Business intelligent report generation method and system based on large model and storage medium

Abstract

The invention provides a business intelligent report generation method and system based on a large model and a storage medium, and the method comprises the steps: obtaining a user report demand, inputting the user report demand into a pre-trained large model, and carrying out the intention recognition, and obtaining a user intention; determining an SQL query statement according to the user intention, and performing data acquisition according to the SQL query statement to obtain target data; and obtaining a report intention in the user intention, determining a target report according to the report intention, and performing data filling on the target report according to the target data to obtain a business intelligent report. According to the embodiment of the invention, the automatic SQL query and commercial intelligent report generation process effectively reduces the generation time of the commercial intelligent report and improves the generation efficiency of the commercial intelligent report.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, system and storage medium for generating business intelligence reports based on a large model. Background Art

[0002] In various business systems, a large amount of business data is stored in the database, and data search is cumbersome. In order to facilitate viewing of business data and facilitate enterprise decision-making, business data needs to be visualized and business intelligence reports (BI) need to be generated.

[0003] In the existing business intelligence report generation process, data analysis and data conversion are generally performed manually, resulting in low efficiency in business intelligence report generation. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a method, system and storage medium for generating business intelligence reports based on a large model, so as to solve the problem of low efficiency in generating business intelligence reports in the prior art.

[0005] The embodiment of the present invention is implemented as follows: a method for generating a business intelligence report based on a large model, the method comprising:

[0006] Obtain user report requirements, and input the user report requirements into the pre-trained large model for intent recognition to obtain user intent;

[0007] Determine an SQL query statement according to the user's intention, and acquire data according to the SQL query statement to obtain target data;

[0008] The report intention in the user intention is obtained, a target report is determined according to the report intention, and data is filled in the target report according to the target data to obtain a business intelligence report.

[0009] Preferably, the method further includes: inputting the user report requirement into the pre-trained large model for intent recognition, and obtaining the user intent and the report intent before:

[0010] Obtaining a model training sample, and inputting the model training sample into the large model for feature extraction to obtain sample features;

[0011] Performing vector conversion on the sample feature to obtain a sample vector, and calculating the vector similarity between the sample vector and a preset vector;

[0012] Determine the preset vector corresponding to the maximum vector similarity as a target vector, and output the intent information corresponding to the target vector as a sample intent;

[0013] The total model loss is determined according to the sample intent, and the parameters of the large model are updated according to the total model loss until the large model converges to obtain the pre-trained large model.

[0014] Preferably, determining the total model loss according to the sample intent comprises:

[0015] Performing feature mapping on the sample vector to obtain a mapping vector, and determining an SQL prediction statement according to the mapping vector;

[0016] Calculating a first model loss according to the sample intent and the sample label of the model training sample, and calculating a second model loss according to the SQL prediction statement and the SQL label of the model training sample;

[0017] The total model loss is determined according to the first model loss and the second model loss.

[0018] Preferably, filling in the target report with data according to the target data to obtain a business intelligence report includes:

[0019] Respectively obtain the subject and frame identifier of the report frame in the target report, and determine the type to be filled in according to the subject and the frame identifier;

[0020] The type to be filled in is matched with the target data to obtain matching data, and the matching data is filled in the corresponding report frame to obtain the business intelligence report.

[0021] Preferably, after filling in the target report with data according to the target data to obtain a business intelligence report, the method further includes:

[0022] Acquire user modification information, and modify the business intelligence report according to the user modification information;

[0023] The modified business intelligence report and the corresponding user report requirement are set as a model fine-tuning sample;

[0024] When the sample quantity of the model fine-tuning samples is greater than the quantity threshold, the parameters of the pre-trained large model are fine-tuned according to the model fine-tuning samples.

[0025] Preferably, obtaining the report intention in the user intention includes:

[0026] Perform entity recognition on the intended words in the user's intention to obtain the intended entity, and determine the target demand word according to the entity type of the intended entity;

[0027] The target requirement words are matched with the requirement list to obtain the report intent.

[0028] Preferably, acquiring data according to the SQL query statement to obtain target data includes:

[0029] Obtaining a query number of the SQL query statement, and determining a target plug-in according to the query number;

[0030] The target plug-in is called to execute the SQL query statement, and the execution result of the SQL query statement is obtained to obtain the target data.

[0031] Another object of an embodiment of the present invention is to provide a business intelligence report generation system based on a large model, the system comprising:

[0032] An intention recognition module is used to obtain user report requirements and input the user report requirements into a pre-trained large model for intention recognition to obtain user intentions;

[0033] A data acquisition module, used to determine an SQL query statement according to the user's intention, and acquire data according to the SQL query statement to obtain target data;

[0034] The report generation module is used to obtain the report intent in the user intention, determine the target report according to the report intent, and fill in the target report with data according to the target data to obtain a business intelligence report.

[0035] Preferably, the intention recognition module is also used for:

[0036] Obtaining a model training sample, and inputting the model training sample into the large model for feature extraction to obtain sample features;

[0037] Performing vector conversion on the sample feature to obtain a sample vector, and calculating the vector similarity between the sample vector and a preset vector;

[0038] Determine the preset vector corresponding to the maximum vector similarity as a target vector, and output the intent information corresponding to the target vector as a sample intent;

[0039] The total model loss is determined according to the sample intent, and the parameters of the large model are updated according to the total model loss until the large model converges to obtain the pre-trained large model.

[0040] The embodiments of the present invention can effectively obtain user intent by inputting user report requirements into a pre-trained large model for intent recognition, automatically determine SQL query statements based on user intent, automatically obtain target data based on the SQL query statements, automatically determine target reports based on report intent, and effectively obtain business intelligence reports by filling in target reports with target data. In this embodiment, the automated SQL query and business intelligence report generation process effectively reduces the generation time of business intelligence reports and improves the generation efficiency of business intelligence reports. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of a method for generating a business intelligence report based on a large model provided by a first embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of the structure of a business intelligence report generation system based on a large model provided by a second embodiment of the present invention;

[0043] Figure 3 It is a schematic diagram of specific implementation steps of a business intelligence report generation system based on a large model provided by a second embodiment of the present invention;

[0044] Figure 4 It is a schematic diagram of the structure of a terminal device provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0046] In order to illustrate the technical solution of the present invention, a specific embodiment is provided below for illustration.

[0047] Embodiment 1

[0048] See also Figure 1 , is a flow chart of a method for generating a business intelligence report based on a big model provided by a first embodiment of the present invention. The method for generating a business intelligence report based on a big model can be applied to any device or system. The method for generating a business intelligence report based on a big model comprises the following steps:

[0049] Step S10, obtaining user report requirements, and inputting the user report requirements into the pre-trained large model for intent recognition to obtain user intent;

[0050] Among them, user report requirements can be input into the pre-trained big model in the form of natural language or voice, and the pre-trained big model can be used to perform intent recognition on user report requirements, which can effectively identify the user intention corresponding to the user report requirements. The user intention includes the user's intention for the content and table format of the generated business intelligence report.

[0051] Optionally, before inputting the user report requirement into the pre-trained large model for intent recognition and obtaining the user intent and the report intent, the method further includes:

[0052] Obtain a model training sample, and input the model training sample into the large model for feature extraction to obtain sample features; wherein the model training sample can be set according to demand, and the sample features include content features, structural features, semantic features, etc.;

[0053] Performing vector conversion on the sample features to obtain a sample vector, and calculating the vector similarity between the sample vector and a preset vector; wherein the preset vector can be set according to requirements, and in this step, the Euclidean distance formula can be used to calculate the vector similarity between the sample vector and the preset vector;

[0054] The preset vector corresponding to the maximum vector similarity is determined as the target vector, and the intent information corresponding to the target vector is output as the sample intent; wherein the vector identifier of the target vector is obtained, and the vector identifier is matched with the intent query table to obtain the intent information corresponding to the target vector, and the intent query table stores the correspondence between different vector identifiers and corresponding intent information;

[0055] The total model loss is determined according to the sample intent, and the parameters of the large model are updated according to the total model loss until the large model converges to obtain the pre-trained large model.

[0056] Further, determining the total model loss according to the sample intent includes:

[0057] Performing feature mapping on the sample vector to obtain a mapping vector, and determining an SQL prediction statement according to the mapping vector; wherein, performing feature mapping on the sample vector according to a preset mapping relationship to obtain a mapping vector, and matching the mapping vector with a statement query table to obtain an SQL prediction statement, wherein the statement query table stores a correspondence between different mapping vectors and corresponding SQL prediction statements;

[0058] Calculating a first model loss according to the sample intent and the sample label of the model training sample, and calculating a second model loss according to the SQL prediction statement and the SQL label of the model training sample;

[0059] The total model loss is determined according to the first model loss and the second model loss; wherein, the first model loss and the second model loss are weighted to obtain the total model loss, and the weighted values ​​corresponding to the first model loss and the second model loss can be set according to requirements.

[0060] Step S20, determining an SQL query statement according to the user's intention, and acquiring data according to the SQL query statement to obtain target data;

[0061] Among them, the SQL query statement corresponding to the user's intention is automatically determined according to the pre-trained large model, and the SQL query statement is executed to obtain data to obtain the target data.

[0062] Optionally, acquiring data according to the SQL query statement to obtain target data includes:

[0063] Obtaining a query number of the SQL query statement, and determining a target plug-in according to the query number; wherein the query number is matched with a plug-in query table to obtain a target plug-in, and the plug-in query table stores a correspondence between different query numbers and corresponding target plug-ins;

[0064] The target plug-in is called to execute the SQL query statement, and the execution result of the SQL query statement is obtained to obtain the target data; wherein, the target plug-in is called to locate the database to obtain the target database, and the SQL query statement is executed in the target database to obtain data to obtain the target data.

[0065] Step S30, obtaining the report intention in the user intention, determining the target report according to the report intention, and filling the target report with data according to the target data to obtain a business intelligence report;

[0066] The report intent is used to indicate the user's format requirements for the business intelligence report to be generated. In this step, the target report is obtained by matching the report intent with the report format table, and the report format table stores the correspondence between different report formats and corresponding target reports.

[0067] Optionally, filling in the target report with data according to the target data to obtain a business intelligence report includes:

[0068] Respectively obtain the subject and frame identifier of the report frame in the target report, and determine the type to be filled in according to the subject and the frame identifier; wherein the subject is used to indicate the subject content in the report frame, and the frame identifier is used to indicate the frame number of the corresponding report frame, and the identifier combination is obtained by combining the subject and the frame identifier, and the identifier combination is matched with the type query table to obtain the type to be filled in, and the type query table stores the corresponding relationship between different identifier combinations and corresponding types to be filled in;

[0069] The type to be filled in is matched with the target data to obtain matching data, and the matching data is filled in the corresponding report frame to obtain the business intelligence report.

[0070] Furthermore, after filling in the target report with data according to the target data to obtain a business intelligence report, the method further includes:

[0071] Acquire user modification information, and modify the business intelligence report according to the user modification information; set the modified business intelligence report and the corresponding user report requirement as a model fine-tuning sample;

[0072] When the number of the model fine-tuning samples is greater than a quantity threshold, fine-tuning parameters of the pre-trained large model according to the model fine-tuning samples;

[0073] Among them, the business intelligence report is modified by the user modifying the information to modify the errors in the business intelligence report, thereby improving the accuracy of the business intelligence report. By setting the modified business intelligence report and the corresponding user report requirements as model fine-tuning samples, the parameter fine-tuning of the pre-trained large model is effectively guaranteed, further improving the accuracy of the pre-trained large model.

[0074] Furthermore, obtaining the report intent in the user intent includes:

[0075] Perform entity recognition on the intent words in the user intent to obtain the intent entity, and determine the target demand word according to the entity type of the intent entity; wherein the entity type of the intent entity is matched with a preset type, and the intent word corresponding to the successfully matched entity type is determined as the target demand word;

[0076] The target requirement words are matched with the requirement list to obtain the report intent; wherein the requirement list stores the correspondence between different target requirement words and corresponding report intents.

[0077] In this embodiment, by inputting user report requirements into the pre-trained large model for intent recognition, the user intent can be effectively obtained, the SQL query statement can be automatically determined based on the user intent, the target data can be automatically obtained based on the SQL query statement, the target report can be automatically determined based on the report intent, and the target report can be filled with data using the target data to effectively obtain a business intelligence report. In this embodiment, the automated SQL query and business intelligence report generation process effectively reduces the generation time of business intelligence reports and improves the generation efficiency of business intelligence reports.

[0078] Embodiment 2

[0079] See also Figure 2 , is a schematic diagram of the structure of a business intelligence report generation system 100 based on a large model provided by a second embodiment of the present invention, including:

[0080] The intention recognition module 10 is used to obtain user report requirements and input the user report requirements into the pre-trained big model for intention recognition to obtain user intention. The user report requirements can be input into the pre-trained big model in natural language or voice, and the pre-trained big model is used to perform intention recognition on the user report requirements, which can effectively identify the user intention corresponding to the user report requirements, including the user's intention for the content and table format of the business intelligence report to be generated.

[0081] Optionally, the intention recognition module 10 is further used to: obtain a model training sample, and input the model training sample into the large model for feature extraction to obtain sample features; wherein the model training sample can be set according to requirements, and the sample features include content features, structural features, semantic features, etc.;

[0082] Perform vector conversion on the sample features to obtain a sample vector, and calculate the vector similarity between the sample vector and a preset vector; wherein the preset vector can be set according to requirements, and in this module, the Euclidean distance formula can be used to calculate the vector similarity between the sample vector and the preset vector;

[0083] The preset vector corresponding to the maximum vector similarity is determined as the target vector, and the intent information corresponding to the target vector is output as the sample intent; wherein the vector identifier of the target vector is obtained, and the vector identifier is matched with the intent query table to obtain the intent information corresponding to the target vector, and the intent query table stores the correspondence between different vector identifiers and corresponding intent information;

[0084] The total model loss is determined according to the sample intent, and the parameters of the large model are updated according to the total model loss until the large model converges to obtain the pre-trained large model.

[0085] Furthermore, the intention recognition module 10 is also used to: perform feature mapping on the sample vector to obtain a mapping vector, and determine a SQL prediction statement according to the mapping vector;

[0086] Calculating a first model loss according to the sample intent and the sample label of the model training sample, and calculating a second model loss according to the SQL prediction statement and the SQL label of the model training sample;

[0087] The total model loss is determined according to the first model loss and the second model loss.

[0088] The data acquisition module 11 is used to determine the SQL query statement according to the user's intention, and to acquire data according to the SQL query statement to obtain the target data. The SQL query statement corresponding to the user's intention is automatically determined according to the pre-trained large model, and the SQL query statement is executed to acquire data to obtain the target data.

[0089] Optionally, the data acquisition module 11 is further used to: acquire a query number of the SQL query statement, and determine a target plug-in according to the query number;

[0090] The target plug-in is called to execute the SQL query statement, and the execution result of the SQL query statement is obtained to obtain the target data.

[0091] The report generation module 12 is used to obtain the report intent in the user's intention, determine the target report according to the report intent, and fill the target report with data according to the target data to obtain a business intelligence report. The report intent is used to indicate the user's format requirements for the business intelligence report to be generated. In this step, the target report is obtained by matching the report intent with the report format table, and the report format table stores the correspondence between different report formats and corresponding target reports.

[0092] Optionally, the report generation module 12 is further used to: respectively obtain the subject and frame identifier of the report frame in the target report, and determine the type to be filled in according to the subject and the frame identifier;

[0093] The type to be filled in is matched with the target data to obtain matching data, and the matching data is filled in the corresponding report frame to obtain the business intelligence report.

[0094] Furthermore, the report generation module 12 is also used to: obtain user modification information, and modify the business intelligence report according to the user modification information;

[0095] The modified business intelligence report and the corresponding user report requirement are set as a model fine-tuning sample;

[0096] When the sample quantity of the model fine-tuning samples is greater than the quantity threshold, the parameters of the pre-trained large model are fine-tuned according to the model fine-tuning samples.

[0097] Furthermore, the report generation module 12 is also used to: perform entity recognition on the intended words in the user's intention to obtain the intended entity, and determine the target demand word according to the entity type of the intended entity;

[0098] The target requirement words are matched with the requirement list to obtain the report intent.

[0099] Please refer to Figure 3 The specific implementation steps of the business intelligence report generation system based on the big model include:

[0100] Perform natural language annotation on the data tables and fields in the model training samples;

[0101] Use the annotated corpus (model training samples) to fine-tune the large model to make it powerful in business data understanding and SQL output;

[0102] Users describe their data analysis needs in natural language;

[0103] The large model receives natural language descriptions (user report requirements) and performs intent recognition and semantic understanding;

[0104] The big model converts the understood intent into SQL query statements;

[0105] The large model calls the plug-in, and the system executes SQL queries to obtain data and obtain the target data;

[0106] The system automatically designs report formats and generates BI reports based on data and user needs;

[0107] Users determine whether the business requirements understood by the big model and the output BI reports are accurate;

[0108] Collect bad cases of model fine-tuning samples in use, regularly fine-tune the model, and optimize the large model's ability to identify user business needs and generate SQL and BI reports.

[0109] In this embodiment, the technical threshold is lowered, and users can describe their data analysis needs in natural language without having to master programming skills. The big model can understand variable natural language descriptions and adapt to different query requirements and report formats. The entire process from natural language description to BI report generation does not require human intervention, and is fully automated. By inputting user report requirements into the pre-trained big model for intent recognition, the user's intent can be effectively acquired, and the SQL query statement can be automatically determined based on the user's intent. The target data can be automatically acquired based on the SQL query statement, and the target report can be automatically determined based on the report intent. By filling in the target report with the target data, a business intelligence report can be effectively obtained. In this embodiment, the automated SQL query and business intelligence report generation process effectively reduces the generation time of business intelligence reports and improves the efficiency of business intelligence report generation.

[0110] Embodiment 3

[0111] Figure 4 2 is a block diagram of a terminal device 2 provided in the third embodiment of the present application. Figure 4 As shown, the terminal device 2 of this embodiment includes: a processor 20, a memory 21, and a computer program 22 stored in the memory 21 and executable on the processor 20, such as a program of a method for generating a business intelligence report based on a large model. When the processor 20 executes the computer program 22, the steps in each embodiment of the method for generating a business intelligence report based on a large model are implemented.

[0112] Exemplarily, the computer program 22 may be divided into one or more modules, which are stored in the memory 21 and executed by the processor 20 to complete the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 22 in the terminal device 2. The terminal device may include, but is not limited to, a processor 20 and a memory 21.

[0113] The processor 20 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0114] The memory 21 may be an internal storage unit of the terminal device 2, such as a hard disk or memory of the terminal device 2. The memory 21 may also be an external storage device of the terminal device 2, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device 2. Further, the memory 21 may also include both an internal storage unit and an external storage device of the terminal device 2. The memory 21 is used to store the computer program and other programs and data required by the terminal device. The memory 21 may also be used to temporarily store data that has been output or is to be output.

[0115] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional unit.

[0116] If the integrated module is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium can be non-volatile or volatile. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable storage medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in computer-readable storage media can be appropriately increased or decreased according to the requirements of legislation and patent practices in the jurisdiction. For example, in some jurisdictions, based on legislation and patent practices, computer-readable storage media do not include electrical carrier signals and telecommunication signals.

[0117] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for generating business intelligence reports based on a large model, characterized in that: The method comprises: Obtain user report requirements, and input the user report requirements into the pre-trained large model for intent recognition to obtain user intent; Determine an SQL query statement according to the user's intention, and acquire data according to the SQL query statement to obtain target data; The report intention in the user intention is obtained, a target report is determined according to the report intention, and data is filled in the target report according to the target data to obtain a business intelligence report.

2. The method for generating a business intelligence report based on a large model as claimed in claim 1, characterized in that: The user report requirement is input into the pre-trained large model for intent recognition, and before obtaining the user intent and the report intent, the method further includes: Obtaining a model training sample, and inputting the model training sample into the large model for feature extraction to obtain sample features; Performing vector conversion on the sample feature to obtain a sample vector, and calculating the vector similarity between the sample vector and a preset vector; Determine the preset vector corresponding to the maximum vector similarity as a target vector, and output the intent information corresponding to the target vector as a sample intent; The total model loss is determined according to the sample intent, and the parameters of the large model are updated according to the total model loss until the large model converges to obtain the pre-trained large model.

3. The method for generating a business intelligence report based on a large model as claimed in claim 2, characterized in that: Determine the total model loss based on the sample intent, including: Performing feature mapping on the sample vector to obtain a mapping vector, and determining an SQL prediction statement according to the mapping vector; Calculating a first model loss according to the sample intent and the sample label of the model training sample, and calculating a second model loss according to the SQL prediction statement and the SQL label of the model training sample; The total model loss is determined according to the first model loss and the second model loss.

4. The method for generating a business intelligence report based on a large model as claimed in claim 1, characterized in that: Filling the target report with data according to the target data to obtain a business intelligence report, including: Respectively obtain the subject and frame identifier of the report frame in the target report, and determine the type to be filled in according to the subject and the frame identifier; The type to be filled in is matched with the target data to obtain matching data, and the matching data is filled in the corresponding report frame to obtain the business intelligence report.

5. The method for generating a business intelligence report based on a large model as claimed in claim 1, characterized in that: After filling in the target report with data according to the target data to obtain a business intelligence report, the method further includes: Acquire user modification information, and modify the business intelligence report according to the user modification information; The modified business intelligence report and the corresponding user report requirement are set as a model fine-tuning sample; When the sample quantity of the model fine-tuning samples is greater than the quantity threshold, the parameters of the pre-trained large model are fine-tuned according to the model fine-tuning samples.

6. The method for generating a business intelligence report based on a large model as claimed in claim 1, characterized in that: Obtain the report intent in the user intent, including: Perform entity recognition on the intended words in the user's intention to obtain the intended entity, and determine the target demand word according to the entity type of the intended entity; The target requirement words are matched with the requirement list to obtain the report intent.

7. The method for generating a business intelligence report based on a large model as claimed in claim 1, characterized in that: Acquire data according to the SQL query statement to obtain target data, including: Obtaining a query number of the SQL query statement, and determining a target plug-in according to the query number; The target plug-in is called to execute the SQL query statement, and the execution result of the SQL query statement is obtained to obtain the target data.

8. A business intelligence report generation system based on a large model, characterized in that: The system comprises: An intention recognition module is used to obtain user report requirements and input the user report requirements into a pre-trained large model for intention recognition to obtain user intentions; A data acquisition module, used to determine an SQL query statement according to the user's intention, and acquire data according to the SQL query statement to obtain target data; The report generation module is used to obtain the report intent in the user intention, determine the target report according to the report intent, and fill in the target report with data according to the target data to obtain a business intelligence report.

9. The business intelligence report generation system based on a large model as claimed in claim 8, characterized in that: The intention recognition module is also used for: Obtaining a model training sample, and inputting the model training sample into the large model for feature extraction to obtain sample features; Performing vector conversion on the sample feature to obtain a sample vector, and calculating the vector similarity between the sample vector and a preset vector; Determine the preset vector corresponding to the maximum vector similarity as a target vector, and output the intent information corresponding to the target vector as a sample intent; The total model loss is determined according to the sample intent, and the parameters of the large model are updated according to the total model loss until the large model converges to obtain the pre-trained large model.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images