A Big Data Metric Query and Analysis Method Based on Large Models

Abstract

The present invention provides a method for querying and analyzing big data metrics based on a large model. The method includes: accessing a data source; after accessing the data source, establishing data metrics based on business requirements; querying the data metrics through the large model in the form of a dialogue to obtain query results and analysis results; and visually displaying the query results and analysis results. The method for querying and analyzing big data metrics based on a large model according to the present invention accesses multiple data sources, establishes corresponding data metrics according to business requirements, introduces the form of AI dialogue, determines the query conditions for the data metrics according to the dialogue information of the dialogue participants and obtains the query results, and visually displays the query results and analysis results according to the personalized chart analysis form selected by the dialogue participants, improving the readability of the data and being more convenient.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and particularly relates to a big data index query and analysis method based on a large model. Background Art

[0002] Data indicators are specific numerical values used in data analysis and business management. They subdivide and quantify business units, making business goals describable, measurable, and decomposable to measure and evaluate the performance of business operations, performance, efficiency, etc. Data indicators are usually obtained by calculating, summarizing, or analyzing raw data. They can provide insights into the business situation and help make data-based decisions. Additionally, the data indicator system is an important part of building a data middle platform, which helps operation and product personnel intuitively see the changes in basic indicators and enables data analysts to carry out data analysis work more conveniently.

[0003] The invention patent with the application number: CN202311465846.8 discloses a financial big data storage and query method based on artificial intelligence. The method includes: obtaining different types of financial transaction data in different transaction data tables of an enterprise to construct long strings of different transaction data tables; adding check codes to each substring according to the data type of each substring; reorganizing the substrings in each long string according to the check codes and lengths of the substrings in each long string to obtain a reorganized string; processing the reorganized string to obtain the encoded data of each reorganized string, and obtaining the importance index of each type of encoded data according to the distribution of the encoded data with different lengths in the encoded data of each reorganized string; performing block processing on each reorganized string to obtain data blocks, and marking and storing all data blocks according to the check codes of the substrings corresponding to the data blocks. The above invention makes the efficiency of data storage and query higher.

[0004] However, the above-mentioned index data is presented to users in the form of data tables or strings, which is not intuitive for users to view, has poor readability, and when querying stored data, it is necessary to follow the set query path for corresponding queries, which is not convenient enough.

[0005] In view of this, there is an urgent need for a big data index query and analysis method based on a large model to at least solve the above deficiencies. Summary of the Invention

[0006] One of the objectives of the present invention is to provide a big data indicator query and analysis method based on a large model, which accesses multiple data sources and establishes corresponding data indicators according to business requirements; introduces the form of AI dialogue, determines the query conditions of data indicators according to the dialogue information of the dialogue personnel, and obtains the query results; visualizes the query results and analysis results according to the personalized chart analysis form selected by the dialogue personnel, improving the readability of the data and making it more convenient.

[0007] A big data indicator query and analysis method based on a large model provided by an embodiment of the present invention includes:

[0008] Access the data source;

[0009] After accessing the data source, establish data indicators based on business requirements;

[0010] Query data indicators through a large model based on the dialogue form, and obtain the query results and analysis results;

[0011] Visualize the query results and analysis results.

[0012] Preferably, accessing the data source includes:

[0013] Determine the required data source type;

[0014] Fill in the connection information according to the template of the connection information preset for the required data source type;

[0015] After completing the filling of the connection information, test the connection status until the connection is successful.

[0016] Preferably, accessing the data source further includes:

[0017] Obtain the target access network docking node;

[0018] Access the data source according to the target access network docking node;

[0019] Among them, obtaining the target access network docking node includes:

[0020] Obtain the first data evaluation information of the pre-access network docking node;

[0021] Determine multiple evaluation parties and evaluation values according to the first data evaluation information;

[0022] Obtain the first access data credibility value according to the evaluation weights and evaluation values of the evaluation parties;

[0023] Obtain the historical access response rate of the pre-access network docking node;

[0024] Normalize and sum the first access data credibility value and the historical access response rate to obtain the first target value;

[0025] Obtain the second data evaluation information of the current access network docking node;

[0026] Determine the second access data trust value according to the second data evaluation information;

[0027] Obtain the local access response rate of the current access network docking node;

[0028] Normalize and sum the second access data trust value and the local access response rate to obtain the second target value;

[0029] When the second target value is greater than or equal to each first target value, use the current access network docking node corresponding to the corresponding second target value as the first node to be reorganized;

[0030] When there is a second target value less than the first target value, obtain the target number of the second target value less than the first target value;

[0031] Use the second target value less than the first target value and the corresponding first target value together as the third target value;

[0032] Sort the third target values from largest to smallest, select the third target values of the first target number, and use them as the fourth target value;

[0033] Obtain the current access network docking node and / or the pre-access network docking node corresponding to the fourth target value, and use them as the second node to be reorganized;

[0034] Use the first node to be reorganized and the second node to be reorganized together as the target access network docking node.

[0035] Preferably, after accessing the data source, based on the business requirements, establish data indicators, including:

[0036] According to the business requirements, obtain the index establishment information, which includes: index name, index type, index source, and calculation logic;

[0037] Establish data indicators according to the index establishment information.

[0038] Preferably, based on the dialogue form, query the data indicators through the large model, and obtain the query results and analysis results, including:

[0039] Obtain the first dialogue information generated based on the dialogue form;

[0040] Based on the preset query condition extraction template, determine the first query condition according to the first dialogue information;

[0041] Query the data indicators through the large model according to the first query condition, and obtain the query results and analysis results.

[0042] Preferably, based on a preset query condition extraction template, determine a first query condition according to the first conversation information, including:

[0043] Based on a preset query condition extraction template, obtain the extraction result of the first query condition according to the first conversation information;

[0044] If the extraction result is successful extraction, determine the corresponding first query condition;

[0045] If the extraction result is failed extraction, determine the missing conversation information according to the query condition extraction template and the first conversation information;

[0046] Generate a supplementary conversation reminder message according to the missing conversation information;

[0047] Display the supplementary conversation reminder message to the conversation user;

[0048] Obtain the supplementary conversation information after the conversation user views the supplementary conversation reminder message;

[0049] Determine the first query condition according to the query condition extraction template, the first conversation information and the supplementary conversation information.

[0050] A big data index query and analysis method based on a large model provided by an embodiment of the present invention further includes:

[0051] Before querying data indexes according to the first query condition through a large model, determine whether a multi-round conversation is triggered;

[0052] If a multi-round conversation is triggered, obtain the second query condition of the second conversation information that has been triggered;

[0053] Query data indexes through the large model according to the first query condition and the second query condition, and obtain a query result and an analysis result.

[0054] Preferably, query data indexes through the large model according to the first query condition and the second query condition, and obtain a query result and an analysis result, including:

[0055] Obtain the verification result of the query result by the conversation user;

[0056] If the verification result is unsuccessful verification, display the third query condition to the conversation user, and obtain the corrected query condition after the conversation user corrects the third query condition;

[0057] Re-obtain the query result based on the corrected query condition.

[0058] Preferably, display the third query condition to the conversation user, and obtain the corrected query condition after the conversation user corrects the third query condition, including:

[0059] Parse the third query condition to determine at least one unit condition;

[0060] Obtain the deletion path of the unit condition and associate the deletion path with the preset operation position of the visual icon of the unit condition;

[0061] Obtain the condition constraint type of the unit condition, where the condition constraint type includes: necessary constraint and non-necessary constraint;

[0062] If the condition constraint type is a necessary constraint, when it is recognized that the number of unit conditions of any condition constraint type is 1 and the deletion path of the corresponding unit condition is triggered, the corresponding unit condition is used as the target unit condition;

[0063] Obtain the preset unit condition candidate list of the target unit condition; the unit condition candidate list includes multiple candidate unit conditions, and the candidate unit conditions are sorted from largest to smallest according to their relevance to the remaining unit conditions;

[0064] Display the unit condition candidate list to the conversation user and remind the conversation user to retain at least one unit condition with a necessary constraint type;

[0065] Determine the selection result of the conversation user and obtain the corrected query condition.

[0066] A big data metric query and analysis method based on a large model provided by an embodiment of the present invention further includes:

[0067] Perform label development according to the data metric to obtain the label development result; the label development result includes: original label, combined label, and external label.

[0068] A big data metric query and analysis method based on a large model provided by an embodiment of the present invention further includes:

[0069] Obtain the label task list according to the label development result;

[0070] Determine the execution result according to the list operation of the label task list;

[0071] Obtain the index link of the task instance display interface in the label task list;

[0072] Obtain the running status of the task instance according to the index link.

[0073] A big data metric query and analysis method based on a large model provided by an embodiment of the present invention further includes:

[0074] Perform single-object panoramic portrait and group analysis according to the label development result;

[0075] Share the label API according to the label development result;

[0076] When sharing the tag API, perform API call statistics;

[0077] When sharing the tag API, obtain the application list of the tag API, and the application list includes: application information and associated API information;

[0078] Parse the associated API information to obtain the associated API operations; among them, the operation types of the associated API operations include: generating interface documents and removing.

[0079] The beneficial effects of the present invention are:

[0080] The present invention accesses multiple data sources, and according to business requirements, establishes corresponding data metrics; introduces the form of AI dialogue, determines the query conditions of the data metrics based on the dialogue information of the dialogue participants and obtains the query results; visualizes the query results and analysis results according to the personalized chart analysis form selected by the dialogue participants, improving the readability of the data and being more convenient.

[0081] Other features and advantages of the present invention will be described in the following specification, and part of them will be obvious from the specification or understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in this application document.

[0082] The technical solution of the present invention will be further described in detail below through the drawings and embodiments. Description of the Drawings

[0083] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0084] Figure 1 It is a schematic diagram of a big data metric query and analysis method based on a large model in an embodiment of the present invention. Detailed Embodiments

[0085] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0086] An embodiment of the present invention provides a big data metric query and analysis method based on a large model, as Figure 1 shown, including:

[0087] Step 1: Access the data source; among them, the data source is: a network docking node that provides query data;

[0088] Step 2: After accessing the data source, based on the business requirements, establish data metrics; among them, the business requirements are: index query requirements, such as: elderly information management query; the data metrics are: data related to the business requirements, such as: the number of elderly people, age, and home address in each region, etc.;

[0089] Step 3: Query the data metrics through the large model in the form of a dialogue to obtain the query result and the analysis result; among them, the form of the dialogue means the form of having a dialogue with the AI; the large model is: an AI model that retrieves data that meets the query conditions from the database according to the query conditions, and is obtained by machine learning using the acquisition records of obtaining data by manually inputting query conditions; the query result is: the result obtained by parsing the dialogue input by the dialogue person, determining the query conditions for querying the data metrics and performing corresponding queries; the analysis result is: the result obtained by analyzing the query result according to the chart analysis form selected by the dialogue person;

[0090] Step 4: Visualize the query result and the analysis result.

[0091] The working principle and beneficial effects of the above technical solution are:

[0092] The present invention accesses multiple data sources, and based on the business requirements, establishes corresponding data metrics; introduces the form of an AI dialogue, determines the query conditions for the data metrics according to the dialogue information of the dialogue person, and obtains the query result; visualizes the query result and the analysis result according to the result selected by the dialogue person's personalized chart analysis form, improving the readability of the data and being more convenient.

[0093] In one embodiment, accessing the data source includes:

[0094] Determine the required data source type; among them, the required data source type is: the types of data sources to be accessed, such as: relational database, NoSQL database;

[0095] Fill in the connection information according to the connection information filling template preset for the required data source type; among them, the connection information filling template is: the information framework of the information required to be filled in when accessing the data source of the required data source type, such as: database address, port, username, password, etc.;

[0096] After completing the filling of the connection information, test the connection status until the connection is successful.

[0097] The working principle and beneficial effects of the above technical solution are:

[0098] To access the data source, first log in to the system and enter the data source access page. Select the required data source type (such as relational database, NoSQL database, etc.), and fill in the relevant connection information, such as database address, port, username, password, etc. After filling in, test the connection to ensure the connection is successful. After the test connection is successful, confirm the access.

[0099] The present invention introduces a connection information filling template to guide the filling of connection information, improving the standardization and success rate of data source connection.

[0100] In one embodiment, accessing the data source further includes:

[0101] Obtain a target access network docking node; wherein, the target access network docking node is: a data source access network node allowed to be accessed;

[0102] Access the data source according to the target access network docking node;

[0103] Wherein, obtaining the target access network docking node includes:

[0104] Obtain the first data evaluation information of the pre-access network docking node; wherein, the pre-access network docking node is: all data source access network nodes obtained by big data; the first data evaluation information is: the evaluation value of the data shared by the data sharing party corresponding to the pre-access network docking node on the data sharing platform;

[0105] Determine multiple evaluators and evaluation values according to the first data evaluation information;

[0106] Obtain the first access data credibility value according to the evaluation weight and evaluation value of the evaluator; wherein, the evaluation weight is determined according to the historical evaluation credit of the evaluator, the lower the historical evaluation credit, the smaller the weight is given, and the sum of the evaluation weights of the evaluators corresponding to the same pre-access network docking node is 1; the first access data credibility value is the sum of the products of the evaluation weight and the evaluation value;

[0107] Obtain the historical access response rate of the pre-access network docking node; wherein, the historical access response rate is: the response speed of data sharing in the history of the pre-access network docking node;

[0108] Normalize and sum the first access data credibility value and the historical access response rate to obtain the first target value;

[0109] Obtain the second data evaluation information of the current access network docking node; wherein, the second data evaluation information is: the evaluation information of the current access network docking node by the local information management personnel;

[0110] Determine the second access data credibility value according to the second data evaluation information; wherein, the second access data credibility value is: the evaluation value of the current access network docking node by the local information management personnel;

[0111] Obtain the local access response rate of the current access network docking node; wherein, the local access response rate is: the response speed of the current access network docking node in local historical data sharing;

[0112] Normalize and sum the second access data credibility value and the local access response rate to obtain the second target value;

[0113] When the second target value is greater than or equal to each first target value, use the current access network docking node corresponding to the corresponding second target value as the first node to be reorganized;

[0114] When there is a second target value less than the first target value, obtain the target number of the second target value less than the first target value; for example: there is a second target value of 85 less than the first target value of 90, there is a second target value of 75 less than the first target values of 90 and 80;

[0115] Use the second target value less than the first target value and the corresponding first target value together as the third target value; wherein, the third target value is, for example: 90, 85, 80, 75;

[0116] Sort the third target values from largest to smallest, select the first target number of the third target values, and use them as the fourth target value; wherein, the third target value is, for example: 90, 85;

[0117] Obtain the current access network docking node and / or the pre-access network docking node corresponding to the fourth target value, and use them as the second node to be reorganized;

[0118] Use the first node to be reorganized and the second node to be reorganized together as the target access network docking node.

[0119] The working principle and beneficial effects of the above technical solution are:

[0120] The present invention screens the target access network docking nodes of the access data source. Specifically, obtain the first data evaluation information of the pre-access network docking nodes obtained by big data, calculate the first access data credibility value according to the evaluation values and evaluation weights of different evaluation parties; obtain the historical access response rate of the pre-access network docking nodes, normalize and sum the first access data credibility value and the historical access response rate to obtain the first target value; obtain the evaluation information of the local information management personnel on the current access network docking node, determine the second access data credibility value, and in addition, obtain the local access response rate of the current access network docking node, normalize and sum the second access data credibility value and the local access response rate to obtain the second target value.

[0121] Compare the second target value with the first target value. If the second target value is greater than or equal to each first target value, the current access network docking node corresponding to the second target value does not need to be removed and is directly used as the first node to be reorganized; if there is a second target value less than the first target value, the current access network docking node corresponding to the second target value has a risk of being removed. Obtain the target number of second target values less than the first target value. At the same time, use the second target value less than the first target value and the corresponding first target value together as the third target value; sort the third target values from largest to smallest, and select the current access network docking nodes and / or pre-access network docking nodes corresponding to the first several fourth target values as the second nodes to be reorganized;

[0122] Re-integrate the first node to be reorganized and the second node to be reorganized as the target access network docking node, dynamically determine a reliable network docking node in real time and obtain data from it, which is more suitable.

[0123] In one embodiment, after accessing the data source, based on business requirements, establish data indicators, including:

[0124] According to business requirements, obtain indicator establishment information, which includes: indicator name, indicator type, indicator source, and calculation logic;

[0125] Establish data indicators according to the indicator establishment information.

[0126] The working principle and beneficial effects of the above technical solution are:

[0127] According to business requirements, the present invention customizes the indicator name, indicator type (such as counting, summing, averaging, etc.), indicator source (i.e., the already accessed data source), calculation logic, etc., and the formulation of data indicators is more flexible.

[0128] In one embodiment, based on a dialogue form, query data indicators through a large model, and obtain query results and analysis results, including:

[0129] Obtain the first dialogue information generated based on the dialogue form; wherein, the first dialogue information is: the dialogue semantics generated by the dialogue user in real time on the AI dialogue interface;

[0130] Based on a preset query condition extraction template, determine the first query condition according to the first dialogue information; wherein, the preset query condition extraction template is formulated according to the historical records of manually formulating query conditions according to query requirements; the first query condition is: the data indicator query condition obtained by comparing the first dialogue information with the query condition extraction template;

[0131] Query data indicators through the large model according to the first query condition, and obtain query results and analysis results.

[0132] The working principle and beneficial effects of the above technical solution are:

[0133] The present invention extracts a template based on the first dialogue information extracted from the AI ​​dialogue and the introduced query condition, determines the query condition of the data indicator, and improves the efficiency of determining the first query condition.

[0134] In one embodiment, extracting a template based on a preset query condition and determining a first query condition according to the first conversation information includes:

[0135] Based on the preset query condition extraction template, obtaining the extraction result of the first query condition according to the first conversation information;

[0136] If the extraction result is successful, determine the corresponding first query condition; if the description factor (e.g., indicator name, application subject) in the first dialogue information matches the condition determination factor (e.g., indicator name, application subject, and query time) corresponding to any pre-selected query condition in the query condition extraction template, then use the corresponding pre-selected query condition as the first query condition;

[0137] If the extraction result is extraction failure, determine the missing dialogue information according to the query condition extraction template and the first dialogue information; wherein the missing dialogue information is: when the description factor in the first dialogue information has the highest matching degree with the corresponding condition determination factor in the query condition extraction template, the description information of the condition determination factor missing from the determined description factors;

[0138] Generate supplementary dialogue reminder information based on missing dialogue information;

[0139] Displaying supplementary conversation reminder information to conversation users;

[0140] Obtain the supplementary conversation information after the conversation user views the supplementary conversation reminder information;

[0141] The template, the first dialogue information and the supplementary dialogue information are extracted according to the query condition to determine the first query condition.

[0142] The working principle and beneficial effects of the above technical solution are:

[0143] When extracting the first query condition in the first dialogue information using the query condition extraction template, there is a situation where the description factors in the first dialogue information do not meet the condition determination factors of any one of the first query condition determination scenarios. Therefore, before determining the first query condition, obtain the extraction result of the first query condition; if the extraction is successful, directly extract it; if the extraction is unsuccessful, obtain the missing condition determination factors that do not match when the description factors in the first dialogue information have the highest matching degree with the corresponding condition determination factors in the query condition extraction template. Based on the missing condition determination factors, obtain the missing dialogue information; based on the missing dialogue information, generate a supplementary dialogue reminder message and remind the dialogue user to supplement the dialogue information; after supplementation, then extract the first query condition based on the query condition extraction template, which improves the rationality of the first query condition determination process.

[0144] In one embodiment, before querying data metrics by the large model according to the first query condition, it further includes:

[0145] Determine whether to trigger a multi-round dialogue; among them, when determining whether to trigger a multi-round dialogue, if the user clicks the preset "Continue to Ask" button on the dialogue interface, enter the multi-round dialogue state and automatically determine the query condition based on the context content;

[0146] If a multi-round dialogue is triggered, obtain the second query condition of the second dialogue information that has been triggered; among them, the second query condition is: the data metric query condition that has been determined above after the multi-round dialogue is triggered;

[0147] According to the first query condition and the second query condition, query the data metric by the large model to obtain the query result and the analysis result.

[0148] The working principle and beneficial effects of the above technical solution are:

[0149] The present invention introduces a question-and-answer form of multi-round dialogue. According to the first query condition and the second query condition, query the data metric by the large model to obtain the query result and the analysis result. Through multi-round dialogue, the system can more accurately capture and understand the user's query intention and needs, so as to provide more accurate query results.

[0150] In one embodiment, querying the data metric by the large model according to the first query condition and the second query condition to obtain the query result and the analysis result includes:

[0151] Obtain the verification result of the query result by the dialogue user; among them, the verification result is: the result of verifying whether the query result meets the query intention of the dialogue user. When verifying, simply pop up a window on the dialogue interface to ask the dialogue user and obtain the feedback information of the dialogue user;

[0152] If the verification result is unsuccessful, display the third query condition to the conversation user and obtain the corrected query condition after the conversation user corrects the third query condition; wherein, the third query condition is: the first query condition and the second query condition that fail the verification; when obtaining the corrected query condition, the conversation user can add or delete the third query condition independently to obtain the corrected query condition.

[0153] Re-obtain the query result based on the corrected query condition.

[0154] The working principle and beneficial effects of the above technical solution are as follows:

[0155] The present invention introduces a verification program for the query result by the conversation user. When the verification is unsuccessful, display the third query condition to the conversation user and obtain the corrected query condition after the conversation user modifies it; re-obtain the query result based on the corrected query condition, which improves the accuracy of the query.

[0156] In one embodiment, displaying the third query condition to the conversation user and obtaining the corrected query condition after the conversation user corrects the third query condition includes:

[0157] Analyze the third query condition to determine at least one unit condition; wherein, the unit condition is: the basic unit that constitutes the third query condition.

[0158] Obtain the deletion path of the unit condition and associate the deletion path with the preset operation position of the visual icon of the unit condition; wherein, the deletion path is: the program in the system that executes the deletion of the unit condition; the visual icon is: the visual element of the unit condition; the preset operation position is: the position preset for the visual icon to indicate the deletion operation, such as: the "x" symbol in the upper right corner of the icon.

[0159] Obtain the condition constraint type of the unit condition. The condition constraint type includes: necessary constraint and non-necessary constraint; wherein, the necessary constraint is: the condition that must be satisfied for executing the data index query; the non-necessary constraint is: the condition that does not have to be satisfied for executing the data index query; for example: when counting the number of births in the past 10 years, the necessary constraints are: time, date of birth, etc., and the non-necessary constraint is: gender, etc.

[0160] If the condition constraint type is a necessary constraint, when it is recognized that the number of unit conditions of any condition constraint type is 1 and the deletion path of the corresponding unit condition is triggered, the corresponding unit condition is used as the target unit condition.

[0161] Obtain a list of candidate unit conditions preset for the target unit condition; the list of candidate unit conditions includes multiple candidate unit conditions, and the candidate unit conditions are sorted from largest to smallest according to their relevance to the remaining unit conditions. Among them, the condition description content of the candidate unit condition is of the same type as that of the target unit condition. For example, when the target unit condition is 2022, the candidate unit conditions can be: 2023, 2021, etc.; the remaining unit conditions are: the remaining conditions in the unit conditions except the target unit condition; the relevance is: the quantitative value of the degree of association;

[0162] Display the list of candidate unit conditions to the dialogue user and remind the dialogue user to retain at least one unit condition whose condition constraint type is a necessary constraint;

[0163] Determine the selection result of the dialogue user and obtain the corrected query condition. Among them, the selection result is: the cancel deletion operation or candidate unit condition selected by the dialogue user.

[0164] The working principle and beneficial effects of the above technical solution are:

[0165] The present invention splits the third query condition to obtain unit conditions; introduces the operation position of the deletion path of the unit condition. When the dialogue user does not need the corresponding unit condition, the corresponding unit condition can be directly deleted based on the query logic, which is more convenient; in addition, the condition constraint type of the unit condition is introduced. When the condition constraint type is a necessary constraint, when it is recognized that the number of unit conditions of any condition constraint type is 1 and the deletion path of the corresponding unit condition is triggered, the corresponding unit condition is used as the target unit condition, and a list of candidate unit conditions for the target unit condition is obtained; the list of candidate unit conditions is displayed to the dialogue user, and the candidate unit conditions in the list of candidate unit conditions are sorted from largest to smallest according to their relevance to the remaining unit conditions. The more likely candidate unit conditions to be selected are presented first, restricting the dialogue user to retain at least one unit condition whose condition constraint type is a necessary constraint, and obtaining the corrected query condition, which improves the standardization of query condition correction.

[0166] In one embodiment, a big data metric query and analysis method based on a large model further includes:

[0167] Perform label development according to the data metrics to obtain label development results; the label development results include: original labels, combined labels, and external labels. Among them, label development is: creating labels for classifying and identifying user groups or data points according to data metrics and user behavior; original labels are calculated and judged based on different data sources; combined labels are calculated and combined based on different labels; external labels are defined based on external label data.

[0168] The working principle and beneficial effects of the above technical solution are:

[0169] The present invention develops labels for data indicators, provides more detailed screening conditions, and has higher reusability.

[0170] In one embodiment, a big data indicator query and analysis method based on a large model further includes:

[0171] According to the label development result, obtain a label task list; wherein, the label task list is a list storing the online status of labels;

[0172] According to the list operation of the label task list, determine the execution result; wherein, the list operation is an operation to control the online and offline status of labels in the label task list;

[0173] Obtain the index link of the task instance display interface in the label task list; wherein, the index link is a preset viewing link for the task instance display interface;

[0174] According to the index link, obtain the running status of the task instance. The running status of the task instance includes: running duration, creation time, update time, etc.

[0175] The working principle and beneficial effects of the above technical solution are as follows:

[0176] This application obtains a label task list according to the label development result; introduces the list operation of the label task list to determine the execution result, reasonably sets the online and offline cycle, reduces the occupation of system resources; constructs a viewing index for the running status of the task instance, and users can directly click on the index link to view the running status of the task instance, which is more convenient.

[0177] In one embodiment, a big data indicator query and analysis method based on a large model further includes:

[0178] According to the label development result, perform single-object panoramic portrait and group analysis; wherein, the single-object panoramic portrait refers to a comprehensive analysis of a single object (such as a single user or a single product), including all relevant data and labels of the object, forming a 360-degree complete view; group analysis is: statistical and analysis of a group of objects (such as a user group or a product set) to identify the characteristics, behavior patterns or trends of the group;

[0179] According to the label development result, share the label API; wherein, the label API is an application programming interface that provides label data or the ability to generate labels to other systems or services in the form of an API;

[0180] When sharing the tag API, API call statistics are performed; among them, the API call statistics are the process of recording and analyzing the usage of the tag API, including information such as the number of calls, call time, call source, etc.;

[0181] When sharing the tag API, obtain the application list of the tag API. The application list includes: application information and associated API information; among them, the application information includes: application ID, application name, etc.; the associated API information includes: API name, status, creator, authorized validity period, and associated API operations, etc.;

[0182] Parse the associated API information to obtain the associated API operations; among them, the operation types of the associated API operations include: generating interface documents and removing.

[0183] The working principle and beneficial effects of the above technical solution are as follows:

[0184] Based on the tag development results, the present invention performs single-object panoramic portrait and group analysis, and generates a shared tag API for external services; in addition, the application list of the tag API is introduced to obtain application information, and the detailed information of the API associated with the application is obtained according to the application information, and its operations directly generate interface documents or remove them, improving the data sharing efficiency and call convenience.

[0185] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A big data index query and analysis method based on a big model, characterized in that: include: Access data sources; After accessing the data source, establish data indicators based on business needs; Acquire first dialogue information generated based on the dialogue form; Extracting a template based on a preset query condition, and determining a first query condition according to the first conversation information; Before querying the data indicator according to the first query condition through the big model, determine whether to trigger multiple rounds of dialogue; If multiple rounds of dialogue are triggered, a second query condition for obtaining the second dialogue information that has been triggered; According to the first query condition and the second query condition, query the data index through the big model to obtain the query result and the analysis result; Obtain the verification result of the query result by the dialog user; If the verification result is that the verification is unsuccessful, parsing the third query condition to determine at least one unit condition; Obtaining a deletion path of the unit condition, and associating the deletion path with an operation position preset by a visual icon of the unit condition; Get the condition constraint type of the unit condition, which includes necessary constraint and non-necessary constraint; If the conditional constraint type is a necessary constraint, when the number of unit conditions of any conditional constraint type is identified to be 1 and the deletion path of the corresponding unit condition is triggered, the corresponding unit condition is used as the target unit condition; Obtaining a unit condition candidate list preset for the target unit condition; the unit condition candidate list includes multiple candidate unit conditions, and the candidate unit conditions are sorted from large to small according to the correlation with the remaining unit conditions; Displaying the candidate list of unit conditions to the dialog user, and reminding the dialog user to keep at least one unit condition whose condition constraint type is a necessary constraint; Determine the selection result of the dialog user and obtain the modified query condition; Re-obtain query results based on the revised query conditions; Visualize the query and analysis results.

2. A big data index query and analysis method based on a big model as claimed in claim 1, characterized in that: Access data sources, including: Determine the type of data source required; Fill in the connection information according to the preset connection information template of the required data source type; After completing the connection information, test the connection status until the connection is successful.

3. A big data index query and analysis method based on a big model as claimed in claim 1, characterized in that: After accessing the data source, establish data indicators based on business needs, including: According to business needs, obtain indicator establishment information, which includes: indicator name, indicator type, indicator source and calculation logic; Establish information based on indicators and establish data indicators.

4. A method for querying and analyzing big data indicators based on a big model as claimed in claim 1, characterized in that: Extracting a template based on a preset query condition and determining a first query condition according to the first conversation information includes: Based on the preset query condition extraction template, obtaining the extraction result of the first query condition according to the first conversation information; If the extraction result is successful, determine the corresponding first query condition; If the extraction result is extraction failure, extract the template and the first dialogue information according to the query condition to determine the missing dialogue information; Generate supplementary dialogue reminder information based on missing dialogue information; Displaying supplementary conversation reminder information to conversation users; Obtain the supplementary conversation information after the conversation user views the supplementary conversation reminder information; The template, the first dialogue information and the supplementary dialogue information are extracted according to the query condition to determine the first query condition.

5. A method for querying and analyzing big data indicators based on a big model as claimed in claim 3, characterized in that: Also includes: Develop labels based on data indicators and obtain label development results; The label development results include: original labels, combined labels and external labels.

6. A method for querying and analyzing big data indicators based on a big model as claimed in claim 5, characterized in that: Also includes: According to the label development results, obtain the label task list; Determine the execution result according to the list operation of the tag task list; Get the index link of the task instance display interface in the tag task list; Get the task instance running status according to the index link.

7. A method for querying and analyzing big data indicators based on a big model as claimed in claim 5, characterized in that: Also includes: Based on the label development results, single-object panoramic profiling and group analysis are performed; Based on the tag development results, share the tag API; When sharing tag API, collect API call statistics; When sharing a tag API, obtain the tag API application list, which includes: application information and associated API information; Parse the associated API information and obtain the associated API operations; wherein the operation types of the associated API operations include: generating an interface document and removing.

Images