An agent-based data analysis method, apparatus, device, and medium
By generating data manipulation code through intelligent agents, the problem of large language models being limited by the length of the context window in enterprise-level data analysis is solved, realizing an efficient and automated data analysis process and ensuring the accuracy and real-time nature of the results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INSPUR ZHUOSHU BIG DATA IND DEV CO LTD
- Filing Date
- 2026-01-07
- Publication Date
- 2026-06-30
AI Technical Summary
Large language models are limited by the length of the context window when processing massive enterprise-level data, which leads to distorted analysis results. Existing methods are difficult to achieve accurate analysis and lack automatic execution and verification mechanisms. Manual coding technology has a high threshold and cannot meet the real-time update and optimization requirements of complex data environments.
An intelligent agent is used to generate executable data operation code. By parsing user requests to obtain key entities, a pre-set intelligent agent is triggered to generate code, execute and verify the results, iterate and update the code until it passes verification, and then provide feedback to the user.
It enables users to perform efficient data analysis without needing to master complex data structures, automates code generation and verification, shortens the analysis cycle, reduces the risk of erroneous decisions, and improves the accuracy and real-time performance of analysis results.
Smart Images
Figure CN122309948A_ABST
Abstract
Description
Technical Field
[0001] This specification relates to the field of artificial intelligence technology, and in particular to a data analysis method, apparatus, device, and medium based on intelligent agents. Background Technology
[0002] With the rapid development of large language models in natural language understanding and generation, their applications in data analysis, business intelligence, and automated report generation are becoming increasingly widespread. However, when processing real-world enterprise-level data, large language models are limited by their fixed context window length, making it impossible to directly load data containing massive amounts of records for end-to-end data analysis.
[0003] Existing methods, such as data sampling, input only a portion of the data into the model, which is prone to distorting the analysis results due to sampling bias and fails to meet the needs of accurate analysis. Manual coding requires users or data analysts to manually write query scripts, which has a high technical threshold, slow response time, and cannot achieve direct conversion from natural language to analysis. Furthermore, while some systems can generate query code from large language models, the lack of automatic execution and verification mechanisms after generation makes it difficult to guarantee the accuracy of the output results. Moreover, with the increasing complexity of data environments, existing methods struggle to form closed-loop analysis and are difficult to update and optimize in a timely manner based on real-world scenarios. Summary of the Invention
[0004] To address the aforementioned issues, this specification provides one or more embodiments of a data analysis method, apparatus, device, and medium based on intelligent agents.
[0005] One or more embodiments of this specification employ the following technical solutions: This specification provides one or more embodiments of a data analysis method based on an intelligent agent, specifically including: The system receives data requests uploaded by users and parses the data requests to obtain key entities; wherein the request information includes: data query requests and data analysis requests. Obtain the metadata of the target data source based on the key entity to trigger the pre-built intelligent agent to generate executable data operation code; Execute the data manipulation code, obtain the execution result, and verify the execution result; If the verification fails, the verification failure information is fed back to the preset intelligent agent to drive the preset intelligent agent to update the data operation code and iterate the execution result; If the verification passes, the execution result will be sent back to the user.
[0006] Optionally, in one or more embodiments of this specification, receiving a data request uploaded by a user terminal to parse the data request and obtain key entities specifically includes: Receive data requests uploaded by the user and obtain the data input format corresponding to the data request; According to the data conversion strategy corresponding to the data input format, the data request is converted to obtain the data request text; The key entities of the data request text are extracted based on natural language processing; wherein, the key entities include: time, dimension, metric, and aggregation method.
[0007] Optionally, in one or more embodiments of this specification, the metadata of the target data source is obtained based on the key entity to trigger a pre-built intelligent agent to generate executable data operation code, specifically including: Based on the target data source identifier corresponding to the key entity, determine the corresponding target data source; Obtain the metadata of the target data source to compare the data size of the metadata with a preset threshold; If the data size exceeds the preset threshold, the preset intelligent agent is triggered to generate executable data operation code; If the data size does not exceed the preset threshold, the data content of the target data source is input into the preset large language model to obtain the execution result and feed it back to the user.
[0008] Optionally, in one or more embodiments of this specification, if the data size exceeds the preset threshold, the preset intelligent agent is triggered to generate executable data operation code, specifically including: If the data size exceeds the preset threshold, the agent mode of the preset intelligent agent is triggered and started. In the proxy mode, the data mode corresponding to the target data source is determined based on the structural information of the target data source; The user requirements defined by the key entities are integrated with the data pattern to construct prompt information, thereby driving the pre-set intelligent agent to generate an executable data operation script that matches the data pattern.
[0009] Optionally, in one or more embodiments of this specification, executing the data manipulation code, obtaining the execution result, and verifying the execution result specifically includes: The data manipulation code is run in a preset isolated execution environment to obtain the execution results; wherein, the execution results include: output data and execution status log; The data manipulation code is syntax-validated based on the execution status log. If the syntax validation passes, then the output data undergoes logical validation; wherein the logical validation includes at least: context constraint validation and rationality validation. If the logic verification passes, it is determined whether the output data matches the user requirements of the key entity, and the verification result of the execution result is obtained.
[0010] Optionally, in one or more embodiments of this specification, verification failure information is fed back to the preset intelligent agent to drive the preset intelligent agent to update the data operation code and iterate the execution result, specifically including: The verification failure information is fed back to the preset intelligent agent, so that the error factors corresponding to the verification failure information can be identified based on the preset intelligent agent; The preset intelligent agent is driven to adjust its prompt words according to the error factors, and the data operation code is regenerated according to the adjusted prompt words; The data manipulation code is updated and generated in a loop, the updated data manipulation code is executed, and the execution result of the updated data manipulation code is verified until the verification is successful and the execution result is obtained.
[0011] Optionally, in one or more embodiments of this specification, the execution result is fed back to the user terminal, specifically including: The verified execution results are format-converted to obtain a natural language summary; Based on the current visualization requirements, the natural language summary is processed to generate corresponding visualization content; The natural language summary is integrated with the visualization content to generate comprehensive feedback content, which is then sent to the user.
[0012] This specification provides one or more embodiments of an agent-based data analysis device, the device comprising: The parsing unit is used to receive data requests uploaded by the user terminal and to parse the data requests to obtain key entities; wherein, the request information includes: data query requests and data analysis requests; The generation unit is used to obtain the metadata of the target data source based on the key entity, so as to trigger the pre-set intelligent agent to generate executable data operation code; An execution unit is used to execute the data manipulation code, obtain the execution result, and verify the execution result; An iteration unit is used to feed back verification failure information to the preset intelligent agent if the verification fails, so as to drive the preset intelligent agent to update the data operation code and iterate the execution result; The feedback unit is used to send the execution result back to the user terminal if the verification is successful.
[0013] This specification provides one or more embodiments of an agent-based data analysis device, the device comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any of the methods described above.
[0014] This specification provides one or more embodiments of a non-volatile computer storage medium storing computer-executable instructions configured to perform any of the methods described above.
[0015] The above-described at least one technical solution adopted in the embodiments of this specification can achieve the following beneficial effects: Based on agent-generated data manipulation code, users do not need to understand complex underlying data table structures. They can efficiently perform complex data analysis simply by describing their requirements using natural language or a simple interface. Automating the process of manually writing data manipulation code greatly shortens the cycle from requirement submission to result acquisition. Through an iterative process, the agent can autonomously discover code errors or result deviations and attempt to fix them, reducing the time spent on manual debugging and intervention. Automatic verification of execution results also reduces the risk of incorrect decisions due to code errors or data problems. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 A flowchart illustrating an agent-based data analysis method provided in an embodiment of this specification; Figure 2 This is a schematic diagram of the overall logic of an agent-based data analysis method provided in the embodiments of this specification. Figure 3 A schematic diagram of the structure of a data analysis device based on an intelligent agent provided in the embodiments of this specification; Figure 4 A schematic diagram of the structure of a data analysis device based on an intelligent agent provided in the embodiments of this specification; Figure 5 This is a schematic diagram of the structure of a non-volatile storage medium provided in the embodiments of this specification. Detailed Implementation
[0017] This specification provides an embodiment of a data analysis method, apparatus, device, and medium based on intelligent agents.
[0018] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.
[0019] like Figure 1 As shown, this specification provides a flowchart illustrating a data analysis method based on intelligent agents. Figure 1 As can be seen, in one or more embodiments of this specification, a data analysis method based on an intelligent agent specifically includes the following steps: S101: Receive data requests uploaded by the user terminal, and parse the data requests to obtain key entities; wherein, the request information includes: data query requests and data analysis requests.
[0020] The request information mainly includes two categories: data query requests and data analysis requests. Data query requests typically refer to instructions from users who want to retrieve a specific set of data from the system, such as "query the sales figures for the first quarter of 2023". Data analysis requests refer to requests from users who want to process, calculate, and mine existing data to generate new conclusions, such as "analyze the year-on-year growth of monthly sales in different regions".
[0021] Specifically, in one or more embodiments of this specification, receiving a data request uploaded by a user terminal and parsing the data request to obtain a key entity includes the following steps: The system receives data requests uploaded by users and obtains the corresponding data input format. Based on the data conversion strategy corresponding to the data input format, it converts the data request to obtain the data request text. This method unifies non-standard or diverse input formats into an intermediate text format that the system can efficiently process. For example, for natural language queries, data conversion strategies might include word segmentation, part-of-speech tagging, and entity linking to generate a structured text containing core semantic information. For structured queries, data conversion strategies might include format validation and parameter extraction to ensure data integrity and correctness.
[0022] This involves extracting key entities from data request text using natural language processing. For example, by using pre-defined extraction rules, key entities can be automatically identified and extracted from the text. These entities are the core elements that constitute a complete analysis or query task, and mainly include: time, dimensions, metrics, aggregation methods, etc.
[0023] S102: Obtain the metadata of the target data source based on the key entity to trigger the pre-set intelligent agent to generate executable data operation code.
[0024] To associate the target data source with metadata and select the optimal processing path based on the actual situation of the data source, ultimately realizing the implementation of user data needs, this specification embodiment obtains the metadata of the target data source based on key entities to trigger a pre-built intelligent agent to generate executable data operation code. Specifically, in one or more embodiments of this specification, obtaining the metadata of the target data source based on key entities to trigger a pre-built intelligent agent to generate executable data operation code includes the following steps: Each key entity is associated with a corresponding target data source identifier, which is a unique identifier that distinguishes different data sources, such as a database name, data warehouse table identifier, or file storage path identifier. In the embodiments described in this specification, based on the target data source identifier corresponding to the key entity, the corresponding target data source identifier is extracted or matched, and then the system's pre-set data source list is searched using this identifier to obtain the target data source corresponding to the user's data request, providing a clear object for subsequent metadata acquisition and data processing.
[0025] After identifying the target data source, the system proactively acquires its metadata. Metadata is the core information describing the data source's attributes, including key metrics related to data scale such as the number of rows, storage usage, and number of fields. After acquiring the metadata, the actual data scale is precisely compared with pre-defined thresholds to determine the size of the data source, providing crucial decision-making support for selecting different processing strategies.
[0026] like Figure 2 As shown, if the data size exceeds a preset threshold, directly processing the raw data would lead to excessive system resource consumption, low processing efficiency, and even task timeouts. Therefore, a pre-built intelligent agent is triggered to generate executable data operation code, ensuring efficient execution of data operations in large-scale data scenarios. If the data size does not exceed the preset threshold, the data content from the target data source is input into a pre-built large language model to obtain the execution result, which is then fed back to the user. This pre-built large language model is an existing large language model, and this process will not be elaborated upon here. This process effectively overcomes the limitations of data size, avoids feeding massive amounts of raw data into a large model, and significantly reduces contextual pressure.
[0027] Furthermore, in one or more embodiments of this specification, if the data size exceeds a preset threshold, a pre-set intelligent agent is triggered to generate executable data operation code, specifically including: Once it is confirmed that the data size of the target data source exceeds a preset threshold, the agent's proxy mode will be triggered and activated. It should be noted that this proxy mode is a dedicated working mode for the agent to operate on large volumes of data. It possesses stronger data source adaptability, logical analysis capabilities, and code generation capabilities, meeting the need for efficient and accurate generation of executable data operation code in large-scale data scenarios, providing mode support for subsequent steps.
[0028] In proxy mode, based on the structural information of the target data source, the corresponding data pattern is determined. The data pattern is a standardized extraction of the target data source's data structure and storage logic, serving as a crucial basis for generating matching data operation scripts and ensuring that the generated code is fully compatible with the target data source's structure. Key entities obtained through parsing are semantically analyzed to clarify their defined user requirements. These user requirements are then integrated with the data pattern to construct prompts that drive a pre-built intelligent agent, generating an executable data operation script that matches the data pattern.
[0029] This process achieves precise matching between user needs and data source characteristics by activating a dedicated agent mode for intelligent agents and refining data patterns. This effectively avoids common problems in large-scale data scenarios, such as code adaptation anomalies, operational lag, and task interruptions, ensuring smooth execution of data operations. Through semantic analysis of key entities, scattered entity information is transformed into logically clear user requirements. Then, through the fusion and verification of requirements and data patterns, the requirement expression is further optimized, ensuring that user requirements can be accurately translated into executable operational logic. This reduces the risk of unexpected data operation results due to misunderstandings of requirements.
[0030] S103: Execute the data operation code, obtain the execution result, and verify the execution result.
[0031] After obtaining the data manipulation code based on the above steps, the embodiments of this specification will execute the data manipulation code, obtain the execution result, and verify the execution result. Specifically, as follows... Figure 2 As shown in one or more embodiments of this specification, executing data manipulation code, obtaining execution results, and verifying the execution results specifically includes: The data manipulation code is run in a pre-defined isolated execution environment, such as a secure sandbox, to obtain execution results. It should be noted that the execution results include output data and execution status logs. To improve the accuracy and reliability of the results, this embodiment also introduces an execution verification mechanism to ensure that the generated code is not only syntactically correct but also logically sound, avoiding answers that are unrealistic. Specifically, firstly, the data manipulation code is subjected to syntax verification based on the execution status logs, such as checking for syntax errors and missing fields. If the syntax verification passes, logical verification is performed on the output data in the execution results, such as verifying negative sales figures or empty results that are not expected to be empty in a certain scenario. This logical verification includes at least: context constraint verification and reasonableness verification. If the logical verification passes, it is determined whether the output data matches the user requirements of key entities, thus obtaining the verification result of the execution results.
[0032] S104: If the verification fails, the verification failure information is fed back to the preset intelligent agent to drive the preset intelligent agent to update the data operation code and iterate the execution result.
[0033] To ensure the security of data operations and the validity of the results, if the verification fails based on step S103 above, as follows: Figure 2 The verification failure information is fed back to the preset intelligent agent to drive the preset intelligent agent to update the data operation code and iterate until the execution result verification is successful. Specifically, in one or more embodiments of this specification, feeding back the verification failure information to the preset intelligent agent to drive the preset intelligent agent to update the data operation code and iterate the execution result includes: The verification failure information is fed back to the pre-built agent, which identifies the error factors corresponding to the failure. For example, grammatical errors include keyword spelling problems and missing sentence structures; logical errors include mismatches between the time range and user requirements, deviations in indicator calculation logic, and conflicts between data and business context. This provides clear optimization directions for subsequent code adjustments. Since verification failures are essentially caused by insufficient error mitigation or unclear constraints in the prompts, the pre-built agent is driven to adjust its prompts based on the identified error factors. After adjustment, the agent will regenerate data manipulation code using the optimized prompts. The prompts are the core instructions driving the agent to generate code, containing key information such as user requirements and data patterns. The intelligent agent will optimize the prompts based on the identified error factors: for example, if the error factor is a grammatical incompatibility, the prompts will be supplemented with the grammatical constraints corresponding to the target data source; if the error factor is a logical deviation such as a mismatch in time range, the prompts will be strengthened with the requirement constraints of key entities and the time boundaries of data operations will be clarified; if the error factor is a conflict between data and business context, the prompts will be supplemented with the corresponding business logic rules.
[0034] After the agent regenerates the data manipulation code, it restarts the previous execution and verification process: the new code is run in a pre-defined isolated execution environment to obtain new execution results, followed by syntax and logic verification. If the verification still fails, new verification failure information is collected and fed back to the agent, which then repeats the process of identifying error factors, adjusting prompts, and regenerating the code. If the verification passes, the iteration stops. This loop continues until the execution result corresponding to the generated data manipulation code successfully passes all verification stages.
[0035] S105: If the verification is successful, the execution result will be fed back to the user.
[0036] To enhance user experience and ensure the readability, intuitiveness, and usability of feedback results, the verified execution results undergo format conversion to obtain a natural language summary. This involves extracting core data information from the execution results, such as key indicator values, time ranges, dimensional classifications, and aggregation results. Then, combining this with key entities relevant to user needs, and following the logical expression of natural language, the fragmented raw data is transformed into a logically coherent and clearly articulated natural language summary. Next, based on the current visualization requirements, the natural language summary is processed to generate corresponding visualization content. The obtained natural language summary is then integrated with the visualization content. On one hand, the natural language summary serves as textual explanations of the visualization content, annotated next to the graphics or at the top of the page, clearly defining the core theme, data range, and key conclusions of the visualization content. On the other hand, interactive functions are added to the visualization content to generate comprehensive feedback content, achieving interaction between text and graphics. The resulting comprehensive feedback content contains both precise textual information and intuitive graphical displays, balancing accuracy and intuitiveness. This comprehensive feedback content is then presented to the user, ensuring that the user can easily and accurately obtain and understand the execution results.
[0037] like Figure 3 As shown in the diagram, this specification provides a schematic diagram of the structure of a data analysis device based on an intelligent agent. Figure 3 As can be seen, in one or more embodiments of this specification, the device includes: The parsing unit 301 is used to receive data requests uploaded by the user terminal and to parse the data requests to obtain key entities; wherein, the request information includes: data query requests and data analysis requests; The generation unit 302 is used to obtain the metadata of the target data source based on the key entity, so as to trigger the preset intelligent agent to generate executable data operation code; The execution unit 303 is used to execute the data operation code, obtain the execution result, and verify the execution result; The iteration unit 304 is used to feed back the verification failure information to the preset intelligent agent if the verification fails, so as to drive the preset intelligent agent to update the data operation code and iterate the execution result; Feedback unit 305 is used to feed back the execution result to the user terminal if the verification is successful.
[0038] like Figure 4 As shown in the diagram, this specification provides a schematic diagram of the structure of a data analysis device based on an intelligent agent. Figure 4 As can be seen, in one or more embodiments of this specification, the device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any of the methods described above.
[0039] like Figure 5 As shown in the diagram, this specification provides a structural schematic of a non-volatile storage medium, which is composed of... Figure 5 As can be seen, one or more embodiments of this specification provide a non-volatile storage medium storing computer-executable instructions 501, which are capable of executing any of the methods described above.
[0040] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0041] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0042] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.
Claims
1. A data analysis method based on intelligent agents, characterized in that, The method includes: The system receives data requests uploaded by users and parses the data requests to obtain key entities; wherein the request information includes: data query requests and data analysis requests. Obtain the metadata of the target data source based on the key entity to trigger the pre-built intelligent agent to generate executable data operation code; Execute the data manipulation code, obtain the execution result, and verify the execution result; If the verification fails, the verification failure information is fed back to the preset intelligent agent to drive the preset intelligent agent to update the data operation code and iterate the execution result; If the verification passes, the execution result will be sent back to the user.
2. The data analysis method based on intelligent agents according to claim 1, characterized in that, Receiving data requests uploaded by the user client, and parsing the data requests to obtain key entities, specifically including: Receive data requests uploaded by the user and obtain the data input format corresponding to the data request; According to the data conversion strategy corresponding to the data input format, the data request is converted to obtain the data request text; The key entities of the data request text are extracted based on natural language processing; wherein, the key entities include: time, dimension, metric, and aggregation method.
3. The data analysis method based on intelligent agents according to claim 1, characterized in that, Based on the key entity, the metadata of the target data source is obtained to trigger the pre-built intelligent agent to generate executable data operation code, specifically including: Based on the target data source identifier corresponding to the key entity, determine the corresponding target data source; Obtain the metadata of the target data source to compare the data size of the metadata with a preset threshold; If the data size exceeds the preset threshold, the preset intelligent agent is triggered to generate executable data operation code; If the data size does not exceed the preset threshold, the data content of the target data source is input into the preset large language model to obtain the execution result and feed it back to the user.
4. The data analysis method based on intelligent agents according to claim 3, characterized in that, If the data size exceeds the preset threshold, the preset intelligent agent is triggered to generate executable data operation code, specifically including: If the data size exceeds the preset threshold, the agent mode of the preset intelligent agent is triggered and started. In the proxy mode, the data mode corresponding to the target data source is determined based on the structural information of the target data source; The user requirements defined by the key entities are integrated with the data pattern to construct prompt information, thereby driving the pre-set intelligent agent to generate an executable data operation script that matches the data pattern.
5. The data analysis method based on intelligent agents according to claim 4, characterized in that, Execute the data manipulation code, obtain the execution result, and verify the execution result, specifically including: The data manipulation code is run in a preset isolated execution environment to obtain the execution results; wherein, the execution results include: output data and execution status log; The data manipulation code is syntax-validated based on the execution status log. If the syntax validation passes, then the output data undergoes logical validation; wherein the logical validation includes at least: context constraint validation and rationality validation. If the logic verification passes, it is determined whether the output data matches the user requirements of the key entity, and the verification result of the execution result is obtained.
6. The data analysis method based on intelligent agents according to claim 1, characterized in that, The verification failure information is fed back to the pre-defined intelligent agent to drive the pre-defined intelligent agent to update the data operation code and iterate the execution result, specifically including: The verification failure information is fed back to the preset intelligent agent, so that the error factors corresponding to the verification failure information can be identified based on the preset intelligent agent; The preset intelligent agent is driven to adjust its prompt words according to the error factors, and the data operation code is regenerated according to the adjusted prompt words; The data manipulation code is updated and generated in a loop, the updated data manipulation code is executed, and the execution result of the updated data manipulation code is verified until the verification is successful and the execution result is obtained.
7. The data analysis method based on intelligent agents according to claim 1, characterized in that, The execution results are then fed back to the user, specifically including: The verified execution results are format-converted to obtain a natural language summary; Based on the current visualization requirements, the natural language summary is processed to generate corresponding visualization content; The natural language summary is integrated with the visualization content to generate comprehensive feedback content, which is then sent to the user.
8. A data analysis device based on intelligent agents, characterized in that, The device includes: The parsing unit is used to receive data requests uploaded by the user terminal and to parse the data requests to obtain key entities; wherein, the request information includes: data query requests and data analysis requests; The generation unit is used to obtain the metadata of the target data source based on the key entity, so as to trigger the pre-set intelligent agent to generate executable data operation code; An execution unit is used to execute the data manipulation code, obtain the execution result, and verify the execution result; An iteration unit is used to feed back verification failure information to the preset intelligent agent if the verification fails, so as to drive the preset intelligent agent to update the data operation code and iterate the execution result; The feedback unit is used to send the execution result back to the user terminal if the verification is successful.
9. A data analysis device based on intelligent agents, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1-7.
10. A non-volatile storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are capable of performing the method described in any one of claims 1-7.