algorithm. The fault tolerance module activates a fault tolerance mechanism if, during the chart generation process, a set of dimension and indicator types compatible with the target chart type cannot be generated. It first attempts to replace the current chart type with a substitute chart type of the same type. If the substitution fails, it is downgraded to the basic chart type for display. The visualization dashboard renderer renders the chart according to the query conditions and generates a visualization dashboard after all target chart objects have been processed.

[0058] Existing technology: Relying solely on keyword matching (e.g., binding a bar chart to a field name containing "sales") cannot understand the user's deeper analytical intent (e.g., time series charts should be prioritized for "trend analysis"). This application employs an embedded model + large language model + domain knowledge graph, which can identify analytical intents such as "comparison," "distribution," and "trend." Example: When a user enters "sales changes in various regions over the past six months," the system automatically selects: Time dimension → line chart (instead of the fixed bar chart of existing technology) and Region dimension → color classification. Improved performance: Chart type matching accuracy increased by 40%+.

[0059] Existing technology: The template is a fixed combination of charts and cannot be nested (e.g., it cannot package "sales indicators + regional distribution map" into an analysis unit). This application supports infinitely nested group types, which improves the layout efficiency of complex analysis scenarios by 3 times and reduces the number of manual adjustments.

[0060] Existing technology: When the field type is incompatible with the chart (e.g., a text field is used for a bar chart), the process is terminated directly with an error. This application adopts a three-level degradation strategy: ① Replace with the same type (bar chart → column chart), ② Simplify the chart type (3D pie chart → flat pie chart), ③ Finally downgrade to a data table. The success rate of large screen generation is increased from 68% in the existing technology to 99%.

[0061] Existing technology: Assuming all databases support the same SQL syntax (e.g., attempting to use window functions on the H2 database will result in an error), the statistical method generation module of this application dynamically adapts, thereby expanding the types of heterogeneous data sources supported by 5 times.

[0062] Existing technology: Each generation is an independent process, which may repeatedly generate the same analysis dimension (such as multiple charts using the "date" dimension). The memory module of this application records the used analysis dimension through hash fingerprint technology: hash("date-sales-sum") = "a1b2c3d4" # Stored in the memory pool, which increases the information density of the large screen by 30% and avoids redundant charts.

[0063] The statistical method generation module can use the following algorithm to take the current target chart type and database type as input and return a collection of "dimension type - indicator type" objects that are compatible with the chart type and database type;

[0064] 1) Dimension type and indicator type.

[0065] Dimension type refers to the way descriptive attributes (i.e., dimensions) are classified in data analysis. Common dimension types include grouping by value, grouping by time period, and grouping by numerical range. Let the complete set of dimension types be: .

[0066] Indicator types refer to the statistical methods used to calculate quantifiable and calculable numerical data in data analysis. Common indicator types include summation, averaging, maximization, minimization, and percentile calculation. Let the complete set of dimension types be: .

[0067] 2) Calculate the dimension types and indicator types supported by the input database type.

[0068] Assuming the target chart type to be processed is t, the dimensions and metrics of this chart type can be defined from the data dictionary based on t, including the number of dimensions: Number of indicators: and the types of metrics supported: ;

[0069] Assuming the current database type is "database", the definitions of dimensions and metrics for this data type can be obtained from the data dictionary based on the database type, including... and the types of metrics supported: .

[0070] Therefore, by performing a simple intersection operation, we can obtain the supported dimension types for the chart type t in the database:

[0071] .

[0072] Supported metric types:

[0073] .

[0074] 3) Returns a collection of "Dimension Type - Indicator Type" objects that are compatible with both chart and database types.

[0075] Permutation calculations can be used to obtain sets of all permutations of different elements from the supported dimension types and supported indicator types, respectively. Performing a Cartesian product on the two sets yields a set of "dimension type - indicator type" objects.

[0076] Dimensional arrangement of sets (ordered non-repeating sampling):

[0077] .

[0078] Index permutation set (ordered non-repeating sampling):

[0079] .

[0080] Complete set of statistical methods (Cartesian product):

[0081] .

[0082] in .

[0083] The chart generation module's algorithm can be implemented by receiving a collection of "dimension type - indicator type" objects and a subset of the data table's field set, and outputting a collection of "dimension type - dimension field - indicator type - indicator field" objects:

[0084] Suppose the subset of fields in the data table is: .

[0085] It is worth mentioning that each field It can be represented as a metadata object that includes field names, field aliases, and field types, for example: These metadata will play a role in subsequent semantic modules.

[0086] The set of fields supported by the calculation dimension type and indicator type.

[0087] Define a validity check function , indicating dimension type d and indicator type Compatibility:

[0088]

[0089] This function can generate a mapping relationship between any dimension type or indicator type and its legal field subset:

[0090] Mapping relationship between dimension type and its valid field subset: .

[0091] Mapping relationship between indicator types and their legal field subsets: .

[0092] Based on the above mapping relationship, given any dimension type d, Given a subset of its valid fields; given any index type m, It is a subset of its valid fields.

[0093] Calculate the collection of objects named "Dimension Type - Dimension Field - Metric Type - Metric Field".

[0094] For any "Dimension Type - Indicator Type" object For each dimension type and indicator type, select one field from the subset of valid fields, and ensure that the fields are not repeated. This will result in an object named "Dimension Type - Dimension Field - Indicator Type - Indicator Field".

[0095] .

[0096] Where i=1… j=1… .

[0097] For a collection of "Dimension Type - Indicator Type" objects, a corresponding collection of "Dimension Type - Dimension Field - Indicator Type - Indicator Field" objects can be generated.

[0098] The aforementioned technological advantages enable this application to significantly surpass existing solutions in terms of automation, analytical effectiveness, and practical applicability.

[0099] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0100] 1. This application significantly surpasses existing solutions in terms of automation, analytical effectiveness, and practical applicability. Through a systematic design that incorporates semantic analysis, templated hierarchical configuration, and intelligent fault tolerance mechanisms, it achieves truly "problem-oriented" automated data visualization generation.

[0101] 2. The memory module records the used "dimension-metric" combinations based on the Hash algorithm, automatically filtering out duplicate analysis logic to ensure that the content on the large screen is compact and information-rich.

[0102] 3. Template-driven + semantic analysis actively generates large screens. Users only need to input a question and select a template, without needing to prepare design materials.

[0103] 4. A three-level mapping mechanism (dimension type → field → indicator) ensures that the chart type strictly matches the data characteristics and analysis objectives.

[0104] 5. Integrate user question semantics throughout the entire process of data table selection, field filtering, and chart matching to ensure that analysis conclusions are directly presented on the large screen. Attached Figure Description

[0105] Figure 1 The flowchart for automatically generating a large data analysis visualization dashboard according to the present invention.

[0106] Figure 2 This is a flowchart illustrating the process of the target chart object entering the chart generation loop in this invention. Detailed Implementation

[0107] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0108] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0109] It should be noted that, unless otherwise specified, the embodiments and features described in this invention can be combined with each other.

[0110] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0111] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are only used for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0112] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. Specific Implementation Example 1:

[0114] Figure 1 , Figure 2 This paper presents a method for automatically generating data visualization dashboards, comprising the following steps: A) Preparing visualization dashboard template configurations; B) The user inputs a question and selects a preferred template from a set of visualization dashboard template configurations, using the user-input question and the selected visualization dashboard template configuration as input for subsequent processing; C) Understanding the user-input question using semantic understanding technology and analyzing it in conjunction with a set of data tables provided by the data management module, outputting the data table most relevant to the user's question; D) Obtaining the field set of the output data table, and again combining it with the user-input question to obtain the field set most relevant to the user's question; E) Using the obtained field set most relevant to the user's question, performing a breadth-first traversal of the user-selected visualization dashboard template configurations, processing the target charts sequentially, and entering the chart generation loop for each target chart object during the traversal of the user-selected visualization dashboard template configurations; F) Activating a fault tolerance mechanism during the chart generation process; G) Finally, generating a visualization dashboard with a style and format consistent with the selected template configuration and whose data is relevant to the user's question through a visualization dashboard renderer.

[0115] Compared to the technical solutions in patent numbers "CN116610404A" and "CN114138222A", this solution adopts intelligent semantic analysis-driven approach. Through an embedding model, knowledge graph, and attention mechanism, it extracts analytical targets (such as "trends" and "comparisons") from user questions and dynamically matches data with chart types. This ensures that the visualization dashboard truly meets the user's analytical needs, solving the core defect of existing technologies where "generated results are disconnected from analytical intent." This defect is particularly evident in patent number "CN116610404A". In its fully automatic mode, Lizhong only matches field names and chart types using a pre-set thesaurus, failing to understand the deeper semantics of user queries. For example, when a user inputs "analyze sales trends in various regions," it might mechanically match only the "sales" field, ignoring the time-series analysis requirement corresponding to "trend," incorrectly generating a static table instead of a line chart. While the semi-automatic mode allows for manual adjustments, it is inefficient and does not meet the goals of automation design. Secondly, the solution in patent number "CN116610404A" uses a flat matching method, such as directly mapping field names to chart types, while this solution employs a three-level mapping mechanism of dimension-indicator-field, ensuring strict compatibility between charts and data. To avoid the problem of "database type mismatch leading to generation failure" in existing technologies, the following is implemented: Dimension type (e.g., grouping, time) → Field (e.g., "region", "date"); Indicator type (e.g., summation, average) → Field (e.g., "sales revenue"). This is dynamically adapted to database characteristics (e.g., ClickHouse supports quantile calculation). Furthermore, this application supports nested template configurations, allowing for infinite levels of nesting through group types and chart arrays. This satisfies complex analysis scenarios (e.g., "total-to-partial" analysis), overcoming the limitations of existing technologies in terms of large-screen layout complexity. Existing technologies only support flat chart combinations. The foreign application also introduces fault tolerance and degradation mechanisms, supporting automatic replacement of chart types (such as bar chart → column chart) or degradation to basic table display, ensuring that the large screen is always available. This effectively solves the problem of direct error reporting when incompatible in "CN116610404A", significantly improving system robustness and addressing the pain point of "overall generation failure due to local errors" in existing technologies. Finally, this solution also dynamically identifies database characteristics (such as MySQL not supporting quantile), enabling cross-database statistical method adaptation. This is completely different from the assumption of consistent database functions in existing technologies, expanding the applicability of the technical solution and supporting heterogeneous data source scenarios. Specific Implementation Example 2:

[0117] This embodiment further explains step A based on specific embodiment 1. The preparation of the visualization large screen template configuration in step A specifically involves pre-constructing a set containing several visualization large screen template configurations. Each visualization large screen template configuration is a list composed of several target chart objects. Each target chart object consists of an id and a type, where the id is associated with the rendering configuration of the visualization large screen renderer; and the type is the target chart type supported by the visualization large screen renderer.

[0118] The rendering configuration includes basic style parameters (CSS / JSON), layout constraints (grid coordinates / responsive rules), data rendering specifications (responsible for converting query data into corresponding charts by dimension and indicator, such as the x-axis as the dimension and the y-axis as the indicator), and other information. The type is the target chart type supported by the visualization dashboard renderer, such as data table, vertical bar chart, horizontal bar chart, indicator, pie chart, etc. It also includes a special group (chart group) type. The chart object of the group type contains a charts array to store the specific charts in the chart group, and there is no limit to the nesting depth of the group and the number of sub-charts, supporting complex dashboard nesting designs.

[0119] Below is an example of configuring a visual dashboard template:

[0120] [{

[0121] "id": "c45a5e1a",

[0122] "type": "Data Table"

[0123] }, {

[0124] "id": "e5262026",

[0125] "type": "Vertical bar chart"

[0126] }, {

[0127] "id": "b6d5ae2e",

[0128] "type": "Horizontal bar chart"

[0129] }, {

[0130] "id": "f1dd10f2-26a7-4f99-bea5-f4ee7fbbe593",

[0131] "type": "group",

[0132] "charts": [ {

[0133] "id": "a9aec4a0",

[0134] "type": "indicator"

[0135] }, {

[0136] "id": "a320807a",

[0137] "type": "indicator"

[0138] } ]

[0140] }, {

[0141] "id": "6f8c864d-2a27-4ad6-bc93-14d69c75309c",

[0142] "type": "group",

[0143] "charts": [{

[0144] "id": "7df96769",

[0145] "type": "indicator"

[0146] }, {

[0147] "id": "61e86f03",

[0148] "type": "pie chart"

[0149] } ]

[0151] } ] Specific Implementation Example 3:

[0154] This embodiment further explains the rendering configuration based on specific embodiment 2. The rendering configuration includes basic style parameters, layout constraints, and data rendering specifications. The type includes data table, vertical bar chart, horizontal bar chart, indicator, pie chart, and also includes a special group type. The chart object of the group type contains a charts array to store the specific charts in the chart group. The nesting depth of the group and the number of sub-charts are unlimited, supporting complex dashboard nesting designs. Specific Implementation Example 4:

[0156] This embodiment further explains step B based on specific embodiment 1. In step B, the user inputs a question, and the input method supports both natural language and structured input. The intelligent semantic analysis module can select an appropriate algorithm based on the input.

[0157] Input methods support both natural language and other languages, such as "Analyze the network alarm situation in the last 5 minutes" and "Product sales in the last three months". Specific Implementation Example 5:

[0159] This embodiment further explains the set of visualization large screen template configurations based on specific embodiment 1. Before selecting a preferred template from the set of visualization large screen template configurations, the visualization large screen template configuration is displayed as an image with actual effect.

[0160] Displaying the visual dashboard template configurations as images with actual effects makes it easier for users to select a preferred template from the collection of visual dashboard template configurations. Specific Implementation Example 6:

[0162] This embodiment further explains step C based on specific embodiment 1. The data table set in step C is an accessible data table set provided by the data management module. The user-input question and the accessible data table set provided by the data management module are received by the intelligent semantic analysis module. The intelligent semantic analysis module is configured to extract key semantic features from the input data through semantic understanding technology. It includes a text vectorization unit based on a word embedding model, a semantic matching unit using an attention mechanism, and a feature enhancement unit containing a domain knowledge graph.

[0163] The data management module provides an accessible collection of data tables, which contains basic information about multiple tables, such as table names and descriptions. During implementation, the semantic analysis module can include a preprocessing layer (using an embedding model to generate question embedding vectors), a feature association layer (calculating the matching degree between field descriptions and questions using cosine similarity), a filtering layer (applying a knowledge graph-based rule engine to filter incompatible fields), and a decision layer (applying an attention mechanism model for semantic matching and returning the optimal decision solution), ultimately outputting the data table most relevant to the user's question. Specific Implementation Example 7:

[0165] This embodiment further explains step D based on specific embodiment 1. In step D, the data management module completes the process of obtaining the field set of the output data table, and the output field set of the data table contains basic information of several fields. Then, the intelligent semantic analysis module completes the process of obtaining the field set most relevant to the user's input question, and the obtained field set most relevant to the user's question is a subset of the field set of the data table. Specific Implementation Example 8:

[0167] This embodiment further explains step E based on specific embodiment 1. The internal process of the chart generation loop in step E is as follows: e1. Start traversing the template configuration and determine if there are any unprocessed chart objects. If so, read the current chart object; e2. Call the statistical method generation module, input the current target chart type and database type, and return a set of "dimension type - indicator type" objects compatible with the chart type and database type; call the chart generation module, receive the set of "dimension type - indicator type" objects and a subset of the data table field set, and output a set of "dimension type - dimension field - indicator type - indicator field" objects; e3. Call the intelligent semantic analysis module, receive the user question and the set of "dimension type - dimension field - indicator type - indicator field" objects, and output "dimension type - dimension field - indicator type - indicator field" objects closely related to the user's question. These objects serve as the query conditions for the target chart; e4. The memory module uses a Hash algorithm to process the generated "dimension type - dimension field - indicator type - indicator field" objects. e5. Store the "Indicator Field" object to remove duplicates and avoid chart repetition; e6. Pass the query conditions to the visualization dashboard renderer for chart rendering and display; e7. Repeat the above steps until all chart objects have been processed.

[0168] The data management module supports multiple data sources, including structured databases such as MySQL and H2, and unstructured databases such as Elasticsearch and Clickhouse. The differences between different database types are mainly handled by the statistical method generation module to ensure that the returned dimension and indicator types are supported by the database. Specific Implementation Example 9:

[0170] This embodiment further explains the fault tolerance mechanism based on specific embodiment 1. The specific application of the fault tolerance mechanism is as follows: During the chart generation process, if it is impossible to generate a set of "dimension type - dimension field - indicator type - indicator field" objects that are compatible with the target chart type, the fault tolerance mechanism is activated. First, it searches for a substitute of the same type. If the substitute fails, it is downgraded to displaying the basic chart. Specific Implementation Example 10:

[0172] A system for automatically generating data visualization dashboards according to any one of the above methods includes a template configuration module, a user input module, a user selection module, a data management module, an intelligent semantic analysis module, a processing module, a memory module, a fault-tolerant module, and a visualization dashboard renderer. The template configuration module is used to pre-build a set containing multiple visualization dashboard template configurations. The user input module receives user-input questions, which can be in natural language or structured input. The template selection module displays the visualization dashboard template configurations as images for user selection and receives the selected template configuration. The intelligent semantic analysis module is configured to extract key semantic features from the input using semantic understanding technology, including a preprocessing layer, a feature association layer, a filtering layer, and a decision layer. The data management module is used to... The system provides basic information and field sets from multiple data tables. The data management module supports various data sources, including structured and unstructured databases. The processing module performs the following processing based on the user-input question and the selected template configuration: a) Selecting the data table most relevant to the user's question from multiple data tables provided by the data management module through the intelligent semantic analysis module; b) Extracting the field set from the selected data table and selecting the subset of fields most relevant to the user's question from the field set through the intelligent semantic analysis module; c) Performing a breadth-first traversal of the target chart objects in the selected template configuration. For each target chart object, it calls the statistical method generation module and the chart generation module to generate query conditions and selects the most relevant object as the query condition through the intelligent semantic analysis module. The memory module stores and deduplicates the query conditions based on a hash algorithm. The fault tolerance module activates a fault tolerance mechanism if, during the chart generation process, a set of dimension and indicator types compatible with the target chart type cannot be generated. It first attempts to replace the current chart type with a substitute chart type of the same type. If the substitution fails, it is downgraded to the basic chart type for display. The visualization dashboard renderer renders the chart according to the query conditions and generates a visualization dashboard after all target chart objects have been processed.

[0173] Existing technology: Relying solely on keyword matching (e.g., binding a bar chart to a field name containing "sales") cannot understand the user's deeper analytical intent (e.g., time series charts should be prioritized for "trend analysis"). This application employs an embedded model + large language model + domain knowledge graph, which can identify analytical intents such as "comparison," "distribution," and "trend." Example: When a user enters "sales changes in various regions over the past six months," the system automatically selects: Time dimension → line chart (instead of the fixed bar chart of existing technology) and Region dimension → color classification. Improved performance: Chart type matching accuracy increased by 40%+.

[0174] Existing technology: The template is a fixed combination of charts and cannot be nested (e.g., it cannot package "sales indicators + regional distribution map" into an analysis unit). This application supports infinitely nested group types, which improves the layout efficiency of complex analysis scenarios by 3 times and reduces the number of manual adjustments.

[0175] Existing technology: When the field type is incompatible with the chart (e.g., a text field is used for a bar chart), the process is terminated directly with an error. This application adopts a three-level degradation strategy: ① Replace with the same type (bar chart → column chart), ② Simplify the chart type (3D pie chart → flat pie chart), ③ Finally downgrade to a data table. The success rate of large screen generation is increased from 68% in the existing technology to 99%.

[0176] Existing technology: Assuming all databases support the same SQL syntax (e.g., attempting to use window functions on the H2 database will result in an error), the statistical method generation module of this application dynamically adapts, thereby expanding the types of heterogeneous data sources supported by 5 times.

[0177] Existing technology: Each generation is an independent process, which may repeatedly generate the same analysis dimension (such as multiple charts using the "date" dimension). The memory module of this application records the used analysis dimension through hash fingerprint technology: hash("date-sales-sum") = "a1b2c3d4" # Stored in the memory pool, which increases the information density of the large screen by 30% and avoids redundant charts.

[0178] The statistical method generation module can use the following algorithm to take the current target chart type and database type as input and return a collection of "dimension type - indicator type" objects that are compatible with the chart type and database type;

[0179] 1) Dimension type and indicator type.

[0180] Dimension type refers to the way descriptive attributes (i.e., dimensions) are classified in data analysis. Common dimension types include grouping by value, grouping by time period, and grouping by numerical range. Let the complete set of dimension types be: .

[0181] Indicator types refer to the statistical methods used to calculate quantifiable and calculable numerical data in data analysis. Common indicator types include summation, averaging, maximization, minimization, and percentile calculation. Let the complete set of dimension types be: .

[0182] 2) Calculate the dimension types and indicator types supported by the input database type.

[0183] Assuming the target chart type to be processed is t, the dimensions and metrics of this chart type can be defined from the data dictionary based on t, including the number of dimensions: Number of indicators: and the types of metrics supported: ;

[0184] Assuming the current database type is "database", the definitions of dimensions and metrics for this data type can be obtained from the data dictionary based on the database type, including... and the types of metrics supported: .

[0185] Therefore, by performing a simple intersection operation, we can obtain the supported dimension types for the chart type t in the database:

[0186] .

[0187] Supported metric types:

[0188] .

[0189] 3) Returns a collection of "Dimension Type - Indicator Type" objects that are compatible with both chart and database types.

[0190] Permutation calculations can be used to obtain sets of all permutations of different elements from the supported dimension types and supported indicator types, respectively. Performing a Cartesian product on the two sets yields a set of "dimension type - indicator type" objects.

[0191] Dimensional arrangement of sets (ordered non-repeating sampling):

[0192] .

[0193] Index permutation set (ordered non-repeating sampling):

[0194] .

[0195] Complete set of statistical methods (Cartesian product):

[0196] .

[0197] in .

[0198] The chart generation module's algorithm can be implemented by receiving a collection of "dimension type - indicator type" objects and a subset of the data table's field set, and outputting a collection of "dimension type - dimension field - indicator type - indicator field" objects:

[0199] Suppose the subset of fields in the data table is: .

[0200] It is worth mentioning that each field It can be represented as a metadata object that includes field names, field aliases, and field types, for example: These metadata will play a role in subsequent semantic modules.

[0201] The set of fields supported by the calculation dimension type and indicator type.

[0202] Define a validity check function , indicating dimension type d and indicator type Compatibility:

[0203]

[0204] This function can generate a mapping relationship between any dimension type or indicator type and its legal field subset:

[0205] Mapping relationship between dimension type and its valid field subset: .

[0206] Mapping relationship between indicator types and their legal field subsets: .

[0207] Based on the above mapping relationship, given any dimension type d, Given a subset of its valid fields; given any index type m, It is a subset of its valid fields.

[0208] Calculate the collection of objects named "Dimension Type - Dimension Field - Metric Type - Metric Field".

[0209] For any "Dimension Type - Indicator Type" object For each dimension type and indicator type, select one field from the subset of valid fields, and ensure that the fields are not repeated. This will result in an object named "Dimension Type - Dimension Field - Indicator Type - Indicator Field".

[0210] .

[0211] Where i=1… j=1… .

[0212] For a collection of "Dimension Type - Indicator Type" objects, a corresponding collection of "Dimension Type - Dimension Field - Indicator Type - Indicator Field" objects can be generated.

[0213] The aforementioned technological advantages enable this application to significantly surpass existing solutions in terms of automation, analytical effectiveness, and practical applicability.

[0214] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatically generating large data visualization dashboards, characterized in that: The process includes the following steps: A. Preparing the visual dashboard template configuration; B. The user inputs a question and selects a preferred template from the set of visual dashboard template configurations, using the user-input question and the selected visual dashboard template configuration as input for subsequent processing. C. After understanding the user's input question through semantic understanding technology and analyzing it in conjunction with the data table set provided by the data management module, output the data table most relevant to the user's question. D. Based on the output data table, obtain the set of fields of the data table, and then combine it with the user's input question to obtain the set of fields most relevant to the user's input question; E. Combining the set of fields most relevant to the user's input question, perform a breadth-first traversal of the user's selected visualization dashboard template configuration, and process the target charts in turn. During the traversal of the user's selected visualization dashboard template configuration, for each target chart object, enter the chart generation loop process. F. Initiate a fault tolerance mechanism during the chart generation process; G. Finally, generate a visualization dashboard that matches the style and format of the selected template and whose data is relevant to the user's problem through the visualization dashboard renderer.

2. The method for automatically generating a data visualization dashboard according to claim 1, characterized in that: The preparation of the visualization large screen template configuration in step A specifically involves pre-constructing a set containing several visualization large screen template configurations. Each visualization large screen template configuration is a list consisting of several target chart objects. Each target chart object consists of an id and a type, where the id is associated with the rendering configuration of the visualization large screen renderer. `type` refers to the target chart type supported by the visualization large screen renderer.

3. The method for automatically generating a data visualization dashboard according to claim 2, characterized in that: The rendering configuration includes basic style parameters, layout constraints, and data rendering specifications. The type includes data tables, vertical bar charts, horizontal bar charts, indicators, pie charts, and also a special group type. The chart object of the group type contains a charts array to store the specific charts in the chart group, and there is no limit to the nesting depth of the group and the number of sub-charts.

4. The method for automatically generating a data visualization dashboard according to claim 1, characterized in that: In step B, the user inputs a question, which can be in the form of natural language input and / or structured input. The intelligent semantic analysis module can select an appropriate algorithm based on the input.

5. The method for automatically generating a data visualization dashboard according to claim 1, characterized in that: Before selecting a preferred template from the collection of visualization template configurations, the visualization template configuration is displayed as an image with actual effects.

6. The method for automatically generating a data visualization dashboard according to claim 1, characterized in that: The data table set in step C is an accessible data table set provided by the data management module. The user-input question and the accessible data table set provided by the data management module are received by the intelligent semantic analysis module. The intelligent semantic analysis module is configured to extract key semantic features from the input data through semantic understanding technology. It includes a text vectorization unit based on a word embedding model, a semantic matching unit using an attention mechanism, and a feature enhancement unit containing a domain knowledge graph.

7. The method for automatically generating a data visualization dashboard according to claim 1, characterized in that: In step D, obtaining the field set of the output data table is done by the data management module, and the output field set of the data table contains basic information of several fields; then, combining the user's input question, obtaining the field set most relevant to the user's input question is done by the intelligent semantic analysis module, and the obtained field set most relevant to the user's question is a subset of the field set of the data table.

8. The method for automatically generating a data visualization dashboard according to claim 1, characterized in that: The internal process of the chart generation loop in step E is as follows: e1. Start traversing the template configuration, determine if there are any unprocessed chart objects, and if so, read the current chart object; e2. Call the statistical method generation module, input the current target chart type and database type, and return a set of "dimension type - indicator type" objects compatible with the chart type and database type; call the chart generation module, receive the set of "dimension type - indicator type" objects and a subset of the data table field set, and output a set of "dimension type - dimension field - indicator type - indicator field" objects; e3. Call the intelligent semantic analysis module, receive the user question and the set of "dimension type - dimension field - indicator type - indicator field" objects, and output "dimension type - dimension field - indicator type - indicator field" objects closely related to the user's concerns. These objects serve as the query conditions for the target chart. e4. The memory module uses a hash algorithm to store and deduplicate the generated "dimension type - dimension field - indicator type - indicator field" objects to avoid chart duplication; e5. The query conditions are passed to the visualization dashboard renderer, which then renders and displays the charts; e6. Repeat the above steps until all chart objects have been processed.

9. The method for automatically generating a data visualization dashboard according to claim 1, characterized in that: The fault tolerance mechanism is applied as follows: During the chart generation process, if a set of "dimension type - dimension field - indicator type - indicator field" objects compatible with the target chart type cannot be generated, the fault tolerance mechanism is activated. First, a replacement of the same type is searched. If the replacement fails, the display is downgraded to the basic chart.

10. A system for generating an automatic data visualization dashboard, used to implement the automatic data visualization dashboard generation method according to any one of claims 1 to 9, characterized in that: The system includes a template configuration module, a user input module, a user selection module, a data management module, an intelligent semantic analysis module, a processing module, a memory module, a fault tolerance module, and a visualization large screen renderer. The template configuration module is used to pre-build a set containing multiple visualization large screen template configurations. The user input module is used to receive user-inputted questions, which can be in natural language or structured input. The template selection module is used to display the visualization large screen template configurations as images for the user to choose from and to receive the selected template configuration. The intelligent semantic analysis module is configured to extract key semantic features from the input using semantic understanding technology, including a preprocessing layer, a feature association layer, a filtering layer, and a decision layer. The data management module provides basic information and field sets for multiple data tables, including structured and unstructured databases; the processing module performs the following processing based on the user's input question and the selected template configuration: a) Selecting the data table most relevant to the user's question from the multiple data tables provided by the data management module through the intelligent semantic analysis module; b. Extract a set of fields from the selected data table, and use the intelligent semantic analysis module to select the subset of fields most relevant to the user's question from the set of fields; c. Perform a breadth-first traversal of the target chart objects in the selected template configuration. For each target chart object, call the statistical method generation module and the chart generation module to generate query conditions, and select the most relevant object as the query condition through the intelligent semantic analysis module. The memory module is used to store and deduplicate query conditions based on a hash algorithm; the fault tolerance module is used to activate the fault tolerance mechanism if a set of dimension and indicator types that are compatible with the target chart type cannot be generated during the chart generation process. It first attempts to replace the current chart type with a substitute chart type of the same type. If the substitution fails, it will be downgraded to the basic chart type for display; the visualization dashboard renderer is used to render the chart according to the query conditions and generate the visualization dashboard after all target chart objects have been processed.

Images