Commercial intelligent analysis method, system and device and storage medium

Abstract

The invention discloses a business intelligent analysis method, system and device and a storage medium, belongs to the technical field of computer science and information, and solves the technical problem that a traditional BI system usually needs professional data analysts and developers to perform complex programming and configuration. The method comprises a low-code platform, an AI Agent agent and a BI system. The method comprises the steps that a service model, a service interface, a service trend display mode and a dialog box for interaction between a user and the AI Agent agent are created through the low-code platform; acquiring data through the service interface and the dialog box; synchronizing the service model and the data through the BI system to form a data analysis data set; importing a business model and data through an AI Agent agent, and creating a large model; obtaining the content asked by the user for the AI Agent agent and a text input by the user; data analysis is carried out based on the large model, and a data analysis result is fed back to the BI system for further analysis; and the BI system returns analysis result data to the low-code platform for data display.

Description

Technical Field

[0001] The present invention relates to the field of computer science and information technology, and more specifically, to a method, system, device and storage medium for business intelligence analysis. Background Art

[0002] With the development of big data and artificial intelligence technology, enterprises have a growing demand for business intelligence (BI). BI systems can help enterprises extract valuable information from large amounts of data to support decision-making. However, traditional BI systems usually require professional data analysts and developers to perform complex programming and configuration, which is not only time-consuming and labor-intensive, but also costly. Summary of the invention

[0003] The technical problem to be solved by the present invention is to address the above-mentioned deficiencies in the prior art. One purpose of the present invention is to provide an efficient, low-cost, and easy-to-use method for business intelligence analysis.

[0004] A second objective of the present invention is to provide a system for business intelligence analysis.

[0005] A third object of the present invention is to provide a computer device.

[0006] A fourth objective of the present invention is to provide a computer storage medium.

[0007] In order to achieve the above-mentioned purpose one, the present invention provides a method for business intelligence analysis, including a low-code platform, an AI Agent, and a BI system, wherein the method is:

[0008] Create business models, business interfaces, business trend display methods, and dialog boxes for users to interact with the AIAgent agent through the low-code platform;

[0009] Acquire data through the business interface and dialog box;

[0010] Synchronizing the business model and data through the BI system to form a data analysis data set;

[0011] Import the business model and data through the AI ​​Agent to create a large model;

[0012] Obtain the content of the user's question to the AI ​​Agent and the text entered by the user;

[0013] Perform data analysis based on the big model, and feed the data analysis results back to the BI system for further analysis;

[0014] The BI system returns the analysis result data to the low-code platform for data display.

[0015] As a further improvement, the implementation process of the large model for data analysis includes:

[0016] Step 1. Pre-train the model, load domain knowledge, generate embeddings, and store them in the vector library;

[0017] Step 2. Preliminary screening: According to the question, a sub-table is retrieved from the vector library. The sub-table contains the minimum dimension and index set involved in the question. The sub-table is submitted to the big model as the context of the prompt to solve the problem of accuracy and too long prompt words.

[0018] Step 3: Fine sorting: matching the sub-table with the indicator library metadata to generate an accurate sub-table;

[0019] Step 4. Submit the precise sub-table and question sentence to the large model to obtain MQL;

[0020] Step 5. Call the BI data analysis interface of the BI system to execute the MQL, and the MQL executes the query process and implementation;

[0021] Step 6. Pass in the indicator model, business model, data analysis data set, and user ID, call the BI data analysis interface to implement data analysis, and return the data analysis results.

[0022] Furthermore, step 1 includes:

[0023] Step 11. Construction of domain knowledge graph: combining graph neural network and natural language processing to construct domain knowledge graph, integrating structured and unstructured data of the low-code platform to form a rich domain knowledge representation;

[0024] Step 12. Adaptive Embeddings Generation: By combining ANNS+embedding with an optimized adaptive learning algorithm, the Embeddings generation strategy is dynamically adjusted according to user behavior and feedback to improve the generalization ability and adaptability of the model.

[0025] Step 13. Multimodal learning integrates multiple data modalities such as text, images, and sounds, and improves the model's understanding and representation capabilities of domain knowledge through multimodal learning.

[0026] Furthermore, step 2 includes:

[0027] Step 21. Context-aware retrieval: Use the BERT pre-trained language model or a variant of BERT pre-trained language model to combine context information for semantic retrieval to improve retrieval accuracy.

[0028] Step 22. Dynamic sub-table generation: Based on the user's historical interaction data, dynamically adjust the sub-table generation strategy to reduce unnecessary information and improve the relevance of the context;

[0029] Step 23. Interactive prompt generation: Develop an interactive prompt generation algorithm to dynamically adjust the prompt words according to the user's input to improve the user interaction experience.

[0030] Furthermore, step 3 includes:

[0031] Step 31. Enhance the indicator library metadata by introducing external data sources and user feedback to continuously enrich and update the indicator library metadata to improve the accuracy of matching;

[0032] Step 32. Generate an intelligent matching algorithm, explore, optimize and encapsulate the intelligent matching algorithm of BERT deep learning, and improve the accuracy and efficiency of the matching algorithm by learning the complex relationship between user queries and indicator libraries;

[0033] Step 33. Multi-dimensional sorting and filtering, implement a multi-dimensional sorting and filtering mechanism, allowing users to filter and sort data based on multiple indicators and dimensions.

[0034] Furthermore, step 4 includes:

[0035] Step 41. Conversion from natural language to MQL: Explore and implement advanced natural language processing techniques to efficiently convert natural language queries into machine-parseable query languages, and improve the depth and accuracy of language understanding by optimizing BERT or GPT or Transformer-XL or BART algorithms;

[0036] Step 42: Context-aware query optimization, combining context information to optimize the generated MQL to improve query accuracy and execution efficiency;

[0037] Step 43. Establish an asynchronous execution and caching mechanism to improve the response speed of the BI data analysis interface and the overall performance of the system by implementing the asynchronous execution mechanism and query result caching;

[0038] Step 44. Establish intelligent error handling and retry logic, and develop intelligent error handling and retry logic to deal with various abnormal situations that may occur during the BI data analysis interface call process.

[0039] Furthermore, step 5 includes:

[0040] Step 51. Establish a data model. In MQL, a data model is an abstract representation of data in a database, defining the structure of the data. The data model includes tables, fields, and relationships.

[0041] Step 52: Establish an indicator model, which is a logical representation of data measurement and analysis. The indicator model is defined by the drag-and-drop function of the low-code platform to extract and calculate data from the data model to generate key performance indicators required for reporting and analysis;

[0042] Step 53. Establish a semantic layer. The semantic layer is an abstract layer above the data model and the indicator model. The semantic layer provides a more understandable query interface. Users can construct queries through the entities and attributes defined in the semantic layer instead of directly operating the physical tables and fields of the database.

[0043] Step 54. Generate SQL. When a user defines a query through MQL, the system will convert the query into a corresponding SQL statement, which will be automatically completed by the MQL parser. The MQL parser will generate an SQL statement that can retrieve the data required by the user's permissions from the database based on the user-defined model and indicators, as well as the mapping relationship of the semantic layer, and the user's role permission relationship, so as to solve the problem of data query permissions;

[0044] Step 55. Execute the query, the generated SQL statement will be sent to the database for execution, and the database will return the query results, which can be further processed and displayed.

[0045] In order to achieve the above-mentioned second objective, the present invention provides a system for business intelligence analysis, including a low-code platform, an AI Agent, a BI system, and also including:

[0046] A creation module, used to create a business model, a business interface, a business trend display method, and a dialog box for users to interact with the AI ​​Agent through the low-code platform;

[0047] A data acquisition module, used to acquire data through the business interface and dialog box;

[0048] A synchronization module, used to synchronize the business model and data through the BI system to form a data analysis data set;

[0049] A large model creation module is used to import the business model and data through the AI ​​Agent to create a large model;

[0050] The question acquisition module is used to obtain the content of the user's question to the AI ​​Agent and the text input by the user;

[0051] The analysis module is used to perform data analysis based on the big model and feed back the data analysis results to the BI system for further analysis; the BI system returns the analysis result data to the low-code platform for data display.

[0052] In order to achieve the third objective above, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method of business intelligence analysis when executing the computer program.

[0053] In order to achieve the fourth objective above, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for business intelligence analysis described above is implemented.

[0054] Beneficial Effects

[0055] Compared with the prior art, the present invention has the following advantages:

[0056] 1. The present invention can improve data processing efficiency and improve analysis efficiency through automatic data collection and analysis by AI Agent.

[0057] 2. The present invention can provide intelligent decision support and provide business-driven intelligent decision suggestions through AI Agent.

[0058] 3. The present invention can reduce data development costs. Low-code technology simplifies the development process of BI applications and reduces costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a flow chart of the present invention;

[0060] Figure 2 This is the workflow diagram of the AI ​​Agent data agent in the present invention. DETAILED DESCRIPTION

[0061] The present invention will be further described below with reference to the specific embodiments in the accompanying drawings.

[0062] See also Figure 1-2 , a method for business intelligence analysis, including a low-code platform, an AI Agent, and a BI system. The method is as follows:

[0063] Through the drag-and-drop method of the low-code platform interface, business models, business interfaces, business trend display methods, and dialog boxes for users to interact with AI Agents can be quickly created. The business trend display method can provide rich data display forms to achieve insights into business trends.

[0064] Get data through business interfaces and dialog boxes;

[0065] The business model and data are synchronized through the BI system to form a data analysis data set; the BI system also provides a more understandable data analysis query interface, namely the BI data analysis interface;

[0066] Import business models and data through AI Agent to create large models;

[0067] Obtain the content of the user's questions to the AI ​​Agent and the text entered by the user; that is, the AI ​​Agent provides an interface for interacting with the low-code platform;

[0068] Rely on big models to perform data analysis and feed the data analysis results back to the BI system for further analysis;

[0069] The BI system returns the analysis result data to the low-code platform for data display. The low-code platform supports more than 30 data display methods to meet diverse data presentation needs. Users can flexibly choose the most appropriate data display form based on business needs and personal preferences.

[0070] The implementation process of data analysis using a large model includes steps 1 to 6:

[0071] Step 1. Pre-train the model, load domain knowledge, generate embeddings (word embeddings), and store them in the vector library.

[0072] Specifically including steps 11 to 13:

[0073] Step 11. Build a domain knowledge graph by combining graph neural network GNN and natural language processing NLP to integrate the structured and unstructured data of the low-code platform to form a rich domain knowledge representation.

[0074] Step 12. Adaptive Embeddings Generation: By combining ANNS+embedding with an optimized adaptive learning algorithm, the Embeddings generation strategy is dynamically adjusted according to user behavior and feedback to improve the generalization ability and adaptability of the model.

[0075] Step 13. Multimodal learning integrates multiple data modalities such as text, images, and sounds, and improves the model's understanding and representation capabilities of domain knowledge through multimodal learning.

[0076] Step 2. Initial screening: According to the question, a sub-table is retrieved from the vector library. The sub-table contains the minimum dimensions and indicator sets involved in the question. The sub-table is submitted to the big model as the context of the prompt to solve the problem of accuracy and too long prompt words, that is, high accuracy and shorter prompt words can be achieved.

[0077] Specifically including steps 21 to 23:

[0078] Step 21. Context-aware retrieval: Use the BERT pre-trained language model or a BERT variant (such as ERNIE) pre-trained language model to perform semantic retrieval in combination with contextual information to improve retrieval accuracy.

[0079] Step 22: Dynamic sub-table generation: Based on the user's historical interaction data, the sub-table generation strategy is dynamically adjusted to reduce unnecessary information and improve the relevance of the context.

[0080] Step 23. Interactive prompt generation: Develop an interactive prompt generation algorithm to dynamically adjust the prompt words according to the user's input to improve the user interaction experience.

[0081] Step 3: Fine sorting: match the sub-table with the indicator library metadata to generate an accurate sub-table.

[0082] Specifically including steps 31 to 33:

[0083] Step 31. Enhance the indicator library metadata by introducing external data sources and user feedback to continuously enrich and update the indicator library metadata to improve the accuracy of matching.

[0084] Step 32. Generate an intelligent matching algorithm, explore, optimize and encapsulate the intelligent matching algorithm of BERT deep learning, and improve the accuracy and efficiency of the matching algorithm by learning the complex relationship between user queries and indicator libraries.

[0085] Step 33. Multi-dimensional sorting and filtering, implement a multi-dimensional sorting and filtering mechanism, allowing users to filter and sort data based on multiple indicators and dimensions.

[0086] Step 4. Submit the precise sub-table and question to the big model to obtain MQL.

[0087] Specifically including steps 41 to 44:

[0088] Step 41. Conversion from natural language to MQL, explore and implement advanced natural language processing techniques to efficiently convert natural language queries into machine-parseable query language MQL, by optimizing algorithms such as BERT or GPT or Transformer-XL or BART to improve the depth and accuracy of language understanding.

[0089] Step 42: Context-aware query optimization, combining context information to optimize the generated MQL to improve query accuracy and execution efficiency.

[0090] Step 43. Establish an asynchronous execution and caching mechanism to improve the response speed of the BI data analysis interface and the overall performance of the system by implementing the asynchronous execution mechanism and query result caching.

[0091] Step 44. Establish intelligent error handling and retry logic, and develop intelligent error handling and retry logic to deal with various abnormal situations that may occur during the BI data analysis interface call process.

[0092] Step 5. Call the BI data analysis interface of the BI system to execute MQL, and MQL executes the query process and implementation;

[0093] Specifically including steps 51 to 55:

[0094] Step 51. Create a data model. In MQL, a data model is an abstract representation of data in a database, defining the structure of the data. The data model includes tables, fields, and relationships. Data models are usually created by data analysts or database administrators to help users understand how data is organized. This solution uses low-code to create data models and synchronize them to BI, reducing the difficulty of model creation.

[0095] Step 52. Establish an indicator model. The indicator model is a logical representation of data measurement and analysis. By using the drag-and-drop function of the low-code platform, it defines how to extract and calculate data from the data model to generate the key performance indicators (KPIs) required for reporting and analysis.

[0096] Step 53. Establish a semantic layer. The semantic layer is an abstract layer above the data model and the indicator model. The semantic layer provides a more understandable query interface. Users can construct queries based on the entities and attributes defined in the semantic layer, rather than directly operating the physical tables and fields of the database.

[0097] Step 54. Generate SQL. When a user defines a query through MQL, the system will convert the query into a corresponding SQL statement. The generation of SQL is automatically completed by the MQL parser. The MQL parser will generate SQL statements that can retrieve the data required by the user's permissions from the database based on the user-defined model and indicators, as well as the mapping relationship of the semantic layer, and the user's role and permission relationship, to solve the problem of data query permissions.

[0098] Step 55. Execute the query, the generated SQL statement will be sent to the database for execution, and the database will return the query results, which can be further processed and displayed.

[0099] Step 6. Pass in the indicator model, business model, data analysis data set, and user ID, call the BI data analysis interface to implement data analysis, and return the data analysis results.

[0100] A business intelligence analysis system, including a low-code platform, an AI Agent, a BI system, and also includes:

[0101] Create a module for creating business models, business interfaces, business trend displays, and dialog boxes for users to interact with AI Agents through a low-code platform;

[0102] The data acquisition module is used to obtain data through the business interface and dialog box;

[0103] The synchronization module is used to synchronize business models and data through the BI system to form a data analysis data set;

[0104] Create a large model module, which is used to import business models and data through AI Agents to create large models;

[0105] The question acquisition module is used to obtain the content of the user's questions to the AI ​​Agent and the text entered by the user;

[0106] The analysis module is used to perform data analysis based on the big model and feed back the data analysis results to the BI system for further analysis; the BI system returns the analysis result data to the low-code platform for data display.

[0107] A computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the above-mentioned method of business intelligence analysis is implemented.

[0108] A computer-readable storage medium stores a computer program, which implements the above-mentioned business intelligence analysis method when executed by a processor.

[0109] The above are only preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the structure of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A method for business intelligence analysis, including a low-code platform, an AI Agent, and a BI system, characterized in that: The method is: Create business models, business interfaces, business trend display methods, and dialog boxes for users to interact with the AIAgent agent through the low-code platform; Acquire data through the business interface and dialog box; Synchronizing the business model and data through the BI system to form a data analysis data set; Import the business model and data through the AI ​​Agent to create a large model; Obtain the content of the user's question to the AI ​​Agent and the text entered by the user; Perform data analysis based on the big model, and feed the data analysis results back to the BI system for further analysis; The BI system returns the analysis result data to the low-code platform for data display.

2. A method for business intelligence analysis according to claim 1, characterized in that: The implementation process of the large model for data analysis includes: Step 1. Pre-train the model, load domain knowledge, generate embeddings, and store them in the vector library; Step 2. Preliminary screening: According to the question, a sub-table is retrieved from the vector library. The sub-table contains the minimum dimension and index set involved in the question. The sub-table is submitted to the big model as the context of the prompt to solve the problem of accuracy and too long prompt words. Step 3: Fine sorting: matching the sub-table with the indicator library metadata to generate an accurate sub-table; Step 4. Submit the precise sub-table and question sentence to the large model to obtain MQL; Step 5. Call the BI data analysis interface of the BI system to execute the MQL, and the MQL executes the query process and implementation; Step 6. Pass in the indicator model, business model, data analysis data set, and user ID, call the BI data analysis interface to implement data analysis, and return the data analysis results.

3. A method for business intelligence analysis according to claim 2, characterized in that: In step 1 include: Step 11. Construction of domain knowledge graph: combining graph neural network (GNN) and natural language processing (NLP) to build a domain knowledge graph, integrating the structured and unstructured data of the low-code platform to form a rich domain knowledge representation; Step 12. Adaptive Embeddings Generation: By combining ANNS+embedding with an optimized adaptive learning algorithm, the Embeddings generation strategy is dynamically adjusted according to user behavior and feedback to improve the generalization ability and adaptability of the model. Step 13. Multimodal learning integrates multiple data modalities such as text, images, and sounds, and improves the model's understanding and representation capabilities of domain knowledge through multimodal learning.

4. A method for business intelligence analysis according to claim 2, characterized in that: In step 2 include: Step 21. Context-aware retrieval: Use the BERT pre-trained language model or a variant of BERT pre-trained language model to combine context information for semantic retrieval to improve retrieval accuracy. Step 22. Dynamic sub-table generation: Based on the user's historical interaction data, the sub-table generation strategy is dynamically adjusted to reduce unnecessary information and improve the relevance of the context; Step 23. Interactive prompt generation: Develop an interactive prompt generation algorithm to dynamically adjust the prompt words according to the user's input to improve the user interaction experience.

5. A method for business intelligence analysis according to claim 2, characterized in that: In step 3 include: Step 31. Enhance the indicator library metadata by introducing external data sources and user feedback to continuously enrich and update the indicator library metadata to improve the accuracy of matching; Step 32. Generate an intelligent matching algorithm, explore, optimize and encapsulate the intelligent matching algorithm of BERT deep learning, and improve the accuracy and efficiency of the matching algorithm by learning the complex relationship between user queries and indicator libraries; Step 33. Multi-dimensional sorting and filtering, implement a multi-dimensional sorting and filtering mechanism, allowing users to filter and sort data based on multiple indicators and dimensions.

6. A method for business intelligence analysis according to claim 2, characterized in that: In step 4 include: Step 41. Conversion from natural language to MQL: Explore and implement advanced natural language processing techniques to efficiently convert natural language queries into machine-parseable query language (MQL), and improve the depth and accuracy of language understanding by optimizing BERT or GPT or Transformer-XL or BART algorithms; Step 42: Context-aware query optimization, combining context information to optimize the generated MQL to improve query accuracy and execution efficiency; Step 43. Establish an asynchronous execution and caching mechanism to improve the response speed of the BI data analysis interface and the overall performance of the system by implementing the asynchronous execution mechanism and query result caching; Step 44. Establish intelligent error handling and retry logic, and develop intelligent error handling and retry logic to deal with various abnormal situations that may occur during the BI data analysis interface call process.

7. A method for business intelligence analysis according to claim 2, characterized in that: In step 5 include: Step 51. Establish a data model. In MQL, a data model is an abstract representation of data in a database, defining the structure of the data. The data model includes tables, fields, and relationships. Step 52: Establish an indicator model, which is a logical representation of data measurement and analysis. The indicator model is defined by the drag-and-drop function of the low-code platform to extract and calculate data from the data model to generate key performance indicators (KPIs) required for reporting and analysis; Step 53. Establish a semantic layer. The semantic layer is an abstract layer above the data model and the indicator model. The semantic layer provides a more understandable query interface. Users can construct queries through the entities and attributes defined in the semantic layer instead of directly operating the physical tables and fields of the database. Step 54. Generate SQL. When a user defines a query through MQL, the system will convert the query into a corresponding SQL statement, which will be automatically completed by the MQL parser. The MQL parser will generate an SQL statement that can retrieve the data required by the user's permissions from the database based on the user-defined model and indicators, as well as the mapping relationship of the semantic layer, and the user's role permission relationship, so as to solve the problem of data query permissions; Step 55. Execute the query, the generated SQL statement will be sent to the database for execution, and the database will return the query results, which can be further processed and displayed.

8. A business intelligence analysis system, comprising a low-code platform, an AI Agent, and a BI system, characterized in that: Also includes: A creation module, used to create a business model, a business interface, a business trend display method, and a dialog box for users to interact with the AI ​​Agent through the low-code platform; A data acquisition module, used to acquire data through the business interface and dialog box; A synchronization module, used to synchronize the business model and data through the BI system to form a data analysis data set; A large model creation module is used to import the business model and data through the AIAgent agent to create a large model; The question acquisition module is used to obtain the content of the user's question to the AI ​​Agent and the text input by the user; The analysis module is used to perform data analysis based on the big model and feed back the data analysis results to the BI system for further analysis; the BI system returns the analysis result data to the low-code platform for data display.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method for business intelligence analysis described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for business intelligence analysis described in any one of claims 1 to 7 is implemented.

Images