Data analysis method, electronic device, storage medium and computer program product

Abstract

The invention discloses a data analysis method, electronic equipment, a storage medium and a computer program product, and relates to the technical field of data analysis, and the data analysis method comprises the following steps: determining a target data set and a target structured query language (SQL) statement according to acquired query information through a first large language model; executing the target SQL statement on the target data set to obtain a data query result; determining a target processing function according to the query information through a second large language model; processing the data query result according to the target processing function to obtain a data processing result; and generating a scene analysis conclusion through a third language model according to the data query result and the data processing result. According to the method and the device, the target SQL statement and the target processing function are automatically determined according to the query information, the method and the device are not limited to limitation of existing scenes, and processing of multiple scenes can be completed in one query process, so that the data analysis efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, and particularly relates to a data analysis method, an electronic device, a storage medium and a computer program product. BACKGROUND

[0002] With the rapid development of information technology, the data index processing process has changed from relying on manual script development to using ETL (Extract-Transform-Load) tools to realize standardized development, which effectively reduces the use threshold of users.

[0003] However, the data index processing and mining process through the ETL tool is usually highly bound to the use scene, and it is difficult to adapt to scene changes. The current data application scene is increasingly complex and variable. When a new scene is added, a large amount of manpower, material resources and time need to be invested to develop or adjust the processing flow of the data index, which seriously restricts the efficiency and effect of data index processing and mining.

[0004] The above content is only used to assist in understanding the technical solutions of the present application and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0005] The main purpose of the present application is to provide a data analysis method, an electronic device, a storage medium and a computer program product, which aims to solve the technical problem of how to improve the efficiency of data analysis.

[0006] To achieve the above purpose, the present application provides a data analysis method, which comprises:

[0007] determining a target data set and a target structured query language (SQL) statement according to the obtained query information through a first large language model;

[0008] executing the target SQL statement on the target data set to obtain a data query result;

[0009] determining a target processing function according to the query information through a second large language model;

[0010] processing the data query result according to the target processing function to obtain a data processing result;

[0011] generating a scene analysis conclusion according to the data query result and the data processing result through a third large language model.

[0012] In an embodiment, the step of determining a target data set and a target structured query language (SQL) statement according to the obtained query information comprises:

[0013] determine a text similarity between the query information and metadata of each data set, and determine a data set with a text similarity greater than a preset similarity threshold as a candidate data set;

[0014] in a case where a number of the candidate data sets does not exceed a preset number, determine the candidate data sets as target data sets;

[0015] in a case where the number of the candidate data sets exceeds the preset number, select the preset number of the candidate data sets as target data sets;

[0016] generate a target SQL statement according to the target data sets.

[0017] In an embodiment, the step of generating a target SQL statement according to the target data sets comprises:

[0018] obtain a historical SQL query statement set corresponding to the target data sets;

[0019] determine a historical SQL query statement with a greatest relevance to the query information in the historical SQL query statement set as a reference SQL statement;

[0020] rewrite the reference SQL statement according to the query information to obtain the target SQL statement.

[0021] In an embodiment, after the step of obtaining a data query result, the method further comprises:

[0022] determine whether the data query result meets a preset condition, wherein the preset condition comprises that a column number of the data query result is less than a preset column number threshold, the data query result only contains one date column and one numerical value column, and columns of the data query result cannot meet a preset calculation requirement;

[0023] in a case where the column number of the data query result is less than the preset column number threshold, perform full-amount processing on the data query result to obtain a data processing result;

[0024] in a case where the data query result only contains one date column and one numerical value column, determine data with a numerical value less than a preset threshold in the numerical value column as target data, aggregate the target data according to dates, and perform full-amount time series statistics on the aggregated target data to obtain a data processing result;

[0025] in a case where columns of the data query result cannot meet the preset calculation requirement, display the data query result;

[0026] In a case where the data query result does not satisfy the preset condition, the step of determining, by the second large language model, the target processing function according to the query information is performed.

[0027] In an embodiment, the target SQL statement includes a date range, and after the step of obtaining the data query result, the method further includes:

[0028] The date range in the target SQL statement is expanded to obtain an updated SQL statement, wherein the date range is a time limit for a query range of the target SQL statement;

[0029] The updated SQL statement is executed on the target data set to obtain a target query result;

[0030] A data mining result is obtained by performing data mining on the target query result through a preset machine learning algorithm;

[0031] An index analysis conclusion is generated by the third large language model according to the data mining result.

[0032] In an embodiment, after the step of generating the scene analysis conclusion, the method further includes:

[0033] The query information is supplemented by the third large language model according to the target data set to obtain a complete question;

[0034] The data query result, the data processing result, and the data mining result are converted into a form of a chart to obtain a plurality of data index analysis charts;

[0035] The complete question, the data index analysis charts, the scene analysis conclusion, and the index analysis conclusion are dynamically assembled according to a preset analysis report template to obtain a data analysis report.

[0036] In an embodiment, the data analysis method further includes:

[0037] For any target data set, the target data set and the query information are converted into task parameters by a preset master process, and are sent to a preset process pool;

[0038] Whether there is an idle child process is determined by the process pool;

[0039] In a case where there is an idle child process, the task parameters are added to a task queue of the process pool to wait for reading and execution by the idle child process;

[0040] The target SQL statement is determined according to the task parameter and the first large language model through the idle sub-process, and the target SQL statement is executed on the target data set to obtain a data query result.

[0041] In addition, to achieve the above object, the present application also provides a data analysis device, which comprises:

[0042] The SQL statement determination module is configured to determine a target data set and a target structured query language (SQL) statement according to the obtained query information through a first large language model.

[0043] The data query module is configured to execute the target SQL statement on the target data set to obtain a data query result.

[0044] The processing function determination module is configured to determine a target processing function according to the query information through a second large language model.

[0045] The data processing module is configured to process the data query result according to the target processing function to obtain a data processing result.

[0046] The analysis conclusion generation module is configured to generate a scene analysis conclusion according to the data query result and the data processing result through a third large language model.

[0047] In addition, to achieve the above object, the present application also provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the data analysis method as described above.

[0048] In addition, to achieve the above object, the present application also provides a storage medium, which is a computer readable storage medium, and a computer program is stored in the storage medium, wherein the computer program is executed by a processor to implement the steps of the data analysis method as described above.

[0049] In addition, to achieve the above object, the present application also provides a computer program product, which comprises a computer program, wherein the computer program is executed by a processor to implement the steps of the data analysis method as described above.

[0050] The one or more technical solutions provided in the application have at least the following technical effects: first, by the first large language model, the target data set and the target SQL (Structured Query Language) statement are determined according to the obtained query information, which avoids syntax errors or logical deviations that may occur when manually writing SQL, reduces the user threshold, and ensures the efficiency of data retrieval; then the target SQL statement is executed on the target data set to obtain a data query result; then by the second large language model, the target processing function is determined according to the query information, which realizes intelligent matching and automatic determination of data processing logic, without relying on manual screening and testing of different processing functions according to the scene, thereby improving the efficiency of the data processing link; then the data query result is processed according to the target processing function to obtain a data processing result, and a scene analysis conclusion is generated according to the data query result and the data processing result by the third large language model. The application can automatically determine the target data set and generate the SQL query statement according to the user input query information, thereby reducing the learning cost of learning the syntax knowledge of the SQL statement and improving the data query efficiency; then the target processing function is automatically called according to the query information, rather than being limited to a fixed scene processing flow in the ETL tool, thereby breaking through the limitation of the fixed processing flow corresponding to each scene in the ETL tool. Therefore, the data index processing flow of the application can realize cross-scene reuse, avoiding repeated development when a new scene is added, thereby improving the efficiency of data analysis. BRIEF DESCRIPTION OF DRAWINGS

[0051] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the application and, together with the specification, serve to explain the principles of the application.

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without creative labor.

[0053] Figure 1 The flowchart provided for the data analysis method embodiment one of the application;

[0054] Figure 2 The architecture diagram of the special process pool provided for the third embodiment of the application;

[0055] Figure 3 The brief flowchart of the data analysis method provided for the third embodiment of the application;

[0056] Figure 4 The module structure diagram of the data analysis device of the application;

[0057] Figure 5 Figure 1 is a schematic diagram of a device structure of a hardware operating environment involved in a data analysis method in an embodiment of the present application.

[0058] The object implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0059] It should be understood that the specific embodiments described herein are merely intended to explain the technical solutions of the present application, and are not intended to limit the present application.

[0060] In order to better understand the technical solutions of the present application, the specific embodiments will be described in detail below with reference to the drawings and the specific embodiments.

[0061] At present, the implementation of data index processing and mining mainly includes two categories of methods: manual script method and ETL tool method. The manual script method relies on professional data analysts to manually write codes to realize customized processing and mining of data, which has high labor cost and slow response speed. The visual components of the ETL tool method are fixed, the tool process is deeply coupled with a specific scene, and a complete processing flow needs to be redeveloped for a new scene, which has low processing efficiency. It can be seen that the current data analysis methods are highly bound to the use scene, and cannot realize cross-scene reuse, resulting in low overall processing efficiency of data analysis.

[0062] The present application provides a solution. First, a first large language model is used to determine a target data set and a target SQL statement according to the obtained query information, which avoids syntax errors or logical deviations that may occur when manually writing SQL, reduces the user threshold, and ensures the efficiency of data retrieval. Then, the target SQL statement is executed on the target data set to obtain a data query result. Then, a second large language model is used to determine a target processing function according to the query information, which realizes intelligent matching and automatic determination of data processing logic, without relying on manual identification and testing of different processing functions according to the scene, thereby improving the efficiency of the data processing link. Then, the data query result is processed according to the target processing function to obtain a data processing result, and a third large language model is used to generate a scene analysis conclusion according to the data query result and the data processing result. The present application can automatically determine a target data set and generate a SQL query statement according to the user input query information, thereby reducing the learning cost of learning SQL statement syntax knowledge and improving the data query efficiency. Then, the target processing function is automatically called according to the query information, rather than being limited to a fixed scene processing flow in the ETL tool, thereby breaking through the limitation of the fixed processing flow corresponding to each scene in the ETL tool. Therefore, the data index processing flow of the present application can realize cross-scene reuse, avoiding repeated development when a new scene is added, thereby improving the efficiency of data analysis.

[0063] It should be noted that the execution subject of the embodiment can be an electronic device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., and a large language model is deployed on the electronic device.

[0064] Based on this, the embodiment of the present application provides a data analysis method, which is described with reference to Figure 1 , Figure 1 The flowchart of the first embodiment of the data analysis method of the present application is shown in FIG. 1.

[0065] In this embodiment, the data analysis method comprises steps S10-S50:

[0066] Step S10, determining a target data set and a target structured query language (SQL) statement according to the obtained query information through a first large language model.

[0067] The large language model (LLM) refers to a pre-trained deep neural network model with natural language understanding and generation capabilities. Common large language models include GPT-4, DeepSeek, Qwen, etc. The first large language model represents a large language model used to determine the target data set and the target SQL statement.

[0068] The query information refers to the text information input by the user to express the data content and analysis requirements that he / she wants to query, which can include keywords, data range, time conditions, etc.

[0069] The target data set refers to a specific data set related to the query information in the preset database, which can be one or multiple.

[0070] SQL (Structured Query Language) is a standard programming language for managing relational databases, and the target SQL statement is an instruction code generated according to the query information for searching data in the target data set.

[0071] Optionally, the above-mentioned first large language model can be directly applied to the data analysis process, or the first large language model can be pre-trained for SQL generation before being applied to the data analysis.

[0072] Exemplarily, first, the query information input by the user is acquired, and the received query information is passed to the first pre-trained large language model as input data of the model; then, the query information is subjected to semantic understanding, key information extraction and the like by the first large language model; then, according to the analysis result of the query information by the first large language model, at least one target data set meeting the requirement is searched for in the database; and simultaneously, a target SQL statement consistent with the query information and capable of accurately querying data from the target data set is generated according to the SQL syntax rules and logic.

[0073] It can be understood that the target SQL statement is generated by the first large language model, which can eliminate the technical threshold of manually writing SQL, so that non-technical personnel can also directly operate.

[0074] Step S20, the target SQL statement is executed on the target data set to obtain a data query result;

[0075] The data query result refers to a data set returned after the target SQL statement is executed, which contains all data records meeting the condition expression in the target SQL statement filtered from the target data set and is organized and presented in a certain format (such as a table, a list, etc.).

[0076] Exemplarily, after the target SQL statement and the target data set are determined, the data records in the target data set are filtered one by one according to the conditions in the target SQL statement, and the data records meeting the conditions are extracted to be arranged into the data query result in a target format, wherein the target format includes markdown, html (hypertext mark-up language), etc.

[0077] Step S30, a target processing function is determined according to the query information by a second large language model;

[0078] The second large language model represents a large language model for determining the target processing function; the first large language model and the second large language model can be the same, or different sizes of large language models can be matched according to the resource consumption degree of the tasks to be executed by each of them.

[0079] The target processing function refers to a pre-defined code logic unit with a specific function, which is used for calculating, converting, cleaning and the like of data, for example, when calculating the sales growth rate, the target processing function can be a function for calculating the same period or same cycle growth rate.

[0080] Optionally, the second large language model is used to perform semantic analysis on the query information input by the user to determine the data processing intention of the user, such as numerical statistics, time series analysis, or time series statistics, etc.; and then the second large language model searches and matches in the preset processing function library according to the analyzed data processing intention to determine the target processing function that meets the data processing intention.

[0081] Optionally, the numerical statistics include summation, frequency statistics, mean, maximum, minimum, median, standard deviation, 1 / 4 quantile, 3 / 4 quantile, skewness, kurtosis, date of maximum value, date of minimum value, and other common numerical statistics.

[0082] Optionally, the time series statistics include day-on-day, week-on-week, month-on-month, year-on-year, moving average, trend slope, abnormal value score, 7-day periodicity, month periodicity, quarter periodicity, year periodicity, 7-day quantile, month quantile, quarter quantile, year quantile, and other common time series statistics.

[0083] Optionally, the time series analysis includes historical trend fitting, abnormal interval identification, future trend prediction, and other analysis and calculation.

[0084] In a feasible implementation, after step S20, the method further includes:

[0085] Step S301: determining whether the data query result meets a preset condition, wherein the preset condition includes that the number of columns of the data query result is less than a preset column threshold, the data query result only contains one date column and one numerical value column, and the columns of the data query result cannot meet a preset calculation requirement.

[0086] Optionally, the preset condition refers to a set of pre-defined logical rules for determining whether the data query result needs to be further processed by the second large language model. The preset condition can involve data dimensions, field types, calculation feasibility, etc., such as requiring the number of columns of the data query result to be less than a preset column threshold, requiring the data query result to only contain one date column and one numerical value column, and requiring the columns of the data query result to be unable to meet a preset calculation requirement.

[0087] The date column refers to a column in the data query result that stores date type data, usually represented by DATE, DATETIME, etc., and records date information related to the data. The numerical value column refers to a column in the data query result that stores numerical value type data, usually represented by INT, FLOAT, DOUBLE, etc., and contains numerical value information that can be subjected to mathematical operations.

[0088] The calculation requirement refers to the minimum data completeness condition required for the execution of the processing function, including input parameter requirements, etc.

[0089] In a case where the number of columns of the data query result is less than the preset column threshold, full processing is performed on the data query result to obtain a data processing result.

[0090] Full processing refers to comprehensive processing operations on all data records in the data query result, including all processing operations that can be performed on the data query result. After the full processing is completed to obtain the data processing result, the step of generating a scene analysis conclusion according to the data query result and the data processing result by the third large language model is directly performed.

[0091] For example, assuming that in a class score calculation scene, the preset column threshold is 3, and the actual data query result only includes a total score column, the total score can be directly processed by summation, mean, maximum, minimum, median, standard deviation, 1 / 4 quantile, 3 / 4 quantile, etc., to obtain and output the data processing result of the total score.

[0092] It can be understood that in a case where the number of columns of the data query result is less than the preset column threshold, since the data volume is small, full processing can be directly performed, bypassing the large language model, reducing resource consumption and time of the large language model for analyzing query information, and being beneficial to improving data analysis efficiency.

[0093] In a case where the data query result only includes one date column and one numerical value column, the data with a numerical value less than a preset threshold in the numerical value column is determined as target data, the target data is aggregated according to the date, full time series statistics is performed on the aggregated target data, and a data processing result is obtained.

[0094] Time series statistics refers to a process of statistical analysis on data arranged in time sequence, which is usually used to observe the change trend of data in the time dimension. The specific time series statistics method is the same as in step S30, and thus is not described herein. After the full time series statistics is completed to obtain the data processing result, the step of generating a scene analysis conclusion according to the data query result and the data processing result by the third large language model is directly performed.

[0095] Optionally, in a case where the data query result only includes one date column and one numerical value column, all rows with a numerical value less than a threshold in the data query result can be summarized according to a unique date to realize data aggregation, and complete time series statistics can be performed on each day (and / or each week, each month).

[0096] In a case where the columns of the data query result cannot meet the preset calculation requirement, the data query result is displayed.

[0097] Exemplarily, the type, quantity, range, etc. of the column values of the data query result are analyzed to determine whether they can meet the preset computing requirement; when it is detected that the column values of the data query result cannot meet the preset computing requirement, data processing is no longer performed, but a display instruction is triggered to convert the data query result into a target format and output to the interactive interface of the electronic device. At the same time, a scene analysis conclusion can be generated directly through the third large language model according to the data query result, without the need for further processing through a processing function.

[0098] It can be understood that in the case where the preset computing requirement is not met, the result is directly displayed without subsequent processing by the user, thereby avoiding invalid computation and system abnormalities.

[0099] In step S305, in the case where the data query result does not meet the preset condition, a step of determining a target processing function according to the query information through the second large language model is performed.

[0100] In this embodiment, by analyzing the target data set in advance, the number of calls to the large language model is reduced, thereby improving the efficiency of data analysis.

[0101] In step S40, the data query result is processed according to the target processing function to obtain a data processing result.

[0102] The data processing result refers to the data output after processing by the target processing function, and is a new data form or content obtained after a series of operations on the original data query result, which can better meet the specific needs or analysis purposes of the user.

[0103] Optionally, the second large language model can autonomously determine whether a processing function is needed and which processing function (such as a mean calculation function) to use according to the query information; and then, after determining the target processing function, the input parameters of the information transmission structure corresponding to the target processing function are determined, for example, in the case where the query information is "calculate the monthly return rate mean of a certain stock in the past year", the target processing function determined by the second large language model can be a mean calculation function, and the information transmission structure of the target processing function needs the return rate information of the stock in the past year as input parameters; and then, the corresponding data is called from the data query result as input parameters, filled into the target processing function, and the processing output of the target processing function is taken as the data processing result.

[0104] In step S50, a scene analysis conclusion is generated through the third large language model according to the data query result and the data processing result.

[0105] The third large language model is configured to generate a scene analysis conclusion based on the data query result and the data processing result. The third large language model can be the same as the first large language model and the second large language model, or can be configured according to the task resource consumption. The present embodiment does not make specific limitations in this regard.

[0106] The scene analysis conclusion refers to a comprehensive analysis and summary of the data query result and the data processing result generated by the third large language model, and is presented in the form of natural language, usually including information such as explanation of data, trend analysis and problem diagnosis.

[0107] For example, the data query result is all order data of a certain product in the past year from a target data set, including order date, product name, sales amount, etc. The data processing result is obtained after processing these order data, such as monthly sales, sales trend, etc. The third large language model combines these data and may generate a scene analysis conclusion: “the monthly average sales of the product is xx, and the sales shows a monthly downward trend”.

[0108] The present embodiment provides a data analysis method, which automatically determines a target data set and generates a SQL query statement according to user input query information, to reduce the learning cost of learning SQL statement syntax knowledge and improve data query efficiency. Then, the target processing function is automatically called according to the query information, rather than being limited to a fixed scene processing flow in the ETL tool, breaking the limitation of the fixed processing flow corresponding to each scene in the ETL tool. Therefore, the data index processing flow of the present application can realize cross-scene reuse, avoiding repeated development when a new scene is added, thereby improving the efficiency of data analysis.

[0109] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as the above embodiment one can be referred to the above introduction, and will not be repeated hereinafter. On this basis, the step of determining the target data set and the target structured query language (SQL) statement according to the obtained query information in step S10 includes:

[0110] In step S11, the text similarity between the query information and the metadata of each data set is determined, and the data set with a text similarity greater than a preset similarity threshold is determined as a candidate data set.

[0111] The metadata of the data set refers to structured information describing the characteristics of the data set, including field names, data types, and data ranges of the data set.

[0112] The text similarity is an index for measuring the similarity of two texts in semantics or content, which is usually calculated by a specific algorithm (such as cosine similarity, Jaccard similarity, etc.), and the value range is usually between 0 and 1.

[0113] The candidate data set refers to a data set with high relevance to the query information after text similarity screening, and is used for further analysis or processing.

[0114] Exemplarily, after obtaining the query information input by the user, the first large language model obtains the metadata of each data set; then, the query information and the metadata are preprocessed to improve the accuracy of similarity calculation; then, a text similarity algorithm is called to calculate the text similarity between the query information and the metadata of each data set, and after the calculation is completed, the text similarity of each data set is compared with a preset threshold value, and the data set with a similarity higher than the threshold value is selected as the candidate data set.

[0115] Step S12, in the case where the number of candidate data sets does not exceed the preset number, the candidate data sets are determined as the target data sets;

[0116] Step S13, in the case where the number of candidate data sets exceeds the preset number, the preset number of candidate data sets are selected as the target data sets;

[0117] The preset number is a pre-set integer value for limiting the maximum number of target data sets, and is stored in the configuration parameters of the program.

[0118] Exemplarily, the target data sets are determined according to whether the number of candidate data sets exceeds the preset number. In the case where the number of candidate data sets does not exceed the preset number, all candidate data sets are directly determined as the target data sets; and in the case where the number of candidate data sets exceeds the preset number, the preset number of candidate data sets are selected as the target data sets according to certain selection logic (such as sorting by similarity, sorting by data set size, sorting by user access frequency, etc.).

[0119] Optionally, in the case where the number of candidate data sets exceeds the preset number, the preset number of candidate data sets can be directly selected as the target data sets according to the text similarity of each candidate data set in the order from high to low.

[0120] Optionally, in the case where the number of candidate data sets exceeds the preset number, the preset number of data sets with the highest access frequency can also be selected from the candidate data sets as the target data sets in combination with the access frequency of the user inputting the current query information.

[0121] Step S14, generating a target SQL statement according to the target data set.

[0122] Optionally, for any target data set, read all metadata of the target data set (including table structure, field type, sample value, etc.); then, align the entities in the query information with the metadata using the first large language model, and generate multiple candidate SQL skeletons (including SELECT, FROM, JOIN, WHERE, ORDER BY, LIMIT, etc. nodes) according to the alignment results; then, complete the possible predicates (such as time range, aggregation function, etc.) to the candidate SQL skeleton through the first large language model, reduce the data query range, and obtain the candidate SQL statement; then, perform syntax checking and null result detection on the candidate SQL statement, if failed, trigger the model to regenerate, if successful, determine the candidate SQL statement as the target SQL statement and output.

[0123] In a feasible implementation, step S14 comprises:

[0124] Step S141, obtaining a historical SQL query statement set corresponding to the target data set;

[0125] The historical SQL query statement refers to the SQL text and execution metadata stored in the audit log corresponding to the target data set, and the set of historical SQL query statements corresponding to each target data set is referred to as a historical SQL query statement set.

[0126] Step S142, determining the historical SQL query statement with the highest relevance to the query information in the historical SQL query statement set as the reference SQL statement;

[0127] The relevance refers to the degree of association between the current query information and the historical SQL query statement calculated by an algorithm (such as text similarity, semantic matching degree), which is used to quantify the matching degree of the two.

[0128] Optionally, according to the identifier of the target data set, all historical SQL query statements executed for the target data set can be retrieved from the query log or the history record table of the database to form the historical SQL query statement set; for each statement in the historical SQL statement set, the relevance value between the current query information and the statement is calculated through a text similarity algorithm (such as cosine similarity), a semantic matching algorithm (such as BERT model) or a rule-based matching algorithm; then, compare the relevance values of all historical SQL statements, and select the historical SQL statement with the highest relevance value as the reference SQL statement.

[0129] Step S143, rewriting the reference SQL statement according to the query information to obtain the target SQL statement.

[0130] Optionally, the reference SQL statement is first converted into a syntax parsing tree by using a SQL parser, and the table name, field, condition and other structures are determined; then the natural language condition in the query information is matched with a rewriting rule set to determine the rule to be applied, wherein the rewriting rule set refers to a set of predefined rules (such as a "time range replacement rule", a "field mapping rule" and the like), which is used to guide how to map the natural language condition in the query information to the specific parameters of the SQL statement, such as replacing "the last three months" with time >= '2025-04-01' AND time <= '2025-07-31'; then, according to the matched rule, the corresponding parameter (such as the time range) in the syntax parsing tree is replaced, and the modified syntax parsing tree is recombined into a complete SQL statement to obtain the target SQL statement.

[0131] Optionally, after obtaining the target SQL statement, the dialect converter can be used for rewriting, such as rewriting the SQL statement adapted to the MySQL database into the SQL statement adapted to the BigQuery SQL database, to support different databases.

[0132] In this embodiment, the related dataset is recommended according to the query information, and the reference SQL statement is determined according to the relevance between the historical SQL query statement of the related dataset and the query information, and the target SQL statement is obtained by rewriting the reference SQL statement, so as to generate a SQL statement that meets the user's demand and is grammatically correct.

[0133] Based on the first and / or second embodiments of the present application, in the third embodiment of the present application, the same or similar contents as the above-mentioned first and second embodiments can be referred to the above description, and will not be repeated hereinafter. On this basis, the target SQL statement contains a date range, and after step S20, it further includes:

[0134] Step S21, expanding the date range in the target SQL statement to obtain an updated SQL statement, wherein the date range is a time limit for the query range of the target SQL statement;

[0135] The date range refers to the time interval expression in the WHERE clause of the target SQL statement, such as BETWEEN start_date AND end_date, which is a time limit for the query range of the target SQL statement.

[0136] The updated SQL statement is a new query instruction generated by modifying the date range based on the target SQL statement, which retains other query conditions (such as tables, fields, etc.) of the target SQL statement and only adjusts the date range to achieve different time span data query requirements.

[0137] Exemplarily, the date range condition in the target SQL statement is extracted and parsed into a computer-processable date format (such as a timestamp, a date object); then, according to a preset date expansion rule, a new date range is calculated, and the original date range in the target SQL statement is replaced with the calculated new date range to generate an updated SQL statement. The preset date expansion rule can be to move the starting date forward by one month, or to automatically expand the interval based on the time semantics, such as expanding a monthly date range to a quarterly date range, and the embodiment is not limited in a specific date expansion rule.

[0138] Step S22, executing the updated SQL statement on the target data set to obtain a target query result;

[0139] The target query result refers to a data set meeting the conditions extracted from the target data set after executing the updated SQL statement, which is the input data of data mining.

[0140] Step S23, performing data mining on the target query result through a preset machine learning algorithm to obtain a data mining result;

[0141] The preset machine learning algorithm is a mathematical model defined in advance for extracting patterns or predicting results from data. Common machine learning algorithms include decision trees, neural networks, clustering algorithms, etc., and the embodiment is not limited to a specific machine learning algorithm.

[0142] The data mining result is the output generated by the machine learning algorithm after analyzing the target query result, which can be historical statistical analysis results, future time series prediction, etc.

[0143] Optionally, before performing data mining on the target query result, it can be processed for data cleaning and feature engineering, such as deleting duplicate records, filling missing values, and encoding categorical variables into numerical values, so as to improve the data quality of the target query result and further improve the prediction effect of the machine learning algorithm.

[0144] Exemplarily, the GMV (Gross Merchandise Volume) prediction can be achieved through the LightGBM (Light Gradient Boosting Machine) algorithm, and the user clustering can be achieved through the MiniBatchKMeans algorithm.

[0145] Step S24, generating an index analysis conclusion according to the data mining result through a third large language model.

[0146] The index analysis conclusion is a natural language text output generated by the third large language model based on the data mining result, using the internal language logic and knowledge system, for explaining the phenomenon, trend or problem reflected by the data mining result.

[0147] It can be understood that, by performing data mining on the target query result after expanding the date range, a longer time span can be covered, thereby providing more comprehensive and accurate trend information and other data mining results.

[0148] In a feasible implementation, after step S50, further comprising:

[0149] Step S51, supplementing the query information according to the target data set by the third large language model to obtain a complete question;

[0150] Optionally, the complete question refers to a natural language sentence completed by the large language model, which contains clear indicators, dimensions, filters, and time ranges, and is used to accurately show the processing target of the current data processing and mining. For example, in the case where the query information is "analyze the test data in May" and the target data set is determined to be "the monthly test record data set of a certain product", the complete question can be obtained as "analyze the test result pass rate and reason distribution of a certain product under the types of function test and performance test from May 1, 2025 to May 31". This facilitates the user to understand the content displayed by the current data query result, data processing result and data mining result.

[0151] Step S52, converting the data query result, data processing result and data mining result into a form of a chart to obtain a plurality of data index analysis charts;

[0152] The data index analysis chart refers to a graphical display result of data, which is used to intuitively present the characteristics, trends and relationships of data indicators. Common chart types include column chart, line chart, pie chart, scatter chart, etc. Each chart is suitable for different data types, such as presenting the data query result in the form of a table, converting the proportional data obtained after processing into a pie chart, and displaying time series data containing dates and numerical values in an Echarts chart.

[0153] It can be understood that, by displaying various types of data in the form of charts, the user can quickly capture the core characteristics and trends of the data, thereby improving the efficiency of data understanding.

[0154] Step S53, dynamically assembling the complete question, each data index analysis chart, scene analysis conclusion and index analysis conclusion according to a preset analysis report template to obtain a data analysis report.

[0155] The preset analysis report template is an electronic document framework with specific structure and format, in which the positions, styles, fonts, and other attributes of various components in the data analysis report are defined, as well as the logical relationship and layout rules between the components, and is usually stored in XML, JSON, or the like, facilitating reading and processing by electronic devices.

[0156] The data analysis report refers to a final document generated by assembling the complete question, data index analysis chart, scene analysis conclusion, and index analysis conclusion according to the preset template, which contains the whole process and results of data analysis, and is used to show and report the data analysis to the user.

[0157] Optionally, the dynamic assembly refers to a template instantiation process based on data driving. For example, the placeholders of the analysis report template can be parsed to correspond to the data contents of the complete question, chart object, scene analysis conclusion, and index analysis conclusion, respectively. If the data value of the index analysis conclusion is empty, the title and display area of the "index analysis conclusion" are deleted. Then, the data analysis report is obtained by automatically rendering according to the size of the data content and adjusting the display effect.

[0158] In this embodiment, the data mining is performed by expanding the date range, and the index analysis conclusion is generated, which improves the completeness of data analysis. After the analysis is completed, the corresponding data results are automatically converted into the form of a chart, and are assembled with the data analysis conclusion to present in the form of a data analysis report, avoiding the tedious process of manually writing a report, shortening the report generation time, and improving the efficiency of data analysis.

[0159] In a feasible implementation manner, the data analysis method further includes:

[0160] In step A10, for any target data set, the target data set and the query information are converted into task parameters by a preset master process, and are sent to a preset process pool.

[0161] The master process refers to a core process in the electronic device responsible for coordinating and managing other sub-processes, having high authority and control ability, and being capable of receiving external input (such as target data set and query information), performing data processing and task allocation, and sending the processing results to the corresponding module or process pool.

[0162] The task parameters refer to the parameter set converted by the master process according to the target data set and the query information, which is used to guide the sub-processes in the process pool to perform specific tasks. These parameters contain data retrieval conditions, data processing methods, output formats, and other information, ensuring that the sub-processes can accurately complete the tasks. The data processing methods include SQL generation, SQL execution, data processing, generation of scene analysis conclusion, data mining, generation of index analysis conclusion, and generation of data analysis report.

[0163] Optionally, after the target data set is determined by the first large language model, since the target data set can include one or more, in order to realize efficient data query and processing, the target data set and the query information can be converted into task parameters by the main process and sent to the preset process pool for any target data set; then the sub-processes in the process pool can execute the subsequent steps of determining the target SQL statement, executing data query and data processing in parallel after obtaining the parameters, so as to realize simultaneous processing of multiple target data sets and simultaneous answering for different target data sets, thereby improving the efficiency of data analysis.

[0164] Step A20, determining whether there is an idle sub-process in the process pool;

[0165] The process pool is a collection of sub-processes created in advance, managed by the main process, used for parallel processing of tasks and improving processing efficiency; and the process pool reuses sub-processes to avoid system overhead caused by frequent creation and destruction of processes, thereby improving system performance and resource utilization.

[0166] The idle sub-process refers to a sub-process in the process pool that is currently not executing any task and is in a waiting state, ready to be assigned a new task by the main process.

[0167] Optionally, the main process can dynamically adjust the number of sub-processes in the process pool according to the number of tasks and the resource situation of the electronic device.

[0168] Optionally, after receiving the task parameters, the process pool can first verify them. For example, the target data set can be verified first to ensure that it is real and accessible, and then it can be determined whether there is an idle sub-process to reduce resource consumption.

[0169] Step A30, in the case where there is an idle sub-process, adding the task parameters to the task queue of the process pool to wait for the idle sub-process to read and execute;

[0170] The task queue refers to a data structure used to store task parameters to be executed, which manages tasks in a certain order (such as first-in, first-out).

[0171] Optionally, in the case where there are multiple idle sub-processes in the process pool, each idle sub-process queries the task queue of the process pool through round robin to determine whether there is a task to be executed.

[0172] Step A40, determining the target SQL statement according to the task parameters and the first large language model by the idle sub-process, and executing the target SQL statement on the target data set to obtain the data query result and subsequent steps.

[0173] In this embodiment, through the idle sub-process, according to the query information and the target data set in the task parameter, the specific implementation of the target SQL statement is determined by means of the first large language model. For reference, the specific implementation of step S14 in the second embodiment is described above, and will not be repeated here.

[0174] Exemplarily, please refer to Figure 2 , Figure 2 An architecture diagram of a dedicated process pool is provided. First, the main process creates a process pool for managing and distributing tasks to multiple sub-processes. Then, the process pool creates multiple sub-processes for parallel processing of tasks, completing the architecture design of the dedicated process pool. Then, the process pool executes the initialization function to ensure that the sub-processes are ready to process tasks. After the sub-processes complete initialization, they send a signal to the process pool indicating that initialization is complete. The main process is notified by the process pool that initialization is complete, and can distribute tasks. Then, the dedicated process pool can be used to execute parallel tasks. After receiving a task request, the main process sends the task that needs to be processed in parallel to the process pool. Then, the process pool checks the task information and sends the task to the sub-process for processing after the check is passed. After the sub-process completes the task processing, it returns the task result to the process pool, which then returns the task result to the main process. Through this architecture, parallel processing of tasks can be achieved, improving the efficiency of data analysis.

[0175] In this embodiment, through the process pool architecture, different target data sets can be processed in parallel for SQL generation, SQL execution, scene calculation, and index analysis. At the same time, different data sets are analyzed and answers for different data sets are generated in parallel, thereby improving the efficiency of data analysis.

[0176] Exemplarily, to help understand the implementation process of the data analysis method obtained after combining the above-mentioned embodiments one and two, please refer to Figure 3 , Figure 3 A brief flowchart of a data analysis method is provided. Specifically:

[0177] First, generate SQL using a large language model and query information to obtain a target SQL statement.

[0178] Then, perform SQL execution to execute the target SQL statement in the target data set to obtain a data query result, wherein the target data set is determined according to the query information before SQL generation.

[0179] Then, process the data query result to obtain a data processing result, wherein the target processing function for data processing is also automatically determined by the large language model and the query information.

[0180] Further, the data query result and the data processing result are integrated into overall data, and are analyzed by the large language model to generate a scene calculation conclusion.

[0181] In addition, the date range in the target SQL statement can be further expanded, and the target query result is obtained after query; further, data mining is performed on the target query result to obtain a data mining result, and the large language model is used to analyze the data mining result to generate an index analysis conclusion.

[0182] Meanwhile, in combination with the architecture design of the special process pool, the serial processing flow from the SQL generation of the current target data set to the generation of the index analysis conclusion can be executed in parallel with the serial processing flow of other data sets to improve the efficiency of data analysis.

[0183] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the data analysis method of the present application. More forms of simple transformation based on this technical concept are within the protection scope of the present application.

[0184] The present application also provides a data analysis device, please refer to Figure 4 The data analysis device comprises:

[0185] The SQL statement determination module 10 is configured to determine the target data set and the target structured query language (SQL) statement according to the obtained query information by using the first large language model.

[0186] The data query module 20 is configured to execute the target SQL statement on the target data set to obtain a data query result.

[0187] The processing function determination module 30 is configured to determine the target processing function according to the query information by using the second large language model.

[0188] The data processing module 40 is configured to process the data query result according to the target processing function to obtain a data processing result.

[0189] The analysis conclusion generation module 50 is configured to generate a scene analysis conclusion according to the data query result and the data processing result by using the third large language model.

[0190] The data analysis device provided by the present application adopts the data analysis method in the above embodiments, and can solve the technical problem of how to improve the efficiency of data analysis. Compared with the prior art, the data analysis device provided by the present application has the same beneficial effects as the data analysis method provided by the above embodiments, and other technical features in the data analysis device are the same as the features disclosed in the above embodiments, which will not be repeated here.

[0191] The electronic device provided in the embodiments of the present application includes at least one processor, and a memory connected with the at least one processor in communication; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the data analysis method in the first embodiment.

[0192] Reference is made below in detail to Figure 5 , which shows a structural schematic diagram of an electronic device suitable for implementing the embodiments of the present application. The electronic device in the embodiments of the present application can include, but is not limited to, a mobile terminal such as a mobile phone, a notebook computer, a digital broadcast receiver, a PDA (Personal Digital Assistant), a PAD (Portable Application Description), a PMP (Portable Media Player), a vehicle terminal (for example, a vehicle navigation terminal), and the like, and a fixed terminal such as a digital TV, a desktop computer, and the like. Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0193] As shown in Figure 5 , the electronic device can include a processing device 1001 (for example, a central processor, a graphic processor, and the like) that can perform various appropriate actions and processes according to programs stored in a read-only memory 1002 or loaded from a storage device 1003 into a random access memory 1004. Various programs and data required for the operation of the electronic device are also stored in the random access memory 1004. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are connected to each other through a bus 1005. An input / output interface 1006 is also connected to the bus. Generally, the following systems can be connected to the input / output interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, and the like; an output device 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, and the like; the storage device 1003 including, for example, a magnetic tape, a hard disk, and the like; and a communication device 1009. The communication device 1009 can allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although the electronic device with various systems is shown in the figure, it should be understood that all the systems shown are not required to be implemented or provided. More or fewer systems can be alternatively implemented or provided.

[0194] In particular, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a read-only memory 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the methods of the embodiments of the present application are performed.

[0195] The electronic device provided by the embodiments of the present application adopts the data analysis method in the above-mentioned embodiments, and can solve the technical problem of how to improve the efficiency of data analysis. Compared with the prior art, the electronic device provided by the present application has the same beneficial effects as the data analysis method provided by the above-mentioned embodiments, and other technical features in the electronic device are the same as the features disclosed in the previous embodiment method, which will not be repeated here.

[0196] It should be understood that parts of the present application can be realized by hardware, software, firmware or a combination thereof. In the description of the above-mentioned embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0197] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0198] The embodiments of the present application provide a computer readable storage medium having stored thereon computer readable program instructions (i.e. computer program) for performing the data analysis method in the above-mentioned embodiments.

[0199] The computer readable storage medium provided by the embodiments of the present application may, for example, be a U disk, but is not limited to an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system or device, or any combination thereof. More specific examples of the computer readable storage medium may include, but are not limited to, an electric connection with one or more conductive wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the embodiments, the computer readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer readable storage medium can be transmitted by any suitable medium, including but not limited to an electric wire, an optical cable, an RF (Radio Frequency), and the like, or any suitable combination thereof.

[0200] The computer readable storage medium described above may be contained in an electronic device, or may exist separately without being assembled into an electronic device.

[0201] The computer readable storage medium described above carries one or more programs, which, when executed by an electronic device, cause the electronic device to: determine, by a first large language model, a target data set and a target structured query language (SQL) statement according to acquired query information; execute the target SQL statement on the target data set to obtain a data query result; determine, by a second large language model, a target processing function according to the query information; process the data query result according to the target processing function to obtain a data processing result; and generate, by a third large language model, a scene analysis conclusion according to the data query result and the data processing result.

[0202] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0203] The flow diagrams and the block diagrams in the drawings are illustrations of architectures, functionalities, and operations of possible implementations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams or block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks may

[0204] The modules involved in the embodiments of the present application can be implemented in the form of software or in the form of hardware. In some cases, the name of the module does not constitute a limitation on the module itself.

[0205] The readable storage medium provided by the embodiments of the present application is a computer readable storage medium, which stores computer readable program instructions (i.e. computer programs) for executing the above data analysis method, and can solve the technical problem of how to improve the efficiency of data analysis. Compared with the prior art, the computer readable storage medium provided by the present application has the same beneficial effects as the data analysis method provided by the above embodiments, and will not be described here.

[0206] The embodiment of the present application further provides a computer program product comprising a computer program, which, when executed by a processor, implements the steps of the data analysis method as described above.

[0207] The computer program product provided by the embodiment of the present application can solve the technical problem of how to improve the efficiency of data analysis. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as those of the data analysis method provided by the above-mentioned embodiments, and are not described here.

[0208] The above only describes some embodiments of the present application, and does not limit the patent scope of the present application. Any equivalent structural transformation made by using the content of the present application specification and drawings, or direct / indirect application in other related technical fields is included in the patent protection scope of the present application.

Claims

1. A data analysis method, characterized by, The data analysis method comprises: determining a target data set and a target structured query language (SQL) statement according to the obtained query information through a first large language model; executing the target SQL statement on the target data set to obtain a data query result; determining a target processing function according to the query information through a second large language model; processing the data query result according to the target processing function to obtain a data processing result; generating a scene analysis conclusion according to the data query result and the data processing result through a third large language model.

2. The data analysis method of claim 1, wherein, The step of determining a target data set and a target structured query language (SQL) statement according to the obtained query information comprises: determining the text similarity between the query information and the metadata of each data set, and determining the data set with a text similarity greater than a preset similarity threshold as a candidate data set; in the case where the number of candidate data sets does not exceed a preset number, determining the candidate data sets as target data sets; in the case where the number of candidate data sets exceeds a preset number, selecting the preset number of candidate data sets as target data sets; generating a target SQL statement according to the target data set.

3. The data analysis method of claim 2, wherein, The step of generating a target SQL statement according to the target data set comprises: obtaining a historical SQL query statement set corresponding to the target data set; determining a historical SQL query statement in the historical SQL query statement set that is most relevant to the query information as a reference SQL statement; rewriting the reference SQL statement according to the query information to obtain the target SQL statement.

4. The data analysis method as claimed in claim 1, characterized in that, After the step of obtaining a data query result, the method further comprises: determining whether the data query result meets a preset condition, wherein the preset condition comprises that the number of columns of the data query result is less than a preset column threshold, the data query result only contains one date column and one numerical value column, and the columns of the data query result cannot meet a preset calculation requirement; in the case where the number of columns of the data query result is less than a preset column threshold, performing full-amount processing on the data query result to obtain a data processing result; in the case where the data query result only contains one date column and one numerical value column, determining data with a numerical value less than a preset threshold in the numerical value column as target data, and aggregating the target data according to the date to obtain a data processing result; in the case where the columns of the data query result cannot meet a preset calculation requirement, displaying the data query result; in the case where the data query result does not meet the preset condition, executing the step of determining a target processing function according to the query information through a second large language model.

5. The data analysis method of claim 1, wherein, The target SQL statement contains a date range, and after the step of obtaining a data query result, the method further comprises: enlarging the date range in the target SQL statement to obtain an updated SQL statement, wherein the date range is a time limit for the query range of the target SQL statement. execute the updated SQL statement on the target data set to obtain a target query result; perform data mining on the target query result through a preset machine learning algorithm to obtain a data mining result; generate an index analysis conclusion through the third large language model based on the data mining result.

6. The data analysis method of claim 5, wherein, After the step of generating the scene analysis conclusion, the method further includes: supplement the query information based on the target data set through the third large language model to obtain a complete question; convert the data query result, the data processing result, and the data mining result into a form of a chart to obtain a plurality of data index analysis charts; dynamically assemble the complete question, the data index analysis charts, the scene analysis conclusion, and the index analysis conclusion based on a preset analysis report template to obtain a data analysis report.

7. The data analysis method of any one of claims 1 to 6, wherein, The data analysis method further includes: for any target data set, convert the target data set and the query information into task parameters through a preset master process, and send the task parameters to a preset process pool; determine whether there is an idle child process through the process pool; in the case where there is an idle child process, add the task parameters to a task queue of the process pool to wait for the idle child process to read and execute; determine a target SQL statement based on the task parameters and the first large language model through the idle child process, and execute the steps of executing the target SQL statement on the target data set to obtain a data query result and subsequent steps.

8. An electronic device, comprising: The device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the data analysis method according to any one of claims 1 to 7.

9. A storage medium, characterized by The storage medium is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the data analysis method according to any one of claims 1 to 7.

10. A computer program product, characterised in that, The computer program product includes a computer program, and the computer program is executed by the processor to implement the steps of the data analysis method according to any one of claims 1 to 7.

Images