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

Abstract

The invention discloses a data analysis method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of data processing.The method comprises the steps that in response to an analysis instruction for an analysis object, a first analysis target and target configuration information of the analysis object are obtained, the target configuration information is used for representing a corresponding relation between at least one second analysis target and a test case, at least one problem recognition mode is packaged in the test case, and the problem recognition mode comprises one or more of image-text recognition and large language model recognition; querying a target test case matched with the first analysis target based on the target configuration information; and calling the target test case to perform problem analysis on the analysis object to obtain a problem analysis result of the analysis object. According to the invention, the analysis problem of the data report problem can be solved.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, specifically to data analysis methods, apparatus, electronic devices, storage media, and program products. Background Technology

[0002] Data reports, as a common and widely used form of data product, are characterized by diverse data formats and ease of modification. During the daily development and iteration of data reports, problems frequently arise due to inconsistent data flow changes, unreasonable manual configurations, and instability of the reporting platform. Therefore, how to quickly analyze the problems existing in data reports has become an urgent issue to be addressed. Summary of the Invention

[0003] In view of this, the present disclosure provides a data analysis method, apparatus, electronic device, storage medium, and program product to solve the analysis problem of data reporting.

[0004] In a first aspect, this disclosure provides a data analysis method, the method comprising:

[0005] In response to an analysis command for an analysis object, the system obtains a first analysis target and target configuration information for the analysis object. The target configuration information is used to characterize the correspondence between at least one second analysis target and a test case. The test case encapsulates at least one problem identification method, which includes one or more of image and text recognition and large language model recognition.

[0006] Based on the target configuration information, query the target test cases that match the first analysis target;

[0007] The target test case is invoked to perform problem analysis on the analysis object, and the problem analysis results of the analysis object are obtained.

[0008] Secondly, this disclosure provides a data analysis apparatus, the apparatus comprising:

[0009] The response module is used to respond to the analysis command for the analysis object, obtain the first analysis target and target configuration information of the analysis object, the target configuration information is used to characterize the correspondence between at least one second analysis target and test cases, and the test cases encapsulate at least one problem identification method, the problem identification method including one or more of image and text recognition and large language model recognition;

[0010] The query module is used to query target test cases that match the first analysis target based on the target configuration information;

[0011] The analysis module is used to call the target test case to perform problem analysis on the analysis object and obtain the problem analysis results of the analysis object.

[0012] Thirdly, this disclosure provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the data analysis method described in the first aspect or any corresponding embodiment.

[0013] Fourthly, this disclosure provides a computer-readable storage medium storing computer instructions for causing a computer to perform the data analysis method described in the first aspect or any corresponding embodiment thereof.

[0014] Fifthly, this disclosure provides a computer program product, including computer instructions for causing a computer to execute the data analysis method described in the first aspect or any corresponding embodiment thereof.

[0015] The data analysis method provided in this disclosure pre-configures test cases for at least one second analysis target. Each test case encapsulates at least one problem identification method from image / text recognition and large language model recognition to adapt to different analysis targets. Furthermore, when data reports need to be analyzed, a first analysis target for the analysis object can be obtained in response to an analysis command for that object. Using the first analysis target and target configuration information, target test cases matching the first analysis target are quickly found. The pre-configured target test cases, containing a problem identification method adapted to the first analysis target, are then invoked to automatically analyze the problem in the analysis object, obtaining the problem analysis results without requiring separate test cases for each data report's analysis object. Therefore, it can quickly analyze problems existing in data reports.

[0016] The beneficial effects of data analysis devices, electronic devices, storage media, and program products correspond to the beneficial effects of data analysis methods, and will not be elaborated upon here. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the specific embodiments of this disclosure or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of an optional application scenario according to an embodiment of the present disclosure;

[0019] Figure 2 This is a flowchart illustrating a first data analysis method according to an embodiment of the present disclosure;

[0020] Figure 3 This is a schematic diagram illustrating the overall execution process of a data analysis method according to an embodiment of the present disclosure;

[0021] Figure 4 This is a flowchart illustrating a second data analysis method according to an embodiment of the present disclosure;

[0022] Figure 5 This is a schematic diagram illustrating a query failure problem according to an embodiment of the present disclosure;

[0023] Figure 6 This is a schematic diagram illustrating a query timeout problem according to an embodiment of the present disclosure;

[0024] Figure 7 This is a flowchart illustrating a third data analysis method according to an embodiment of the present disclosure;

[0025] Figure 8 This is a flowchart illustrating a manual inspection process;

[0026] Figure 9 This is a schematic diagram of a chart showing significant data fluctuations within a recent day, according to an embodiment of the present disclosure.

[0027] Figure 10 This is a schematic diagram of a stagnant chart data according to an embodiment of the present disclosure;

[0028] Figure 11 This is a schematic diagram of a large language model recognition process according to an embodiment of the present disclosure;

[0029] Figure 12 This is a schematic diagram of a first prompt word according to an embodiment of the present disclosure;

[0030] Figure 13 This is a schematic diagram illustrating a problem analysis result according to an embodiment of the present disclosure;

[0031] Figure 14 This is a schematic diagram of a message card according to an embodiment of the present disclosure;

[0032] Figure 15 This is a schematic diagram of a problem analysis report according to an embodiment of the present disclosure;

[0033] Figure 16 This is a structural block diagram of a data analysis apparatus according to an embodiment of the present disclosure;

[0034] Figure 17 This is a structural block diagram of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0036] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0037] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0038] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0039] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0040] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0041] Data reports, as a common and widely used form of data product, are characterized by diverse data formats and ease of modification. During the daily development and iteration of data reports, issues frequently arise due to inconsistent data flow, unreasonable manual configuration, and platform instability. These issues include unusable report data (stability problems), unreliable report data (accuracy problems), and inconsistent report data (consistency problems). Therefore, quickly analyzing the problems in data reports has become an urgent issue to address.

[0042] In related technologies, the following two problem analysis methods are proposed:

[0043] The first approach, based on a user interface (UI) automation framework, simulates user login and click behaviors. It verifies whether the report page loads as expected by checking the presence of elements and text markers in specific locations. However, this problem analysis method requires simulating user behavior to verify the business functionality of the report page. For data reporting products, user behavior involves observing the entire report, analyzing key metrics, and drawing conclusions about business issues. This user behavior is difficult to simulate, leading to a mismatch in the problem analysis scenario. Furthermore, since data reporting is a data-driven product, this first problem identification method focuses more on the problem characteristics of the report while ignoring the data itself, making it difficult to accurately identify accuracy and consistency issues. Moreover, in data reporting products, a single report page often contains hundreds of charts and metrics. The first problem identification method struggles to quickly generate automated test cases for all charts, failing to significantly improve testing efficiency in practice and instead greatly increasing the maintenance costs of automated test cases.

[0044] The second approach involves building automated problem analysis capabilities at the data layer. However, this method relies on the stability monitoring of the reporting product platform and regular checks by data reporting personnel. The stability testing accuracy of the reporting product platform is low, making it difficult to detect query failures for specific metrics. Manual checks are inefficient and infrequent, easily overlooking problems. Furthermore, this method requires adding accuracy detection functionality to the data reporting chain (such as the data reporting interface layer), but data-level accuracy detection struggles to effectively cover chart-level accuracy issues; for example, problems caused by improper data report configuration may go undetected. Additionally, existing volatility derivation models for curve fluctuations cannot fully cover visually identifiable anomalies, making it difficult to detect such anomalies through accuracy checks. Moreover, this method requires building an indicator marketplace to clearly define how the same indicator is used in different scenarios and configure relevant detection functions based on these usage patterns. Maintaining an indicator marketplace is costly; without maintenance, the entire detection functionality becomes worthless. Furthermore, inconsistencies in historical indicators caused by backtracking are difficult to detect using an indicator marketplace.

[0045] In view of this, according to embodiments of this disclosure, a data analysis method is provided. This data analysis method aims to combine image recognition and Large Language Model (LLM) technology to build an automated report inspection solution to automatically identify stability, accuracy, and consistency issues in reports. It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be executed in a different order than that presented here.

[0046] As one optional application scenario of this disclosure embodiment, such as Figure 1 As shown, the entire data analysis scenario includes a client 10 and a server 20 of the testing platform. In the client 10, the user can configure the analysis object of the data report to be analyzed and the first analysis target of the analysis object. The server 20 is configured with at least one test case corresponding to a second analysis target. Each test case encapsulates a problem identification method, such as image / text recognition and large language model recognition. The server 20 can respond to analysis commands for the analysis object, calling the target test case matching the first analysis target to perform problem analysis on the analysis object and obtain the problem analysis results.

[0047] This embodiment provides a first data analysis method, which can be used in a data report testing platform, such as the server side of the testing platform. Figure 2 This is a flowchart illustrating the first data analysis method according to an embodiment of the present disclosure, as shown below. Figure 2 As shown, the process includes the following steps:

[0048] Step S201: In response to the analysis command for the analysis object, obtain the first analysis target and target configuration information of the analysis object. The target configuration information is used to characterize the correspondence between at least one second analysis target and test cases. The test cases encapsulate at least one problem identification method, which includes one or more of image and text recognition and large language model recognition.

[0049] In practical applications, a user-based automated testing framework can be built upon open-source functional automation testing frameworks (such as Robot Framework). The client of the testing platform displays a configuration page on top of this user-friendly automated testing framework. Users can configure the analysis objects to be analyzed in the target data report, the primary analysis target of each object, and the inspection plan for the analysis objects, generating task configuration instructions. The server of the testing platform responds to these task configuration instructions, generating inspection tasks and their execution conditions. Then, if the execution conditions of the inspection tasks are met, the server generates analysis instructions for the analysis objects.

[0050] In addition, users can also configure the analysis object to be analyzed and the first analysis target of the analysis object through the configuration page displayed on the client, generate analysis instructions for the analysis object, and send them to the server.

[0051] Optionally, the first analysis objective is used to characterize the problem to be analyzed for the analysis object. The first analysis objective includes one or more of the following: stability issues, accuracy issues, and consistency issues. Stability issues include one or more of the following: query failure issues and query timeout issues. Accuracy issues include one or more of the following: numerical values ​​dropping to zero, null values, and outliers. Consistency issues include one or more of the following: cross-page consistency issues and historical page consistency issues. The above issues can be adjusted according to the actual situation and are not limited here.

[0052] Optionally, the second analysis objective is used to characterize the problem to be analyzed. The second analysis objective includes one or more of the following: stability issues, accuracy issues, and consistency issues. Stability issues include one or more of the following: query failure issues and query timeout issues. Accuracy issues include one or more of the following: numerical values ​​dropping to zero, null values, and outliers. Consistency issues include one or more of the following: cross-page consistency issues and historical page consistency issues. The above issues can be adjusted according to the actual situation and are not limited here.

[0053] like Figure 3As shown, in practical applications, testers can conduct surveys with data report developers and users to collect information on the scope and definitions of common problems in data reports. These include stability issues, accuracy issues, and consistency issues. Stability issues include query failures and query timeouts. Accuracy issues include values ​​dropping to zero, null values, and outliers. Consistency issues include cross-page consistency and consistency across historical pages. Furthermore, the characteristics of each type of problem are extracted. Based on these characteristics, and combined with the problem identification capabilities provided by image and text recognition and large language model recognition, modular and automated test case development is completed, resulting in test cases for each type of problem. This leads to the correspondence between each type of problem and its test cases, i.e., the aforementioned target configuration information. The second analytical objective mentioned above is... Figure 3 The various issues shown demonstrate how to quickly configure test cases for different data report analysis objects using target configuration information.

[0054] Step S202: Based on the target configuration information, query the target test cases that match the first analysis target.

[0055] The second analysis objective includes the first analysis objective. The test cases corresponding to the second analysis objective that matches the first analysis objective are queried from the target configuration information to determine the target test cases.

[0056] Step S203: Call the target test case to perform problem analysis on the analysis object and obtain the problem analysis results of the analysis object.

[0057] Specifically, such as Figure 3 As shown, the image-text recognition is configured with corresponding question text features for the second analysis target. If the target test case encapsulates image-text recognition, it is called to capture an image of the page containing the analysis object, and image-text recognition is performed on the image of the analysis object to obtain the target text features of the analysis object. Then, based on the target text features and question text features, question analysis is performed on the analysis object to obtain the question analysis results.

[0058] Specifically, such as Figure 3 As shown, the large language model is configured with corresponding prompt words for the second analysis target. These prompt words instruct the large language model to perform problem analysis on the image of the analysis object. Furthermore, if the target test case encapsulates large language model recognition, the target test case is invoked to capture an image of the page containing the analysis object. Using the large language model and the prompt words, problems are identified in the image of the analysis object, yielding the problem analysis result. Here, the large language model is a visual large language model.

[0059] After obtaining the problem analysis results from the analyzed object, the server feeds back the problem analysis results to the target object, such as testers, developers, and users.

[0060] The data analysis method provided in this embodiment pre-configures test cases for at least one second analysis target. These test cases encapsulate at least one problem identification method from image / text recognition and large language model recognition to adapt to different analysis targets. Furthermore, when data reports need to be analyzed, the method can respond to analysis commands for the analysis object and obtain the first analysis target of the analysis object. Using the first analysis target and target configuration information, it quickly finds target test cases that match the first analysis target, calls the pre-configured target test cases that match the first analysis target, and automatically performs problem analysis on the analysis object to obtain the problem analysis results, without needing to configure separate test cases for each data report's analysis object. Therefore, it can quickly analyze problems existing in data reports.

[0061] This embodiment provides a second data analysis method, which can be used in data report testing platforms, such as the server side of a testing platform. Figure 4 This is a flowchart illustrating the second data analysis method according to an embodiment of the present disclosure, as shown below. Figure 4 As shown, the process includes the following steps:

[0062] Step S401: In response to the analysis command for the analysis object, obtain the first analysis target and target configuration information of the analysis object. The target configuration information is used to characterize the correspondence between at least one second analysis target and test cases. The test cases encapsulate at least one problem identification method, which includes one or more of image and text recognition and large language model recognition. For details, please refer to... Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0063] Step S402: Based on the target configuration information, query the target test cases that match the first analysis target. For details, please refer to [link to relevant documentation]. Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0064] Step S403: Call the target test case to perform problem analysis on the analysis object and obtain the problem analysis results of the analysis object.

[0065] In some optional implementations, step S403 above includes:

[0066] Step S4031: If the target test case includes image and text recognition, call the target test case to extract the first image of the object being analyzed from the page where the object is located.

[0067] The image and text recognition is used for problem analysis of preset simple questions. Optionally, the preset simple questions include stability problems (such as query failure and query timeout problems in stability problems), accuracy problems (such as the problem of values ​​dropping to zero, null value problems, and outlier problems).

[0068] Specifically, the page containing the analysis object and its position within that page are located based on an element selector. Here, the page refers to the data report page containing the analysis object. A screenshot of the analysis object is automatically taken, and the first image of the analysis object is extracted.

[0069] Step S4032: Perform image-text recognition on the first image based on the image-text recognition model to obtain the target text features of the first image.

[0070] For example, see Figure 5 The first image is Figure 5 Taking the image shown as an example, image-text recognition is performed on the first image using an image-text recognition model, and the target text feature of the first image is "Query failed". See also Figure 6 The first image is Figure 6 Taking the image shown as an example, image recognition is performed on the first image using an image recognition model, and the target text feature of the first image is "No data available". Furthermore, if the first image contains charts, image recognition can also be performed on the first image using an image recognition model to obtain the target text feature of the first image, such as indicators with a value of zero in the chart, or other types of zero values, such as 0.0, 0.00%, etc.

[0071] Step S4033: Obtain the pre-configured problem text features of the first analysis target.

[0072] In practical applications, various question text features corresponding to the second analysis objectives can be collected. Then, the question text features corresponding to the first analysis objective can be queried from these various question text features corresponding to the second analysis objectives.

[0073] Optionally, the issue text features for query failures include text features such as: "loading failed", "query failed", "reload component", "reload page", "please try", "sorry, loading error", etc.

[0074] Optionally, the issue text features for query timeout issues include text features such as "No data available" and "Query timed out".

[0075] Optionally, the problem text features for problems where the value drops to zero include various types of zero values.

[0076] Step S4034: Based on the matching results between the target text features and the question text features, obtain the question analysis results of the analysis object.

[0077] Specifically, if the target text features match the problem text features, it indicates that the analyzed object has a problem corresponding to the first analysis target, and the analysis result for the analyzed object is determined to be that the analysis object fails the verification. If the target text features do not match the problem text features, it indicates that the analyzed object does not have a problem corresponding to the first analysis target, and the analysis result for the analyzed object is determined to be that the analysis object passes the verification. Of course, in practical applications, if the target test case also encapsulates large language model recognition, then it is necessary to combine the problem analysis results of the large language model recognition to determine the problem analysis result for the analyzed object.

[0078] For example Figure 5 As shown, if the first analysis target is a query failure problem, and the target text feature in the first image of the analysis object is identified as "query failure" through the image recognition model, and the problem text feature corresponding to the query failure problem includes "query failure", then it is determined that the analysis object has a query failure problem, so as to obtain the problem analysis result of the analysis object.

[0079] The data analysis method provided in this embodiment includes an image and text recognition method, which can be used to identify simple problems and improve the efficiency of simple problem identification.

[0080] This embodiment provides a third data analysis method, which can be used in data report testing platforms, such as the server side of a testing platform. Figure 7 This is a flowchart illustrating the third data analysis method according to an embodiment of the present disclosure, as shown below. Figure 7 As shown, the process includes the following steps:

[0081] Step S701: In response to the analysis command for the analysis object, obtain the first analysis target and target configuration information of the analysis object. The target configuration information is used to characterize the correspondence between at least one second analysis target and test cases. The test cases encapsulate at least one problem identification method, which includes one or more of image and text recognition and large language model recognition. For details, please refer to... Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0082] Step S702: Based on the target configuration information, query the target test cases that match the first analysis target. For details, please refer to [link to relevant documentation]. Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0083] Step S703: Call the target test case to perform problem analysis on the analysis object and obtain the problem analysis results of the analysis object.

[0084] In some optional implementations, step S703 above includes:

[0085] Step S7031: If the target test case encapsulates large language model recognition, call the target test case to extract the second image of the analysis object from the page where the analysis object is located.

[0086] Among them, large language model recognition is used for problem analysis of preset difficult problems. Optionally, preset difficult problems include abnormal fluctuations in curves, consistency problems across pages, and consistency problems of historical pages.

[0087] Specifically, the page containing the analysis object and its position within that page are located based on an element selector. Here, the page refers to the data report page containing the analysis object. A screenshot of the analysis object is automatically taken, and a second image of the analysis object is extracted.

[0088] Step S7032: Obtain the first prompt word. The first prompt word is used to instruct the large language model to perform problem analysis on the second image.

[0089] For each analysis object, a corresponding first prompt word can be obtained.

[0090] See Figure 8 In related technologies, manually reviewing data reports to identify difficult issues mainly involves the following steps: Researchers or data analysts open the data report dashboard and cycle through the charts in the data report. They observe the trends of the curves in the charts and generate problem analysis results. Finally, they collect the problem analysis results and produce a problem analysis report.

[0091] Specifically, accuracy issues include large fluctuations in indicators, outliers in indicators (such as negative values, percentage anomalies, etc.), and large fluctuations in chart data within the most recent preset time period (such as one day) (see...). Figure 9 ), chart data stagnation (see) Figure 10 )wait.

[0092] The manual review process for issues related to significant fluctuations in indicators includes: opening the data report dashboard, polling the charts in the data report, observing whether the volatility indicator meets expectations, and generating an analysis result. The manual review process for issues related to outliers in indicators includes: opening the data report dashboard, polling the charts in the data report, observing the indicator data, month-on-month values, growth rates, and other indicators, and generating an analysis result. The manual review process for significant fluctuations in chart data within the recent preset time period includes: opening the data report dashboard, polling the charts in the data report, observing the curve fluctuations, and generating an analysis result based on historical fluctuations and the business scenario. The manual review process for stagnant chart data includes: opening the data report dashboard, polling the charts in the data report, observing the curve fluctuations, and generating an analysis result based on historical fluctuations and the business scenario.

[0093] In practical applications, based on experts' experience with various secondary analysis objectives corresponding to large language models, prompts for the large language model can be developed to enable the identification of difficult problems in data reports using the large language model. Specifically, prompts for multiple rounds of problem analysis can be configured for different secondary analysis objectives to perform multiple rounds of problem analysis on the secondary images of the analysis object.

[0094] Optionally, the first prompt includes dependency information for problem analysis, anomaly classification information, and at least one analysis method. Dependency information includes one or more of the following: curve trends, coordinate axis data, and standard curve values, which can be adjusted according to actual conditions. Anomaly classification information includes classification information for one or more of the following anomaly categories: spike anomalies, trough anomalies, data mutation anomalies, data stagnation anomalies, periodic mutation anomalies, local noise anomalies, and anomaly point cluster anomalies, which can be adjusted according to actual conditions. Analysis methods include one or more of the following: problem judgment conditions, boundary conditions, relaxation strategies, and output formats. The relaxation strategy is used to characterize the identification of non-performance problems. The analysis methods also include precautions and calculation methods for each analysis method.

[0095] Optionally, the classification information for spike anomalies includes a significant increase in data. The classification information for trough anomalies includes a significant decrease in data. The classification information for data mutation anomalies includes a significant change in data trend compared to historical trends, i.e., the data trend changes from a stable or declining trend to an upward trend, or vice versa. The classification information for data stagnation anomalies includes data remaining at the same value within a recent preset time period. The classification information for periodic anomalies includes a sudden change in the periodic pattern of the data. The classification information for local noise anomalies includes sudden, irregular, and large fluctuations in the curve within a certain interval. The classification information for anomaly cluster anomalies includes the sudden appearance of multiple consecutive large outlier values ​​in the data.

[0096] Step S7033: Input the first prompt word and the second image into the large language model, perform problem analysis on the second image, and obtain the problem analysis results of the analysis object.

[0097] Understandably, the first cue word is used to instruct the large language model to perform problem analysis on the second image based on the dependency information of problem analysis, anomaly classification information, and at least one analysis method, so as to obtain the problem analysis results of the analyzed object.

[0098] The data analysis method provided in this embodiment includes a large language model recognition method. Therefore, it is possible to use the large language model recognition method to identify difficult problems, so as to achieve accurate identification of difficult problems.

[0099] In some optional implementations, step S4034 includes: inputting the first prompt word and the second image into the large language model, performing multiple rounds of problem analysis on the second image to obtain the problem analysis results of the analysis object; wherein, the input data for the problem analysis in the remaining rounds other than the first round of problem analysis includes the output results of the previous round of problem analysis, and the output results include the problems existing in the second image and their judgment basis.

[0100] Specifically, in each round of problem analysis, the large language model classifies the analyzed object into anomalies based on dependency information, anomaly classification information, and analysis methods to identify the problems present in the second image. Furthermore, it derives the criteria for determining the problems present in the second image based on dependency information and analysis methods.

[0101] The data analysis method provided in this embodiment uses the output results of the previous round of problem analysis as the input data for the next round of problem analysis. Therefore, it can perform drill-down analysis on the problems existing in the second image of the analysis object to improve the accuracy of the problem analysis results.

[0102] In some optional implementations, the first prompt word and the second image are input into a large language model, and multiple rounds of question analysis are performed on the second image to obtain the question analysis results of the analysis object, including:

[0103] Step a1: In the first round of problem analysis, the first prompt word and the second image are input into the large language model, and the problem analysis is performed on the second image to obtain the output results of the first round of problem analysis.

[0104] The output of the first round of problem analysis includes the problems identified in the second image during the first round of problem analysis, as well as the basis for judging these problems.

[0105] Step a2: In the remaining rounds of problem analysis, a second prompt word is generated based on the output of the previous round of problem analysis. The second prompt word is input into the large language model, and the problem analysis of the second image is performed in combination with the problem samples in the target storage space to obtain the output results of the remaining rounds of problem analysis. The problem samples include image samples and the problems existing in the image samples.

[0106] Specifically, see Figure 11 In the remaining rounds of problem analysis, for the problem analysis other than the last round (such as...) Figure 11 The second round of problem analysis (shown) generates a second prompt word based on the output of the previous round of problem analysis. The second prompt word is input into the large language model, and the problem analysis of the second image is performed in combination with the problem samples in the target storage space to obtain the output results of the corresponding round of problem analysis.

[0107] Optionally, the second prompt word for the final round of question analysis also includes preset question filtering conditions. The second prompt word for the final round of question analysis is used to instruct the large language model to filter the questions in the output results of the previous round of question analysis based on the preset question filtering conditions.

[0108] Among them, the preset problem filtering conditions are used to ignore minor anomalies and historical anomalies. That is, the preset problem filtering conditions are used to ignore problems whose changes were less than the preset range in the previous round of problem analysis, as well as problems identified in the past.

[0109] Specifically, see Figure 11 For the final round of problem analysis, a second prompt word is generated based on the output of the previous round of problem analysis. The second prompt word is input into the large language model, and the problems in the output of the previous round of problem analysis are filtered in combination with the preset problem filtering conditions to obtain the output of the final round of problem analysis.

[0110] Step a3: Determine the problem analysis results of the analysis object based on the output results of the last round of problem analysis.

[0111] The problem analysis results include the verification results of the analyzed object and the corresponding judgment criteria. The verification results are obtained based on the problems in the output of the last round of problem analysis, and the judgment criteria are the judgment criteria for the problems in the output of the last round of problem analysis.

[0112] The data analysis method provided in this embodiment combines problem samples in the target storage space to perform problem analysis on the second image during the problem analysis process of the large language model. Therefore, it can use problem samples to perform retrieval enhancement generation on the large language model to improve the accuracy of the problem analysis results.

[0113] For example, see Figure 12 , Figure 12 This is the first clue word corresponding to an anomaly in the curve graph. The first clue word includes information on the dependencies for problem analysis, anomaly classification information, and analysis methods. The analysis methods include judgment conditions, boundary conditions, relaxation strategies, and output format. For example... Figure 12As shown, the judgment conditions include: Method 1: "Due to the nature of the business, there is no need to focus on historical anomalies on the curve; only analyze whether there are abnormal fluctuations in the data of the latest day in the curve." In addition, the judgment conditions also include: Method 2: "Please analyze and judge based on the curve trend in the graph and experience. If there is labeled data in the graph, please analyze the fluctuations based on the curve trend and experience," Method 3: "If there are coordinate axes in the graph and labeled data, where the x-axis data represents the □ period and the y-axis data represents the indicator, please analyze and judge based on the xy-axis data," and Method 4: "For anomalies identified through Method 1 and Method 2, please find the data before the anomaly occurred on the corresponding curve, as well as the maximum and minimum values ​​of the anomaly area. Calculate the percentage of anomalies according to the precautions and calculation methods below and convert it to a positive number for use in relaxing the judgment strategy." The precautions for Method 4 include: Point 1: "If there are no data labels in the image, please skip the calculation of the anomaly percentage!"; Point 2: "The specific values ​​of the data before the anomaly, the maximum value of the anomaly, and the minimum value of the anomaly should be estimated based on their positions, the labels on the y-axis, and the labels on the curve!"; Point 3: "Data on the same curve is usually only labeled at the far left and far right of the curve. These two data points only represent the oldest and newest dates. Data at points in the middle of the curve should be estimated based on their positions and the labels on the y-axis; do not directly use the labels at the two ends!"; Point 4: "The labels on different curves in the image are different colors; please distinguish them and do not confuse the data from different curves!"; Point 5: "If multiple labels are adjacent and closely arranged vertically in the image, they represent data from different curves, not the same curve; please distinguish them and do not confuse the data from different curves!" Method 4 calculates the following: "If the y-axis data is percentage data, then the percentage of anomalies = the data at the outlier (the labeled maximum or minimum value of the outlier area) - the data before the anomaly. For example, if the data before the curve anomaly is approximately 20%, and the data at the outlier is approximately 55%, then the percentage of anomalies for the curve is 55% - 20% = 35%, and the result becomes 35% when converted to a positive number." "If the y-axis data is not percentage data, then the percentage of anomalies = (data at the outlier - data before the anomaly) / data before the anomaly * 100%. For example, if the data before the curve anomaly is approximately 3000, and the data at the outlier is approximately 1500, then the percentage of anomalies for the curve is (1500 - 3000) / 3000 * 100% = -50%, and the result becomes 50% when converted to a positive number."The relaxed strategies include: Strategy 1: "If the fluctuation trend of the curve has a certain periodicity, or is not significantly different from the fluctuation pattern of the first half of the curve, please consider this fluctuation to be normal." Additionally, it includes Strategy 2: "If the difference between the labeled values ​​at both ends of the curve is large, please analyze the date length on the x-axis and the overall trend of the curve. If the date length is greater than two months, and the growth or decline trend is relatively stable, please consider this fluctuation to be normal." Strategy 3: "The indicator has gradually increased from 2 million to 3.5 million over 3 months. Although the volatility exceeds the range, it is still normal fluctuation." "If a data stagnation problem is identified, as long as the labeled values ​​displayed on the curve are not completely consistent, regardless of the magnitude of the difference, please consider this fluctuation to be normal." Output formats include: "If any curve in the image has any problem, please return: Verification failed and the reason for the failure." "If there are no abnormal problems in the image, please return: Verification passed." "If the image cannot be judged, please return: Unable to judge."

[0114] Furthermore, Figure 12 After the first prompt word shown is input into the large language model, the following results can be obtained: Figure 13 The problem analysis results are shown. Among them, such as... Figure 13 As shown, the problem analysis results include the verification results of the analyzed object (such as verification failure) and the basis for judging the verification results (such as "the actual value curve has no fluctuations from April 6, 2025 to June 21, 2025 and the duration exceeds one month, indicating a data stagnation problem").

[0115] In some optional implementations, the data analysis method of this disclosure further includes: in response to a task configuration instruction for a target data report, generating an inspection task for the target data report and execution conditions for the inspection task, wherein the inspection task includes an analysis object in the target data report and a first analysis target of the analysis object; and, if the execution conditions of the inspection task are met, generating an analysis instruction for the analysis object.

[0116] Specifically, the test platform's client displays a configuration page based on the automated testing framework within the user interface. Users configure the analysis objects to be analyzed in the target data report, their primary analysis targets, and the inspection plans for these objects, generating task configuration instructions. The server responds to these task configuration instructions for the target data report, generating inspection tasks and their execution conditions. The server periodically executes these inspection tasks based on the execution conditions, generating analysis instructions for the analysis objects to achieve automated inspection of the analysis objects in the target data report.

[0117] The data analysis method provided in this embodiment offers configuration functionality for inspection tasks on target data reports. Therefore, it is possible to configure corresponding inspection tasks for different target data reports to achieve automated inspection of the analysis objects within the target data reports and automatically identify analysis objects with problems in the target data reports.

[0118] In some optional implementations, the data analysis method of this disclosure further includes: generating a problem analysis report based on the problem analysis results; and pushing the problem analysis report to the target object.

[0119] Specifically, by regularly performing inspection tasks, multiple problem analysis results can be obtained for the analyzed objects. The obtained problem analysis results are statistically analyzed to obtain a problem analysis report.

[0120] In practical applications, the following can be adopted: Figure 14 The message card format shown indicates that the robot pushes problem analysis reports to the target object. The message card displays the number of successful verifications for the analyzed object (e.g., 42) and the number of failed verifications (e.g., 2). It also displays the problem analysis results for each failed verification object, including the corresponding test cases, the reasons for the failure, and suggested fixes. Furthermore, it displays controls for inspection tasks and problem analysis reports. Users can click the inspection task control to view the task, or click the problem analysis report control to view the detailed problem analysis report.

[0121] In practical applications, the following can also be used: Figure 15 The problem analysis report, as shown, is pushed to the target audience. The report includes the results of the inspection task and analytical indicators, enabling the target audience to determine whether there are any anomalies in the data reports.

[0122] The data analysis method provided in this embodiment can automatically collect the problem analysis results and notify the target object of the problem analysis results.

[0123] As a concrete application example, an automated testing platform client is installed on a computer. Users can configure inspection tasks and execution conditions for target data reports through the client. The inspection task includes the analysis object in the target data report and the first analysis target of the analysis object. When the execution conditions of the training task are met, the server generates analysis instructions for the analysis object, uses the data analysis method disclosed herein to perform problem analysis on the analysis object, and obtains the problem analysis results. The server generates a problem analysis report based on the problem analysis results and pushes the problem analysis report to the client corresponding to the target object for display on the client.

[0124] This disclosed data analysis method enables the construction of a user interface-based automated testing platform based on an open-source functional automation testing framework (such as Robot Framework). Within this platform, modular encapsulation of test cases for image / text recognition and large language model recognition expands the platform's testing capabilities, supporting automated testing of data reporting products. Furthermore, automated testing capabilities are used to analyze common data issues in data reports, such as stability, accuracy, and consistency problems. Simultaneously, problem features, such as problematic text features, can be extracted from test cases using expert experience. These features are then provided to the image / text recognition model or large language model, simulating the testing behavior of testers reviewing data reports. This achieves automated report inspection, ensuring the stability, accuracy, and consistency of data reports. Specifically, for large language model recognition, multi-round prompts can be generated based on abnormal curve fluctuations and expert testing experience, enabling the large language model to perform multiple rounds of problem analysis, effectively identifying difficult issues in reports. Building upon this foundation, supporting automation capabilities can be developed for automated data report inspection, such as supporting the rapid generation of automated test cases for different data reports and the rapid creation of inspection tasks, saving testers' manpower costs. It also automatically performs scheduled inspection tasks to proactively identify issues in the data reports, improving the efficiency of problem analysis. During the inspection process, it can automatically render the problem analysis results for the data reports, generating a problem analysis report, which is then pushed to users via a chatbot for automatic notification of the results. It also provides regular problem analysis reports to relevant personnel, such as developers and testers, enabling them to evaluate and resolve issues in the data reports, achieving rapid response in data report maintenance. Ultimately, it achieves a one-stop automated inspection capability for data reports, improving the testing efficiency of testers.

[0125] This embodiment also provides a data analysis device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0126] This embodiment provides a data analysis device, such as... Figure 16 As shown, it includes:

[0127] The response module 1601 is used to respond to the analysis command for the analysis object, obtain the first analysis target and target configuration information of the analysis object, the target configuration information is used to characterize the correspondence between at least one second analysis target and the test case, and the test case encapsulates at least one problem identification method, which includes one or more of image and text recognition and large language model recognition.

[0128] The query module 1602 is used to query target test cases that match the first analysis target based on the target configuration information.

[0129] Analysis module 1603 is used to call the target test cases to perform problem analysis on the analysis object and obtain the problem analysis results of the analysis object.

[0130] In some alternative implementations, the analysis module 1603 includes:

[0131] The first extraction unit is used to call the target test case when the target test case includes image and text recognition, and extract the first image of the analysis object from the page where the analysis object is located.

[0132] The image and text recognition unit is used to perform image and text recognition on the first image based on the image and text recognition model to obtain the target text features of the first image.

[0133] The feature acquisition unit is used to acquire the pre-configured question text features of the first analysis target.

[0134] The feature matching unit is used to obtain the problem analysis results of the analysis object based on the matching results between the target text features and the problem text features.

[0135] In some alternative implementations, the analysis module 1603 includes:

[0136] The second extraction unit is used to call the target test case and extract the second image of the analysis object from the page where the analysis object is located, when the target test case encapsulates a large language model recognition.

[0137] The prompt word acquisition unit is used to acquire the first prompt word, which is used to instruct the large language model to perform problem analysis on the second image.

[0138] The model recognition unit is used to input the first prompt word and the second image into the large language model, perform problem analysis on the second image, and obtain the problem analysis results of the analyzed object.

[0139] In some optional implementations, the model recognition unit includes:

[0140] The model recognition subunit is used to input the first prompt word and the second image into the large language model, and to perform multiple rounds of problem analysis on the second image to obtain the problem analysis results of the analysis object. The input data for the problem analysis in the remaining rounds, except for the first round of problem analysis, includes the output results of the previous round of problem analysis. The output results include the problems existing in the second image and the judgment basis.

[0141] In some optional implementations, the model recognition subunit is specifically used for: in the first round of problem analysis, inputting the first prompt word and the second image into the large language model, performing problem analysis on the second image, and obtaining the output result of the first round of problem analysis; in the remaining rounds of problem analysis, generating the second prompt word based on the output result of the previous round of problem analysis, inputting the second prompt word into the large language model, and performing problem analysis on the second image in conjunction with the problem samples in the target storage space, and obtaining the output result of the remaining rounds of problem analysis, wherein the problem samples include image samples and the problems existing in the image samples; and determining the problem analysis result of the analysis object based on the output result of the last round of problem analysis.

[0142] In some optional implementations, the second prompt word for the final round of question analysis also includes preset question filtering conditions. The second prompt word for the final round of question analysis is used to instruct the large language model to filter questions in the output of the previous round of question analysis based on the preset question filtering conditions.

[0143] In some optional embodiments, the data analysis apparatus of this disclosure further includes:

[0144] The configuration module is used to respond to the task configuration instructions for the target data report, generate the inspection task of the target data report and the execution conditions of the inspection task. The inspection task includes the analysis object in the target data report and the first analysis target of the analysis object.

[0145] The instruction generation module is used to generate analysis instructions for the analysis object when the execution conditions of the inspection task are met.

[0146] In some optional embodiments, the data analysis apparatus of this disclosure further includes:

[0147] The report generation module is used to generate problem analysis reports based on the problem analysis results.

[0148] The push module is used to push problem analysis reports to the target audience.

[0149] The data analysis apparatus provided in this disclosure can execute the data analysis method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects for executing the method. The data analysis apparatus of this disclosure pre-configures test cases for at least one second analysis target. Each test case encapsulates at least one problem identification method from image / text recognition and large language model recognition to adapt to different analysis targets. Furthermore, when data reports need to be analyzed, it can respond to analysis instructions for the analysis object and obtain the first analysis target of the analysis object. Using the first analysis target and target configuration information, it quickly finds target test cases matching the first analysis target, calls the pre-configured target test cases that match the first analysis target, and automatically performs problem analysis on the analysis object to obtain the problem analysis results, without needing to configure test cases separately for each data report's analysis object. Therefore, it can quickly analyze problems existing in data reports.

[0150] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0151] Figure 17 This is a structural block diagram of an electronic device provided in an embodiment of the present disclosure.

[0152] The following is a detailed reference. Figure 17 The diagram illustrates a structural block diagram suitable for implementing an electronic device according to embodiments of the present disclosure. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 1701, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1702 or a program loaded from memory 1708 into random access memory (RAM) 1703. The RAM 1703 also stores various programs and data required for the operation of the electronic device. The processor 1701, ROM 1702, and RAM 1703 are interconnected via a bus 1704. An input / output (I / O) interface 1705 is also connected to the bus 1704.

[0153] Typically, the following devices can be connected to I / O interface 1705: input devices 1706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 1707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 1708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1709. Communication device 1709 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 17 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0154] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory 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 communication device 1709, or installed from memory 1708, or installed from ROM 1702. When the computer program is executed by processor 1701, it performs the functions defined in the data analysis method of embodiments of this disclosure.

[0155] Figure 17 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0156] This disclosure also provides a computer-readable storage medium in which the methods described in this disclosure can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium may also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the data analysis methods shown in the above embodiments are implemented.

[0157] A portion of this disclosure can be applied to computer program products, such as computer program instructions, which, when executed by a computer, can invoke or provide methods and / or technical solutions according to this disclosure through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, and installation package files. Accordingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions; the computer compiling the instructions and then executing the corresponding compiled program; the computer reading and executing the instructions; or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0158] Although embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A data analysis method, characterized in that, The method includes: In response to an analysis command for an analysis object, the system obtains a first analysis target and target configuration information for the analysis object. The target configuration information is used to characterize the correspondence between at least one second analysis target and a test case. The test case encapsulates at least one problem identification method, which includes one or more of image and text recognition and large language model recognition. Based on the target configuration information, query the target test cases that match the first analysis target; The target test case is invoked to perform problem analysis on the analysis object, and the problem analysis results of the analysis object are obtained.

2. The method according to claim 1, characterized in that, The step of calling the target test case to perform problem analysis on the analysis object and obtaining the problem analysis results of the analysis object includes: If the target test case includes the image and text recognition, the target test case is invoked to extract the first image of the analysis object from the page where the analysis object is located. Based on the image-text recognition model, the first image is subjected to image-text recognition to obtain the target text features of the first image; Obtain the pre-configured question text features of the first analysis target; Based on the matching results between the target text features and the question text features, the question analysis results of the analysis object are obtained.

3. The method according to claim 1, characterized in that, The step of calling the target test case to perform problem analysis on the analysis object and obtaining the problem analysis results of the analysis object includes: When the target test case encapsulates the large language model recognition, the target test case is invoked to extract the second image of the analysis object from the page where the analysis object is located. Obtain the first prompt word, which is used to instruct the large language model to perform problem analysis on the second image; The first prompt word and the second image are input into the large language model, and the second image is analyzed to obtain the analysis result of the analysis object.

4. The method according to claim 3, characterized in that, The step of inputting the first prompt word and the second image into the large language model, performing problem analysis on the second image, and obtaining the problem analysis result of the analyzed object includes: The first prompt word and the second image are input into the large language model, and multiple rounds of problem analysis are performed on the second image to obtain the problem analysis results of the analysis object; wherein, the input data for the problem analysis in the remaining rounds of problem analysis, except for the first round of problem analysis, includes the output results of the previous round of problem analysis, and the output results include the problems existing in the second image and their judgment basis.

5. The method according to claim 4, characterized in that, The step of inputting the first prompt word and the second image into the large language model, and performing multiple rounds of question analysis on the second image to obtain the question analysis results of the analysis object includes: In the first round of problem analysis, the first prompt word and the second image are input into the large language model, and the second image is analyzed to obtain the output result of the first round of problem analysis; In the remaining rounds of problem analysis, a second prompt word is generated based on the output of the previous round of problem analysis. The second prompt word is input into the large language model, and the second image is analyzed in conjunction with the problem samples in the target storage space to obtain the output of the remaining rounds of problem analysis. The problem samples include image samples and the problems existing in the image samples. The problem analysis result of the analyzed object is determined based on the output of the last round of problem analysis.

6. The method according to claim 5, characterized in that, The second prompt word in the final round of problem analysis also includes preset problem filtering conditions. The second prompt word in the final round of problem analysis is used to instruct the large language model to filter the problems in the output results of the previous round of problem analysis based on the preset problem filtering conditions.

7. The method according to claim 1, characterized in that, Also includes: In response to a task configuration instruction for a target data report, an inspection task for the target data report and the execution conditions for the inspection task are generated. The inspection task includes the analysis object in the target data report and the first analysis target of the analysis object. If the execution conditions of the inspection task are met, an analysis instruction is generated for the object being analyzed.

8. The method according to claim 1, characterized in that, Also includes: A problem analysis report will be generated based on the problem analysis results. The problem analysis report is pushed to the target audience.

9. A data analysis device, characterized in that, The device includes: The response module is used to respond to the analysis command for the analysis object, obtain the first analysis target and target configuration information of the analysis object, the target configuration information is used to characterize the correspondence between at least one second analysis target and test cases, and the test cases encapsulate at least one problem identification method, the problem identification method including one or more of image and text recognition and large language model recognition; The query module is used to query target test cases that match the first analysis target based on the target configuration information; The analysis module is used to call the target test case to perform problem analysis on the analysis object and obtain the problem analysis results of the analysis object.

10. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the data analysis method of any one of claims 1 to 8 by executing the computer instructions.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the data analysis method according to any one of claims 1 to 8.

12. A computer program product, characterized in that, Includes computer instructions for causing a computer to perform the data analysis method according to any one of claims 1 to 8.

Images