A software quality evaluation method, related device, storage medium and program product

Abstract

The application discloses a software quality evaluation method, related equipment, a storage medium and a program product. The method is used for collecting quality data of a target software in a historical time period, extracting numerical features of the quality data, and then performing index prediction based on software attributes of the target software, the numerical features of the quality data and the quality data to obtain reference indexes for quality evaluation of the target software. The correlation degree analysis is performed on each reference index, and a quality index is selected from each reference index based on the correlation degree analysis result. The correlation degree between any two quality indexes is less than a correlation degree threshold. Finally, an expected index value interval of the quality index is configured, and the software quality of the target software is evaluated based on the quality index and the expected index value interval to obtain a quality evaluation result of the target software. Thus, the objectivity and accuracy of the quality evaluation result can be improved.

Description

Technical Field

[0001] This application relates to the field of software development technology, and in particular to a software quality assessment method, related equipment, storage medium, and program product. Background Technology

[0002] With the rapid development of software development technology, a wide variety of software products are emerging in the market. Since the quality of software products directly impacts their competitiveness and market share, accurate quality assessment has become crucial. Generally, software product quality assessment requires pre-selecting relevant quality indicators and then evaluating the software product's quality by testing the values ​​of those indicators. However, current software quality assessment methods rely on professionals combining historical experience to formulate quality indicators and evaluation standards, resulting in assessments that are not entirely objective or rigorous, and consequently, inaccurate results. Summary of the Invention

[0003] This application provides a software quality assessment method, related equipment, storage medium, and program product, which can automatically select software to be assessed.

[0004] On the one hand, embodiments of this application provide a software quality assessment method, including:

[0005] Collect quality data of the target software over a historical period and extract the numerical features of the quality data. The target software refers to software for which quality assessment is required.

[0006] Based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data, index prediction is performed to obtain reference indicators for quality assessment of the target software.

[0007] A correlation analysis was performed on each reference indicator, and a quality indicator was selected from the reference indicators based on the correlation analysis results. The correlation between any two quality indicators was less than the correlation threshold.

[0008] Configure the expected value range of the quality indicators, and perform software quality assessment on the target software based on the quality indicators and the expected value range to obtain the quality assessment result of the target software.

[0009] In one implementation, the correlation analysis results are used to indicate the correlation between various reference indicators; the selection of quality indicators from the reference indicators based on the correlation analysis results includes:

[0010] Iterate through each of the reference indicators and obtain the correlation between the currently traversed reference indicator and other reference indicators from the correlation analysis results;

[0011] Add reference indicators with a correlation degree greater than or equal to the correlation degree threshold to the same indicator group to obtain at least one indicator group;

[0012] Reference indicators that have never been added to an indicator group and quality indicators selected from each indicator group, wherein no quality indicators belonging to the same indicator group are selected.

[0013] Furthermore, embodiments of this application provide a software quality assessment apparatus, comprising:

[0014] The data acquisition unit is used to collect quality data of the target software within a historical time period and extract the numerical characteristics of the quality data. The target software refers to software that has a quality assessment requirement.

[0015] The indicator prediction unit is used to predict indicators based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data, so as to obtain reference indicators for quality evaluation of the target software.

[0016] The indicator selection unit is used to perform correlation analysis on each reference indicator and select quality indicators from the reference indicators based on the correlation analysis results. The correlation between any two quality indicators is less than the correlation threshold.

[0017] A quality assessment unit is used to configure the expected value range of the quality indicators, and to perform software quality assessment on the target software based on the quality indicators and the expected value range, so as to obtain the quality assessment result of the target software.

[0018] In one implementation, the quality data includes at least two data items, and the numerical characteristics of the quality data include the numerical characteristics of each data item; when the indicator prediction unit performs indicator prediction based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data to obtain reference indicators for quality assessment of the target software, it may specifically perform the following:

[0019] Based on the software attributes and the quality data, quality keywords are extracted to obtain the quality keywords corresponding to the quality data. Each quality keyword is used to describe a quality assessment dimension.

[0020] Based on the data items and their numerical characteristics, a quality fluctuation analysis is performed to obtain the quality fluctuation index of each data item. The quality fluctuation index is used to indicate the probability that the corresponding data item will cause quality fluctuation.

[0021] The large language model is invoked to generate reference indicators for quality assessment of the target software based on the quality assessment dimensions described by the quality keywords and the target data items whose quality fluctuation index is greater than the index threshold.

[0022] In another embodiment, the number of quality indicators is at least two; the quality assessment unit performs a software quality assessment on the target software based on the range of the quality indicators and the expected indicator values, and obtains the quality assessment result of the target software, which can be specifically executed as follows:

[0023] Based on the at least two quality indicators, data dimensionality reduction is performed to obtain potential quality indicators;

[0024] Obtain the correlation coefficient between each quality indicator and the potential quality indicator, and use the quality indicators with a correlation coefficient greater than a preset threshold as the associated quality indicators of the potential quality indicators;

[0025] The potential quality indicators, the associated quality indicators of the potential quality indicators, and the expected value range of each associated quality indicator are used as evaluation parameters in the software quality assessment task. Based on the evaluation parameters, the target software is evaluated to obtain the quality assessment result of the target software.

[0026] In another embodiment, the evaluation parameters further include the indicator weights of potential quality indicators, and the quality evaluation unit can also be used to perform:

[0027] Obtain the information retention rate of the potential quality indicators for the at least two quality indicators;

[0028] Based on the information retention rate, the indicator weights of the potential quality indicators are configured, and the information retention rate is positively correlated with the indicator weights of the potential quality indicators.

[0029] In another embodiment, the evaluation parameters further include the weights of the associated quality indicators; the quality data includes at least two data items and historical data values ​​for each data item within the historical time period, and the associated quality indicator is related to at least one data item in the quality data; the quality evaluation unit can also be used to perform:

[0030] Based on the data items related to the associated quality indicator in the quality data and their respective historical data values, the historical data values ​​of the associated quality indicator are determined.

[0031] Based on the number of associated quality indicators of the potential quality indicators and the historical data values ​​of the associated quality indicators, the information entropy of the associated quality indicators is determined.

[0032] Based on the information entropy of the associated quality indicators, the indicator weights of the associated quality indicators are configured, and the indicator weights of the associated quality indicators are negatively correlated with the information entropy.

[0033] In another implementation, when the quality assessment unit performs a software quality assessment on the target software based on the assessment parameters and obtains the quality assessment result of the target software, it may specifically perform the following:

[0034] Obtain the current index values ​​of the target software under each associated quality index;

[0035] For each of the associated quality indicators, the quality indicator score is determined based on its current indicator value, expected indicator value range, and indicator weight, and the total quality indicator score corresponding to each associated quality indicator of the potential quality indicator is obtained.

[0036] The product of the potential quality indicators and the total score of the quality indicators is used to determine the software quality score of the target software, and a quality assessment result containing the software quality score is generated.

[0037] In another implementation, when the quality assessment unit determines the quality indicator score for each of the associated quality indicators based on its corresponding current indicator value, expected indicator value range, and indicator weight, it may specifically perform the following:

[0038] The expected value range of the associated quality indicators is divided into at least one sub-range, and a quality indicator score is assigned to each sub-range.

[0039] From the at least one sub-interval, determine the target sub-interval to which the current index value of the associated quality index belongs;

[0040] The product of the quality index score corresponding to the target sub-interval and the index weight of the associated quality index is determined as the quality index score of the associated quality index.

[0041] In another embodiment, the software quality assessment apparatus further includes an assessment optimization unit, which, after obtaining the quality assessment result of the target software, can be used to perform:

[0042] Collect new quality data of the target software within a target time period, wherein the target time period includes the time period following the historical time period;

[0043] Based on the new quality data, an update strategy for the associated quality indicators is generated, and an intelligent analysis model is invoked to perform a feasibility analysis on the update strategy based on the existing quality data and the new quality data, thereby obtaining the credibility of the update strategy.

[0044] The associated quality indicators are updated based on an update strategy where the confidence level is higher than the confidence level threshold, so that the updated associated quality indicators can be used to evaluate the software quality of the target software; wherein, the indicator update includes at least one of indicator weight update, expected indicator value range update, and indicator content update.

[0045] In another implementation, when the indicator selection unit selects quality indicators from the various reference indicators based on the correlation analysis results, it may specifically perform the following:

[0046] The large language model is invoked to perform indicator analysis on each reference indicator based on the correlation analysis results and the quality data, and the indicator analysis results are used to indicate one or two of the indicator importance and indicator optimization strategies.

[0047] Based on the results of the indicator analysis and the correlation analysis, the quality indicator is selected from the various reference indicators.

[0048] In another embodiment, the quality data includes at least two data items and historical data values ​​for each data item within a historical time period, and the quality indicator is related to at least one data item in the quality data; when configuring the expected value range of the quality indicator, the quality assessment unit may specifically perform the following:

[0049] Based on the historical data values ​​corresponding to the data items related to the quality indicators in the quality data, the numerical distribution characteristics of the quality indicators are determined.

[0050] Based on the numerical distribution characteristics, the numerical distribution density interval of the quality indicator is determined, and based on the numerical distribution density interval, the expected value range of the quality indicator is determined.

[0051] In another implementation, when the data acquisition unit is used to collect quality data of numerical type corresponding to the target software to be evaluated within a historical time period, it can specifically perform the following:

[0052] Obtain numerical software data of the target software within a historical time period, wherein the software data includes at least one of code modification data, operational data, and test data;

[0053] The software data is subjected to outlier detection based on distribution hypothesis to obtain a first outlier dataset, and the first outlier dataset is subjected to outlier detection based on machine learning to obtain a second outlier dataset.

[0054] Obtain the data source of each abnormal data in the first abnormal dataset and the second abnormal dataset;

[0055] The intelligent analysis model is invoked to optimize the software data based on the first abnormal dataset, the second abnormal dataset, and the data sources of each abnormal dataset, to obtain the quality data.

[0056] In another aspect, embodiments of this application provide a computer device, including:

[0057] A memory, wherein a computer program is stored;

[0058] A processor for loading the computer program to implement the method as described in the first aspect.

[0059] In another aspect, embodiments of this application also provide a computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed as described in the first aspect.

[0060] In another aspect, embodiments of this application also provide a computer program product, the computer program product including computer instructions, wherein a processor of a computer device reads the computer instructions and executes the method as described in the first aspect.

[0061] This application first predicts reference indicators for quality assessment of the target software based on its software attributes, historical quality data, and numerical characteristics. Then, by performing correlation analysis on the predicted reference indicators, indicators with correlation less than a threshold are selected as quality indicators. Finally, the software quality is assessed based on these quality indicators and their expected value ranges. This ensures that the quality indicators used in the assessment are comprehensively determined based on multiple objectively existing software data (software attributes, historical quality data, and their numerical characteristics). Using such quality indicators for software quality assessment can cover a more comprehensive range of quality influencing factors and avoid subjective biases in the assessment results due to insufficient experience or cognitive biases. This guarantees the objectivity of the quality indicators and assessment results and can improve the accuracy and reliability of the assessment results to a certain extent. Furthermore, in the embodiments of this application, the correlation between the various quality indicators is lower than the correlation threshold, which can avoid the use of redundant or duplicate quality indicators during software quality assessment. This allows the software quality assessment to focus on independent and key quality indicators, ensuring that the quality assessment results are not distorted due to the overlap of quality indicators, and further improving the credibility of the quality assessment results. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a schematic diagram illustrating a software quality assessment principle provided in an embodiment of this application;

[0064] Figure 2 This is a schematic diagram of a software quality assessment environment provided in an embodiment of this application;

[0065] Figure 3 This is a schematic flowchart of a software quality assessment method provided in an embodiment of this application;

[0066] Figure 4 This is a schematic diagram of a software data acquisition method provided in an embodiment of this application;

[0067] Figure 5 This is a schematic diagram illustrating a method for filtering software data according to an embodiment of this application;

[0068] Figure 6 This is a schematic diagram of a reference index generation method provided in an embodiment of this application;

[0069] Figure 7 This is a schematic diagram illustrating a method for selecting quality indicators provided in an embodiment of this application;

[0070] Figure 8 This is a schematic diagram illustrating a method for determining a range of expected indicator values ​​provided in an embodiment of this application;

[0071] Figure 9 This is a schematic flowchart of another software quality assessment method provided in the embodiments of this application;

[0072] Figure 10 This is a schematic diagram of the factor loading matrix provided in the embodiments of this application;

[0073] Figure 11 This is a schematic diagram illustrating the generation principle of a potential quality indicator provided in an embodiment of this application;

[0074] Figure 12 This is a schematic diagram illustrating the relationship between potential quality indicators and associated quality indicators provided in an embodiment of this application;

[0075] Figure 13 This is a schematic diagram illustrating a software quality assessment principle provided in an embodiment of this application;

[0076] Figure 14 This is a schematic diagram illustrating the configuration principle of an expected index value range provided in an embodiment of this application;

[0077] Figure 15 This is a schematic diagram of the logical structure of a software quality assessment device provided in an embodiment of this application;

[0078] Figure 16 This is a schematic diagram of the structure of a software quality assessment device provided in an embodiment of this application;

[0079] Figure 17 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0080] It should be noted in advance that, in order to enable those skilled in the art to better understand the technical solutions proposed in the embodiments of this application, the embodiments of this application will be described clearly and completely in conjunction with one or more accompanying drawings. Furthermore, the accompanying drawings shown in the embodiments of this application are merely illustrative examples; for instance, the execution order of each step in the drawings can be adaptively adjusted according to the actual application scenario.

[0081] Furthermore, in the embodiments of this application, the block diagrams, modules, and units shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. Each module or unit can be part of a larger module or unit that includes the functionality of that module or unit. That is, the terms "module" or "unit" mentioned in the embodiments of this application refer to a computer program or part of a computer program with a predetermined function, which can work together with other related parts to achieve a predetermined goal. It can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof, or implemented in different network and / or processor devices and / or microcontroller devices. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units.

[0082] Specifically, this application provides an automated execution scheme for objectively evaluating software quality. The main principles employed in this scheme for software quality evaluation can be found in [link to relevant documentation]. Figure 1 .like Figure 1Based on the principles illustrated, the technical solution proposed in this application, when dealing with target software requiring quality assessment, can first predict reference indicators for quality assessment of the target software based on the software attributes, historical quality data, and numerical characteristics of the quality data. Then, by performing correlation analysis on the predicted reference indicators, each reference indicator with a correlation degree less than the correlation degree threshold is selected as a quality indicator. Finally, expected indicator value ranges are configured for the quality indicators, and the software quality is assessed based on the quality indicators and their expected indicator value ranges to obtain the quality assessment results of the target software.

[0083] Quality data refers to data related to the target software that can be used to measure software quality. Quality data can include data generated at various stages of the target software's lifecycle, such as software development data, software testing data, and software operation data. This application does not impose specific limitations on this. Software attributes are used to characterize at least one feature of the target software, and this at least one feature includes business characteristics. For example, business characteristics can be the type of business provided by the target software (such as social networking, audio / video playback, image processing, or gaming) or the type of object the target software serves (such as teenagers, the elderly, women, men, students, or job seekers). Optionally, software attributes can also be used to characterize features such as code complexity, test coverage, and user scenario diversity. User scenario diversity refers to the diverse usage scenarios and differences in needs arising from different factors such as time, location, identity, task objectives, and environmental conditions when business objects use the product or service.

[0084] In this solution, the quality indicators used for quality assessment of the target software are selected from reference indicators. These reference indicators are derived through comprehensive prediction based on three objective software data sources: software attributes, historical quality data, and their numerical characteristics. Automated indicator prediction using multiple objective data sources not only ensures that the selected quality indicators cover a wider range of quality influencing factors but also avoids subjective biases caused by insufficient human experience or cognitive biases, guaranteeing the objectivity of the quality indicators and thus improving the objectivity and accuracy of the quality assessment results. Furthermore, the correlation between the selected quality indicators is below a correlation threshold, avoiding the use of redundant or duplicate quality indicators during software quality assessment. This allows the software quality assessment to focus on independent and critical quality indicators, ensuring that the assessment results are not distorted due to overlapping indicators, thereby guaranteeing the credibility of the quality assessment results. In other words, the quality indicators in this application are achieved by combining data-driven indicator prediction and data analysis-driven indicator selection, making the quality assessment process of the target software essentially a quantifiable, data-driven process. This ensures that the target software receives an objective quality assessment, thereby guaranteeing the accuracy and reliability of the quality assessment results.

[0085] In one feasible implementation, the above technical solution can be deployed in, for example, Figure 2 The application is shown in the software quality assessment environment. Figure 2 In the diagram, boxes marked with 201 indicate n data sources related to the target software, where n is a positive integer. Different data sources are used to provide data of different qualities, such as... Figure 2 Data source 1 can be used to provide code update records, data source 2 can be used to provide operational data, and data source 3 can be used to provide test data. Figure 2 In the diagram, boxes marked with 202 show a variety of computer devices that can be used to execute this application. These computer devices can also run the target software and / or provide support services such as data storage and data calculation for the target software. This application does not impose specific limitations on the type and number of computer devices.

[0086] based on Figure 1 The present application proposes a software quality assessment method based on the illustrated software quality assessment principles. A schematic flowchart of this method can be found here. Figure 3 And can be made by Figure 2 To execute using any of the computer devices shown. For example... Figure 3 As shown, the method may specifically include steps S301-S304:

[0087] S301. Collect quality data of the target software within a historical time period and extract the numerical characteristics of the quality data. The target software refers to software for which quality assessment is required.

[0088] In specific embodiments, the target software can be a batch of software with the same quality assessment requirements, or it can be a single software with quality assessment requirements. Optionally, the quality assessment requirements can be specifically used to describe the quality assessment objectives (such as assessing the overall quality of the software, assessing the performance quality of the software when facing peak traffic, etc.) or quality assessment dimensions (such as performance dimension, code dimension, market response dimension, etc.). Quality data refers to data related to the target software and that can be used to measure the software quality, and can specifically include at least one data item, with each data item corresponding to a software data dimension.

[0089] In one exemplary implementation, the composition of the quality data may be as shown in Table 1.

[0090] Table 1

[0091]

[0092]

[0093] As shown in Table 1, the software data dimensions used by computer equipment when collecting quality data can include software development, software testing, software operation, software feedback, and so on. It should be noted that the data items collected under each software data dimension in this embodiment need to be representable using numerical data types in order to extract the numerical features of the quality data. These numerical features can include one or both of the following: numerical statistical features (such as mean, extreme values, standard deviation, mode, etc.) and numerical distribution features (such as Poisson distribution, normal distribution, gamma distribution, etc.).

[0094] In one implementation, when collecting quality data for target software, the computer device can first acquire numerical software data of the target software over a historical period, and then use mathematical analysis and machine learning to filter the software data to obtain quality data. Specifically, the software data can include at least one of code modification data, operational data, and test data, and the data sources for different types of software data can be different. The data source can specifically indicate the data source providing the software data, and / or indicate the functional module (or business module) to which the software data belongs.

[0095] For example, computer equipment can be referenced. Figure 4 The method shown is used to obtain software data. For example... Figure 4Software data can, for example, include business platform data, externally reported data, and cached data in a database. Computer devices can employ different data interaction methods when acquiring different types of software data. For instance, when collecting software data from business platform data, this can be achieved by configuring message subscription events on the Application Programming Interface (API). These message subscription events are used to store business data as software data when the API detects a transmission of business data. Similarly, when collecting software data from externally reported data, network communication protocols (such as TCP / IP) can be used to transmit the externally reported data to the corresponding storage space, thereby achieving the acquisition of software data. These are just a few examples; further details are omitted here.

[0096] After collecting the software data, the computer equipment can first perform outlier detection based on distributional hypothesis to obtain a first outlier dataset. Then, it can further perform machine learning-based outlier detection on the first outlier dataset to obtain a second outlier dataset. The distributional hypothesis outlier detection can be based on the 3σ principle in mathematical analysis, while the machine learning outlier detection can include at least one of Isolation Forest (IF), Local Outlier Factor (LOF), and Support Vector Machine (SVM). Next, the data sources of each outlier data point in the first and second outlier datasets are obtained. Then, an intelligent analysis model is invoked to optimize the software data based on the first and second outlier datasets and their respective data sources to obtain quality data. It can be understood that since the software data contains data items and their values, the outlier data obtained from outlier detection can be either data items or data values.

[0097] Among them, the intelligent analysis model refers to a neural network model with data analysis capabilities. These capabilities can be acquired through training the model with relevant sample data. As an example, when optimizing software data, the intelligent analysis model can first infer the data production environment of each abnormal data point based on its data source (e.g., peak traffic periods, low-pressure system testing phases, scenarios with intense competition from similar products). Then, it obtains the outlier judgment rules associated with each data production environment and predicts the confidence level of the abnormal data based on its corresponding abnormal dataset and outlier judgment rules. Finally, it removes abnormal data with a confidence level higher than a confidence threshold (e.g., 90%) from the software data to obtain the final quality data.

[0098] In one exemplary implementation, the computer device can be configured according to... Figure 5 The method shown is used to filter software data. For example... Figure 5 The computer equipment can first perform outlier detection on the software data based on the 3σ principle. Based on the results of this outlier detection, it can further perform one or more outlier detection methods, such as IF, LOF, and SVM. Then, it combines the outlier detection results from both stages (including all detected outliers), the business characteristics of the target software, and the data sources of each outlier to perform intelligent analysis, obtaining the final outliers that need to be removed. Quality data is then obtained by removing these outliers from the software data. Optionally, the computer equipment can also perform other data filtering processes on the software data, such as null value removal, illegal format removal, and duplicate feature removal, to further ensure the usability of the quality data.

[0099] It is worth mentioning that, to ensure the reference value of quality data and its numerical characteristics, the embodiments of this application typically record relevant software data of the target software at multiple historical moments within a historical time period, making the screening of quality data based on a large amount of software data. When the collected software data has a large amount of variable information (such as a large number of data items and their data values), the software data can essentially be regarded as a high-dimensional dataset. Based on this characteristic, the present applicant proposes that when performing anomaly detection on large-scale software data, anomaly detection methods that do not require data distribution assumptions (such as isolated forest detection, i.e., IF) can be preferentially selected to improve the efficiency of anomaly detection.

[0100] S302. Based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data, predict the indicators to obtain reference indicators for quality assessment of the target software.

[0101] In a specific embodiment, based on the description of step S301 above, it is easy to understand that the quality data may include at least two data items and one or more historical data values ​​corresponding to each data item within a historical time period. In this case, the numerical features of the quality data may include the numerical features of each data item, which are obtained by extracting numerical features based on the corresponding historical data values.

[0102] In one exemplary implementation, when determining reference indicators for the target software based on software attributes, numerical characteristics of quality data, and quality data, the computer device may refer to, for example, Figure 6 The principle shown is achieved by performing the following steps (1)-(3):

[0103] (1) Extract quality keywords based on software attributes and quality data to obtain the quality keywords corresponding to the quality data.

[0104] Specifically, a natural language processing model can be invoked to identify the text characters contained in the quality data. Then, these text characters are semantically understood in combination with software attributes. Based on the semantic understanding results and these text characters, at least one quality keyword is generated, and each quality keyword is used to describe a quality assessment dimension.

[0105] (2) Based on each data item and its numerical characteristics, quality fluctuation analysis is performed to obtain the quality fluctuation index of each data item. The quality fluctuation index is used to indicate the probability that the corresponding data item will cause quality fluctuation.

[0106] Specifically, to ensure the accuracy of the quality fluctuation index, quality fluctuation analysis can be performed using a relevant neural network model. This model can be pre-trained based on quality scores and quality data samples. When performing quality fluctuation analysis for any data item, the trained model first obtains the meaning of the data item, and then combines this meaning with the numerical characteristics of the data item to perform multi-dimensional data analysis to determine whether the data item is likely to cause quality fluctuations. Generally, when the meaning of a data item is close to the quality description information, the greater the fluctuation amplitude represented by the numerical characteristics, the more likely the data item is to cause quality fluctuations.

[0107] (3) Call the large language model based on the quality assessment dimensions described by quality keywords, and the target data items whose quality fluctuation index is greater than the index threshold, to generate reference indicators for the target software in quality assessment.

[0108] Specifically, the reference indicators generated in this application embodiment are related to at least one target data item in the quality data. As an exemplary implementation, after determining the quality fluctuation index of each data item, the computer device can first select target data items whose quality fluctuation index is greater than an index threshold from the data items included in the quality data. Then, it can further determine target data items corresponding to the quality assessment dimension from the selected target data items. Finally, it can create one or more reference indicators based on the target data items under each quality assessment dimension to obtain reference indicators for quality assessment of the target software. By creating one or more reference indicators based on target data items under the same quality assessment dimension, information integration of multiple target data items can be achieved, thereby minimizing the number of reference indicators and reducing subsequent storage and computational costs.

[0109] Alternatively, in other implementations, the computer device may directly use the target data items corresponding to the quality assessment dimensions as reference indicators to maximize the scope of quality assessment that the reference indicators can cover, thereby improving the accuracy of the quality assessment results to a certain extent.

[0110] S303. Perform correlation analysis on each reference indicator, and select quality indicators from each reference indicator based on the correlation analysis results. The correlation between any two quality indicators shall be less than the correlation threshold.

[0111] In a specific embodiment, since the reference indicators are related to at least one data item in the quality data, the quality indicators selected from the various reference indicators based on the correlation analysis results should also be related to at least one data item in the quality data. The correlation analysis results are used to indicate the correlation between the various reference indicators. The correlation can specifically include one or more of the following: Pearson Product-Moment Correlation Coefficient (PPMCC), Spearman's Rank Correlation Coefficient (SRCC), and Kendall Rank-Order Correlation Coefficient (KRCC). As an example, when the correlation includes multiple correlation coefficients, the correlation coefficient with the highest confidence level can be used as the reference correlation when selecting the quality indicator. The confidence level can be determined based on the differences between different correlation coefficients. For example, if a certain correlation coefficient differs significantly from the other two correlation coefficients, then the confidence level of that correlation coefficient is considered low.

[0112] In one feasible implementation, when a computer device selects quality indicators from reference indicators based on the correlation analysis results, it can first traverse each reference indicator and obtain the correlation degree between the currently traversed reference indicator and other reference indicators from the correlation analysis results. Then, reference indicators with a correlation degree greater than or equal to the correlation degree threshold are added to the same indicator group to obtain at least one indicator group corresponding to the reference indicators of the target software. Then, quality indicators are selected from the reference indicators that have not been added to the indicator group and from each indicator group. Among the selected quality indicators, there are no quality indicators belonging to the same indicator group.

[0113] Specifically, when selecting quality indicators from reference indicators that have not been added to the indicator group and from each indicator group, each reference indicator that has not been added to the indicator group can be selected as a quality indicator, and quality indicators can be selected from each indicator group in turn. The i-th indicator group used to select quality indicators does not contain any quality indicators that have been selected from the previous i-1 groups, where i is a positive integer, so that there are no quality indicators belonging to the same indicator group among the selected quality indicators, thereby ensuring that the correlation between any two quality indicators is less than the correlation threshold.

[0114] For example, suppose there are five reference indicators (1-5). After grouping these indicators based on their correlation, we get three indicator groups (with duplicates removed): indicator group A (based on reference indicator 1), indicator group B (based on reference indicator 2), and indicator group C (based on reference indicator 4). Indicator group A = {reference indicator 1, reference indicator 2}, indicator group B = {reference indicator 2, reference indicator 1, reference indicator 4}, and indicator group C = {reference indicator 4, reference indicator 2}. Then, the reference indicators not added to any indicator group are reference indicators 3 and 5. In this case, the computer device can directly use reference indicators 3 and 5 as quality indicators, and then select quality indicators sequentially from indicator groups A, B, and C. If reference indicator 1 is selected as a quality indicator from indicator group A, then reference indicator 1 cannot exist in other indicator groups used to select quality indicators. That is, the next indicator group used to select quality indicators should be indicator group C rather than indicator group B, so as to ensure that the currently selected quality indicator exists in a different indicator group from other selected quality indicators.

[0115] In another feasible implementation, when selecting quality indicators from reference indicators based on the correlation analysis results, the computer equipment can also follow the method described above. Figure 7 The principle illustrated is used to achieve this. Specifically, the large language model is first invoked to perform indicator analysis on each reference indicator based on the correlation analysis results and quality data to obtain the indicator analysis results. Then, based on the indicator analysis results and the correlation analysis results, quality indicators are selected from each reference indicator. The indicator analysis results are used to indicate one or both of the indicator importance and indicator optimization strategies.

[0116] When the indicator analysis results are used to indicate importance, reference indicators with a correlation degree lower than the correlation degree threshold can be further selected from reference indicators with importance higher than the corresponding threshold as quality indicators. Alternatively, reference indicators with an importance degree higher than the corresponding threshold can be further selected from all reference indicators with a correlation degree lower than the correlation degree threshold as quality indicators. When the indicator analysis results are used to indicate indicator optimization strategies, the corresponding reference indicators can be optimized based on the indicator optimization strategy. Then, a new correlation analysis can be performed on each optimized reference indicator. Finally, quality indicators can be selected based on the new correlation analysis results and the optimized reference indicators. The specific selection method can be found in the principle shown in the previous implementation method, and will not be detailed here.

[0117] It should be noted that the Large Language Model (LLM) in this application embodiment is a generic term. That is to say, the Large Language Model mentioned in this step and the Large Language Model mentioned in step S302 above can be the same neural network model or different neural network models.

[0118] S304. Configure the expected value range of the quality indicators, and conduct a software quality assessment of the target software based on the quality indicators and the expected value range to obtain the quality assessment result of the target software.

[0119] In a specific embodiment, the expected indicator value range refers to the range of values ​​to which the expected quality indicator corresponds in the target software. This range typically covers all common indicator values ​​of the corresponding quality indicator. That is, in practical applications, indicator values ​​outside this range can be considered as abnormal indicator values. The handling of abnormal indicator values ​​(such as removal, prioritization, etc.) can be determined according to specific needs and is not limited here.

[0120] In one feasible implementation, the expected value range of each quality indicator can be set based on empirical values. For example, it can be configured manually. Alternatively, when the historical quality indicators used by the target software include the current quality indicators, the expected value range corresponding to the historical quality indicators can be directly used as the expected value range of the current quality indicators.

[0121] In another feasible implementation, since quality data can include at least two data items and each data item's historical data values ​​within a historical time period, and the quality indicator is related to at least one data item in the quality data, this application embodiment proposes that the computer device can first determine the numerical distribution characteristics of the quality indicator based on the historical data values ​​of the data items related to the quality indicator, then determine the numerical distribution density interval of the quality indicator based on the numerical distribution characteristics, and finally determine the expected indicator value range of the quality indicator based on the numerical distribution density interval.

[0122] In determining the numerical distribution characteristics of a quality indicator, one can first obtain the numerical relationship between the quality indicator and its related data items. Then, based on the historical data values ​​of each related data item, the historical indicator values ​​can be determined. Finally, through statistical analysis of the historical indicator values, the numerical distribution characteristics of the quality indicator can be obtained. The numerical distribution characteristics of a quality indicator can be used to characterize the probability distribution followed by the indicator value, such as Poisson distribution, normal distribution, chi-square distribution, gamma distribution, uniform distribution, etc. The dense interval of the numerical distribution of a quality indicator refers to the numerical interval composed of indicator values ​​with a high probability of occurrence. For example, a computer device can calculate the distribution probability of the quality indicator based on the probability distribution indicated by the numerical distribution characteristics, using the corresponding distribution function (such as a probability mass function, probability density function, cumulative probability function, etc.). When the distribution probability is greater than the probability threshold associated with the distribution function, the dense interval of the numerical distribution is determined based on the input parameters corresponding to the distribution function.

[0123] As an example, when determining the expected value range of a quality indicator based on a densely distributed interval of numerical values, the computer device can directly use the densely distributed interval as the expected value range, or it can further calculate the quartiles of the densely distributed interval after determining it, and then determine the expected value range based on the quartiles. For an example, the method of determining the expected value range based on quartiles can be found in [link to relevant documentation]. Figure 8 .

[0124] like Figure 8 The computer equipment can first calculate the median, upper quartile, and lower quartile of the densely distributed numerical range, then use the difference between the upper quartile and the lower quartile as the interquartile range (represented by IQR), and then calculate the upper bound of the expected index value range according to Equation 1, and the lower bound of the expected index value range according to Equation 2.

[0125] Ran_Max=max(Q3+1.5*IQR,Max) Formula 1

[0126] In Equation 1, Ran_Max refers to the upper bound of the expected index value range (e.g., ...). Figure 8 The upper bound of the interval is defined as follows: max(*) is the maximum value function, which takes the maximum value within the parentheses as the function value; Q3 refers to the upper quartile; IQR refers to the interquartile range; Max refers to the upper bound of the interval where the numerical distribution is dense.

[0127] Ran_Min=mim(Q1-1.5*IQR, Min) Equation 2

[0128] In Equation 2, Ran_Min refers to the lower bound of the expected index value range (e.g., ...). Figure 8The lower bound of the interval is defined as follows: min(*) is the minimum value function, which takes the minimum value within the parentheses as the function value; Q1 refers to the lower quartile; IQR refers to the interquartile range; Min refers to the lower bound of the interval where the numerical distribution is dense.

[0129] After determining the expected range of each quality indicator through the above methods, the large language model can be further invoked to conduct a retrospective analysis and trend prediction on the historical data of each quality indicator. Based on the indicator fluctuation records indicated by the retrospective analysis results and the indicator change trends indicated by the trend prediction results, the expected range of each quality indicator can be comprehensively adjusted to ensure the scientific validity (or accuracy) of the expected range when it is put into use in the future.

[0130] This application first predicts reference indicators for quality assessment of the target software based on its software attributes, historical quality data, and numerical characteristics. Then, by performing correlation analysis on the predicted reference indicators, indicators with correlation less than a threshold are selected as quality indicators. Finally, the software quality is assessed based on these quality indicators and their expected value ranges. This ensures that the quality indicators used in the assessment are comprehensively determined based on multiple objectively existing software data (software attributes, historical quality data, and their numerical characteristics). Using such quality indicators for software quality assessment can cover a more comprehensive range of quality influencing factors and avoid subjective biases in the assessment results due to insufficient experience or cognitive biases. This guarantees the objectivity of the quality indicators and assessment results and can improve the accuracy and reliability of the assessment results to a certain extent. Furthermore, in the embodiments of this application, the correlation between the various quality indicators is lower than the correlation threshold, which can avoid the use of redundant or duplicate quality indicators during software quality assessment. This allows the software quality assessment to focus on independent and key quality indicators, ensuring that the quality assessment results are not distorted due to the overlap of quality indicators, and further improving the credibility of the quality assessment results.

[0131] Based on the above Figure 3 In addition to the corresponding implementation methods, this application also proposes another software quality assessment method, the schematic flowchart of which can be found in [reference needed]. Figure 9 And it can still be executed by the aforementioned computer devices. Specifically, such as Figure 9 As shown, the method may include steps S901-S906:

[0132] S901. Collect quality data of the target software within a historical time period and extract the numerical characteristics of the quality data. The target software refers to software for which quality assessment is required.

[0133] S902. Based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data, predict the indicators to obtain reference indicators for quality assessment of the target software.

[0134] S903. Perform correlation analysis on each reference indicator, and select quality indicators from each reference indicator based on the correlation analysis results. The correlation between any two quality indicators shall be less than the correlation threshold.

[0135] In specific embodiments, the specific implementation of steps S901-S903 can be found in the implementation described in steps S301-S303 above, and will not be repeated here.

[0136] S904. Configure the expected value range of the quality indicators, and perform data dimensionality reduction based on each quality indicator to obtain the potential quality indicators.

[0137] In a specific embodiment, the configuration method for the expected indicator value range is the same as that described in step S304 above, and will not be repeated here. When there are two or more quality indicators, the computer device can first perform data dimensionality reduction processing on the quality indicators to obtain potential quality indicators. Here, a potential quality indicator is an abstract concept; specifically, it can be understood as an indicator that cannot be directly observed or obtained, but can be indirectly inferred from multiple quality indicators. One potential quality indicator can cover the quality assessment range corresponding to one or more quality indicators.

[0138] In one feasible implementation, potential quality indicators can be obtained through dimensionality reduction methods such as factor analysis (FA) or principal component analysis (PCA). Taking factor analysis as an example, performing factor analysis on the quality indicators of the target software yields a factor loading matrix (see diagram). Figure 10 This factor loading matrix can be used to indicate at least one latent indicator (typically each column of the matrix corresponds to one latent indicator), where each matrix element is the correlation coefficient between a quality indicator and a latent indicator, such as... Figure 10 x_1 in the matrix represents the correlation coefficient between potential indicator 1 and the original indicator ① (or quality indicator ①). The entire factor loading matrix is ​​used to indicate the correlation coefficient between each quality indicator and each potential indicator.

[0139] After obtaining the factor loading matrix, the sum of squared correlation coefficients of each latent indicator in the factor loading matrix can be used as the eigenvalue of that latent indicator. It should be noted that the eigenvalue here represents the number of quality indicators that a latent indicator can explain. For example, an eigenvalue greater than or equal to 1 indicates that the explanatory power of the latent factor is at least equivalent to that of a single original quality indicator. Based on this, after determining the eigenvalues ​​of each latent indicator, the computer device can select latent quality indicators from the latent indicators according to the "eigenvalue greater than 1 criterion (Kaiser-Criterion Principle, KC principle)," ensuring that the eigenvalues ​​of the selected latent quality indicators are greater than or equal to 1. This allows for effective dimensionality reduction of the quality indicator set through latent quality indicators.

[0140] In another implementation, potential quality indicators can be obtained through intelligent analysis of quality indicators by a neural network model (such as the aforementioned large language model) based on its learned data processing experience.

[0141] Based on the two implementation methods described above, it is easy to see that the principle of obtaining potential quality indicators in the embodiments of this application can be exemplarily referred to [reference needed]. Figure 11 .like Figure 11 As shown, when it is necessary to obtain potential quality indicators based on a set of quality indicators corresponding to the target software (composed of selected quality indicators), the computer equipment can first perform a factor analysis sufficiency test on the quality indicator set to determine whether the quality indicator set is suitable for factor analysis. If it is suitable for factor analysis, then factor analysis is used to determine the potential quality indicators; if it is not suitable for factor analysis, then a neural network model is used to determine the potential quality indicators. Optionally, the factor analysis sufficiency test can include the Kaiser-Meyer-Olkin Measure (KMO) and Bartlett's Test of Sphericity (Bartlett's Test of Sphericity). The KMO test is used to detect the degree to which the quality indicator set is suitable for factor analysis, and the Bartlett's Test of Sphericity is used to detect whether the quality indicator set conforms to a multivariate normal distribution. As an example, if the quality indicator set passes both the KMO test and the Bartlett's Test of Sphericity, then the quality indicator set can be considered suitable for determining potential quality indicators using factor analysis.

[0142] In specific applications of this application, the computer device can further obtain the indicator weights of the potential quality indicators after obtaining them. In one exemplary implementation, if the potential quality indicators are determined by a neural network model through intelligent analysis, the indicator weights of the potential quality indicators can be further predicted by the neural network model based on the meaning of the potential quality indicators. Specifically, the neural network model can predict the priority of each potential quality indicator according to the quality assessment requirements of the target software, and then assign weights to each potential quality indicator based on the priority of the potential indicators.

[0143] In one exemplary implementation, if the potential quality indicators are determined using factor analysis, the indicator weights of the potential quality indicators can be determined based on the information retention (or variance explained) of the potential quality indicators, and the information retention is positively correlated with the indicator weights. Information retention typically indicates the amount of information retained by a potential quality indicator from the total information conveyed by all quality indicators, and specifically, it can be the ratio of the retained information to the total information.

[0144] For example, information retention can be calculated using the correlation coefficients associated with potential quality indicators, as represented by the factor loading matrix. For instance, the correlation coefficients associated with potential quality indicator 1 refer to... Figure 10 Given a set of correlation coefficients labeled 1001, the information retention of potential quality index 1 can be based on the correlation coefficients labeled 1001. Figure 10 The correlation coefficient is calculated using the 1001-labeled value. As an example, a computer device can use the sum of squares of the correlation coefficients corresponding to the potential quality indicator in the factor loading matrix as the information retention of that potential quality indicator.

[0145] Optionally, information retention can also be calculated using the correlation coefficients associated with potential quality indicators, as represented by the rotated factor loading matrix. The rotation can be orthogonal or oblique. Orthogonal rotations can include Varimax, Quartimax, etc.; oblique rotations can include Promax, Direct Oblimin, etc. By rotating the factor loading matrix, the factor structure can be simplified, ensuring that a quality indicator is strongly correlated with only one potential quality indicator, thereby improving the clarity of the potential quality indicator's meaning and ultimately enhancing the accuracy of subsequent quality assessment results based on the potential quality indicator.

[0146] It is worth noting that in practical applications, the sum of the information retention rates corresponding to each selected potential quality indicator is not necessarily 1 (for example, the sum of information retention rates may only be 80%). That is to say, these potential quality indicators cannot fully represent all quality indicators. However, in order to obtain more easily interpretable indicator weights and subsequent related data processing results (such as software quality scores), this application proposes that after obtaining the sum of the information retention rates of each potential quality indicator, the computer device can normalize the information retention rate of each potential quality indicator based on the sum of the information retention rates, and determine the indicator weights according to the normalized information retention rates.

[0147] S905. Obtain the correlation coefficient between each quality indicator and the potential quality indicator, and use the quality indicators with a correlation coefficient greater than a preset threshold as the associated quality indicators of the potential quality indicators.

[0148] In a specific embodiment, the relationship between potential quality indicators and associated quality indicators can be as follows: Figure 12 As shown. That is, there is at least one potential quality indicator, and each potential quality indicator can correspond to one or more associated quality indicators, and the associated quality indicators are essentially the previously selected quality indicators. It should be noted that, for the convenience of describing and understanding the specific implementation methods of the embodiments of this application, unless otherwise specified, the following description will use one potential quality indicator as an example to illustrate the relevant technical implementation details, and the processing methods of other potential quality indicators are the same.

[0149] As described in the implementation method in step S904 above, the correlation coefficient between each quality indicator and the potential quality indicator can be determined by the matrix elements corresponding to the potential quality indicator in the corresponding factor loading matrix. In this step, the computer device can use quality indicators with a correlation coefficient greater than a preset threshold (e.g., 80%) as the associated quality indicators of the potential quality indicator. Still using... Figure 10 For example, suppose Figure 10 If potential indicator 1 is identified as a potential quality indicator, and both x_1 and y_1 are greater than 80%, then the computer equipment can use the original indicator ① and the original indicator ② as the associated quality indicators of the potential quality indicator.

[0150] Furthermore, the computer equipment can also obtain the weights of various related quality indicators. It can be understood that since related quality indicators are essentially still quality indicators, and quality indicators are related to at least one data item in the quality data, related quality indicators should also be related to at least one data item. In this case, the computer equipment can determine the weight of the related quality indicator based on the historical data values ​​of the data items related to the related quality indicator.

[0151] As an example, computer equipment can first determine the historical data value of a related quality indicator based on the various data items related to the quality indicator in the quality data and their historical data values. Then, based on the historical data value of the related quality indicator and the number of related quality indicators of the potential quality indicator, the information entropy of the related quality indicator is determined. Subsequently, according to the weighting principle that the weight of the related quality indicator is negatively correlated with the information entropy, the weight of the related quality indicator is configured based on its information entropy. The weight calculation method based on information entropy theory can quantify the importance of each related quality indicator, ensuring the scientific nature of the subsequently constructed quality assessment system.

[0152] S906. Potential quality indicators, associated quality indicators of potential quality indicators, and the expected value range of each associated quality indicator are used as evaluation parameters in the software quality assessment task. Based on the evaluation parameters, the target software is evaluated to obtain the quality assessment result of the target software.

[0153] In a specific embodiment, the evaluation parameters may further include the indicator weights of the potential quality indicators and the indicator weights of the associated quality indicators determined in step S905. In this case, the computer device can specifically perform a software quality evaluation on the target software based on the potential quality indicators, the indicator weights of the potential quality indicators, the associated quality indicators of the potential quality indicators, and the indicator weights and expected indicator value ranges of each associated quality indicator, thereby obtaining the quality evaluation result of the target software. Optionally, the quality evaluation result may include the current quality status, and may further include quality improvement suggestions and future trend predictions, so that relevant personnel can optimize the target software in a timely manner, or optimize the quality evaluation system (including evaluation parameters, numerical relationships between evaluation parameters, etc.).

[0154] As an example, the computer device can first obtain the current index values ​​of the target software under various related quality indicators. Then, for each related quality indicator, it determines the quality index score based on its corresponding current index value, expected index value range, and index weight. Further, it obtains the total quality index score corresponding to each related quality indicator of the potential quality indicator, and determines the software quality score of the target software by multiplying the index weight of the potential quality indicator by the total quality index score. This process then generates a quality assessment result containing the software quality score.

[0155] In calculating the quality index score of a related quality indicator, the expected value range of the related quality indicator can first be divided into at least one sub-interval, and a quality index score can be assigned to each sub-interval. As an example, the quality index score corresponding to each sub-interval can be determined based on the cumulative probability of each indicator value occurring within that sub-interval, and the quality index score can be positively correlated with the cumulative probability. Then, the computer device can determine the target sub-interval to which the current indicator value of the related quality indicator belongs from the at least one sub-interval, and then determine the quality index score of the related quality indicator as the product of the quality index score corresponding to the target sub-interval and the indicator weight of the related quality indicator.

[0156] In one exemplary implementation, the software quality score can be calculated as shown in Equation 3, and the score for each quality indicator can be calculated as shown in Equation 4.

[0157]

[0158] In Equation 3, Q represents the software quality score, Li represents the i-th potential quality indicator, m is the total number of potential quality indicators identified for the target software, Wj represents the indicator weight of the j-th associated quality indicator, Sj represents the quality indicator score corresponding to the j-th associated quality indicator, and n is the number of associated quality indicators of the i-th potential quality indicator. This represents the total score of the quality indicators corresponding to each associated quality indicator of the i-th potential quality indicator.

[0159] Sj=K[B(Mj)] Equation 4

[0160] In Equation 4, Sj represents the quality index score corresponding to the j-th associated quality index, Mj represents the current index value of the j-th associated quality index, B(Mj) represents the target sub-interval to which Mj belongs, and K[] represents the quality index score corresponding to the target sub-interval.

[0161] In the practical application of this application, the indicator weights, expected indicator value ranges, and related indicator content (such as changing indicator X to indicator Y) can all be updated, and the specific updates can be achieved by collecting new quality data and performing data analysis on it.

[0162] Taking the update of associated quality indicators as an example, in one exemplary implementation, the computer device can collect new quality data corresponding to the target software within a target time period after a certain period of time, and then generate an update strategy for the associated quality indicators based on the new quality data. The target time period includes the time period following the historical time period mentioned in step S901, and may further include that historical time period. Exemplarily, the computer device can first execute the aforementioned steps S901-S905 based on the collected new quality data to determine new associated quality indicator information. Optionally, the associated quality indicator information may include one or more of the following: indicator content (such as indicator name, indicator meaning, etc.), expected indicator value range, and indicator weight. Then, the computer device can determine the update instruction information corresponding to the associated quality indicator based on the difference between the new associated quality indicator information and the existing associated quality indicator information, and then generate an update strategy for the associated quality indicators based on the determined update instruction information. The update instruction information can specifically indicate one or more of the indicator content, expected indicator value range, indicator weight, etc., that needs to be updated (e.g., added, deleted, or modified).

[0163] When testing the implementation effect of the above-mentioned technical means, the technicians found that the method of generating update strategies based on limited quality data has certain cognitive limitations, resulting in the generated update strategies not being entirely reliable. To avoid using unreliable update strategies to incorrectly update related quality indicators, this application proposes that after obtaining the update strategy, an intelligent analysis model can be further invoked to perform a feasibility analysis on the update strategy based on the original quality data and the new quality data, thus obtaining the reliability of the update strategy. Optionally, the intelligent analysis model can be a commercially available question-answering model, or a large language model trained on quality data samples from similar software. Such intelligent analysis models can typically combine their learned rich knowledge to perform comprehensive and multi-level trend analysis on limited quality data, determining the rationality and feasibility of the update strategy based on the trend, and presenting it as a reliability score. If the reliability of the update strategy is higher than the reliability threshold, the update strategy can be used to update the related quality indicators. This ensures that the updated related quality indicators help improve the accuracy of the quality assessment results obtained from the software quality assessment of the target software. The updating of related quality indicators includes at least one of the following: updating indicator weights, updating expected indicator value ranges, and updating indicator content.

[0164] This application first predicts reference indicators for quality assessment of the target software based on its software attributes, historical quality data, and numerical characteristics. Then, it performs correlation analysis on these predicted reference indicators to select those with correlation less than a threshold as quality indicators. Next, it extracts one or more potential quality indicators based on these indicators and selects associated quality indicators from each potential indicator. Finally, it combines these potential and associated quality indicators to comprehensively assess the software quality. By simultaneously focusing on potential and associated quality indicators, the quality problems of the target software can be more comprehensively identified during the quality assessment. For example, a software's potential quality indicator might include user experience, and its associated quality indicator might include CPU utilization. In this case, even if the software performs well on the associated quality indicator, there might still be issues with the user experience, meaning the overall quality of the software still needs improvement. By comprehensively assessing the software quality using both potential and associated quality indicators, the software quality score becomes more realistic due to the presence of potential quality indicators.

[0165] Based on the above Figure 3 as well as Figure 9 The software quality assessment method shown in this application also provides a software quality assessment principle, which can be exemplarily referred to in [reference needed]. Figure 13 .like Figure 13 As shown, this principle mainly involves six steps, S1301-S1306, among which:

[0166] S1301. Collect the quality data corresponding to the target software.

[0167] Specifically, computer equipment can collect data from relevant data sources through various methods, such as interface requests, cache reads, data reporting, and direct database reads, based on the software quality priorities of the target software. This collected data can then be used as quality data. Software quality priorities can include key evaluation dimensions (code quality, traffic response quality, etc.) and key evaluation stages in the software lifecycle (such as the operation phase and development phase). Understandably, the quality data collected for the same software will differ depending on the specific software quality priorities.

[0168] Therefore, in practical application scenarios, quality data can originate from various R&D stages, from business requirements to online operations. Generally, the quality data in this embodiment can come from the following software development stages: pre-sales or customer requirement collection, requirement writing, requirement presentation, technical solution review, development initiation, software development, integration testing of software modules and subsystems, product requirement experience, test initiation and requirement testing, online release after successful test verification, and online operation. Based on this, the quality data in this embodiment mainly includes business requirement data, business defect data, code data, test data, performance data, and operational data. This data generally exists in business platforms, databases, cache databases, etc., and can be automatically obtained through business platform interface requests, data reporting by the software itself, or by reading data from the database. The obtained quality data can be uniformly stored in a preset database, ultimately forming the quality data.

[0169] Because quality data may contain data items in different formats, and the quality of quality data is crucial in data analysis and quality assessment, data cleaning and reorganization are necessary. Data reorganization is an optional operation, generally used when the size of the cleaned quality data is small, to expand the quality data and ensure the integrity and logical consistency of the quality data used for quality generation. Optionally, data reorganization can include data appending, data insertion, data deletion, and data modification. Data cleaning can include one or more of the following processing methods:

[0170] (1) Extreme Outlier Removal. This method is mainly used to identify and remove extreme outliers in the data. These values ​​are usually caused by system problems, human error, or other abnormal situations. In this embodiment, distribution analysis methods (such as outlier detection based on the 3σ principle) and machine learning methods (such as IF, LOF, and SVM) can be used to comprehensively detect outliers in the quality data. After the two analyses are completed, a neural network model can be further invoked to combine the data source and the results of the two outlier analyses to jointly determine the outlier values ​​that need to be removed, and perform the final removal operation in the quality data.

[0171] (2) Illegal Format Removal. This method is mainly used to remove data that does not conform to the expected data format. In general, regular expressions or format validation tools can be used to verify the data format.

[0172] (3) Illegal value removal. This method is mainly used to remove illegal values ​​contained in quality data by combining business rules and data range. Illegal values ​​refer to values ​​that exceed the reasonable range or do not comply with business rules.

[0173] (4) Null value removal. This method is mainly used to remove data items containing null values ​​to avoid bias in analysis results caused by null values. In the embodiments of this application, methods such as deleting null values ​​or filling null values ​​can be used for processing.

[0174] Data filtering and data reorganization can effectively improve the quality and reliability of data, providing a solid and reliable data foundation for subsequent data analysis and quality assessment system construction.

[0175] S1302. Combining mathematical statistical analysis with in-depth analysis of neural network models, quality indicators are automatically generated based on quality data and software attributes.

[0176] Specifically, mathematical statistical analysis is used to analyze the numerical characteristics of each data item in quality data, while deep analysis using neural network models (usually large models) is used to predict quality indicators applicable to the target software based on numerical characteristics, software attributes, and quality data. By combining the natural language processing, machine learning, language generalization, and numerical analysis capabilities of large models with the advantages of traditional statistical methods in terms of data processing complexity, more comprehensive and accurate quality indicators can be extracted more efficiently. Moreover, because large models have rich technical knowledge reserves, when predicting quality indicators based on this technical knowledge, it is usually possible to analyze current business needs and potentially important future quality factors, and then refer to this information to generate quality indicators, ensuring that the obtained quality indicators are forward-looking.

[0177] In one implementation, mathematical statistical analysis also involves using correlation analysis algorithms (such as PPMCC, SRCC, KRCC, etc.) to perform correlation analysis on the initially generated quality indicators to obtain the correlation degree between the various quality indicators. Based on this, the large model can also be used to revise and / or evaluate the initially generated quality indicators based on the correlation degree between the various quality indicators, ultimately obtaining quality indicators with higher rationality and feasibility.

[0178] S1303. Combining mathematical statistical analysis and neural network models, configure the expected value range of each quality indicator based on quality data.

[0179] Specifically, configuring the expected indicator value range is for the purpose of conducting a logically clear and systematic evaluation of software quality. Configuring a reasonable expected indicator value range helps improve the accuracy of the quality evaluation results.

[0180] In one exemplary implementation, the principle for configuring the expected indicator value range can be found in [reference needed]. Figure 14 .like Figure 14Mathematical statistical analysis is specifically used to analyze the data of various quality indicators based on quality data, focusing on both statistical analysis and data distribution. This allows for the preliminary calculation of the dense distribution intervals of each quality indicator value, which are then used as the expected indicator value intervals. Neural network models are primarily used to combine the market environment and software characteristics of the target software, performing retrospective analysis and trend prediction based on historical software data to optimize the expected indicator value intervals.

[0181] Mathematical statistics analysis is highly rigorous, while neural network models have significant advantages in handling complex data and multivariate analysis. Applying both to the configuration stage of expected indicator value ranges can yield more accurate and reliable expected indicator value ranges.

[0182] S1304. Extract potential quality indicators and their related quality indicators, and determine the indicator weight of each indicator.

[0183] Specifically, after obtaining the set of quality indicators, factor analysis can be performed to study the internal relationships between the various quality indicators. This allows for the identification of more fundamental, but not directly observable, latent indicators hidden within the set. At least one latent quality indicator is then selected from these latent indicators, and this selected indicator should reflect most of the information in the set of quality indicators. To increase the interpretability of the latent quality indicators and facilitate the understanding of subsequent quality assessment results, computer equipment can utilize large language models to define each latent quality indicator (e.g., name it, assign it physical meaning, etc.).

[0184] The specific methods for extracting potential quality indicators and their associated quality indicators can be found in the relevant embodiments in steps S904-S905 above, and will not be described in detail here.

[0185] S1305. Based on each indicator, its weight, and the baseline, a quality assessment system is constructed to evaluate the quality of the target software.

[0186] Specifically, the quality assessment system consists of assessment parameters and the numerical relationships between them. The assessment parameters include potential quality indicators and their weights, related quality indicators of potential quality indicators and their weights, and the expected value range of each related quality indicator. The numerical relationships include those shown in Equations 3 and 4 above, which will not be detailed here.

[0187] Optionally, after constructing the quality assessment system, its accuracy can be verified. Specifically, the market response to the target software can be compared with the quality assessment results. If there is a significant difference, it indicates that the accuracy of the quality assessment system is low and needs to be optimized.

[0188] S1306. Continuously collect new quality data and iteratively optimize the quality assessment system by referring to the new quality data and quality assessment results.

[0189] Specifically, after the quality assessment system is built, new quality data can be collected periodically to iteratively optimize the system, ensuring that the system is as closely aligned with the actual business environment as possible, thereby guaranteeing the reference value of the quality assessment results.

[0190] In this embodiment, the evaluation parameters used for quality assessment of the target software are automatically generated based on objectively existing software data, combining traditional data analysis with large-scale model-assisted optimization. This reduces manual intervention in the quality evaluation system construction process, thereby improving the objectivity of quality indicators and assessment results. Furthermore, because this embodiment enables automated construction of the quality evaluation system, its construction is no longer limited by subjective cognitive biases and lack of experience among relevant personnel, greatly facilitating the construction of quality evaluation systems for various types of software. In addition, this embodiment enables automated construction and iterative optimization of the quality evaluation system, which can be widely applied to various businesses and software products requiring quality assessment. It covers the entire lifecycle of quality management from software development to final release. Moreover, the introduction of neural network models significantly improves the intelligence and automation level of quality assessment, helping software developers achieve software quality management goals more efficiently.

[0191] Based on the aforementioned method embodiments, this application also proposes a software quality assessment device, the structure of which can be exemplarily described in [reference needed]. Figure 15 .like Figure 15 As shown, a software quality evaluation device can include a user interface, memory, network interface, and processor. The processor can further include a data acquisition module, a network module, a human-computer interaction module, and a system construction module. Among these:

[0192] A user interface (GUI) is used to receive user input and interact with the user. Optionally, the GUI can be a graphical user interface (GUI) or a command-line interface (CLI), allowing users to input parameters, requirements, and feedback related to software quality assessment via input devices such as a keyboard, mouse, or touchscreen, and can exist on output devices such as a monitor. The GUI can also be a voice interface or a virtual reality interface. Voice interfaces primarily use speech recognition technology to interact with the user through voice input and feedback, while virtual reality interfaces use virtual reality (VR) technology to interact with the user through VR devices, providing a more intuitive user experience.

[0193] The storage unit is used to store data and program instructions, supporting the operation of various functional modules. The storage unit can contain distributed storage systems or cloud storage service modules. Distributed storage systems use distributed storage systems (such as Hadoop HDFS) to store and manage large-scale data; cloud storage services use cloud storage services (such as AWS S3, Google Cloud Storage) to store and manage data, providing higher storage scalability and reliability.

[0194] A network interface is used for data communication with external data sources and remote servers, supporting various network protocols (such as HTTP, HTTPS, TCP / IP, etc.). For example, a network interface can be an Ethernet interface, a Wi-Fi module, a cellular network module, etc. It is worth noting that network interfaces can also include Bluetooth interfaces and Near Field Communication (NFC) interfaces. Bluetooth interfaces use Bluetooth technology to communicate with external devices via Bluetooth connections, while NFC interfaces use NFC technology to exchange data with external devices. Designing a network interface in software quality assessment equipment ensures the timely collection of comprehensive quality data and enhances the equipment's integration capabilities.

[0195] A processor executes program instructions stored in memory to perform related data processing and analysis tasks. Specifically, the processor coordinates the work of various modules, including data acquisition, feature extraction, indicator system construction, and result output. Processors can be central processing units (CPUs), graphics processing units (GPUs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), etc.

[0196] The data acquisition module is used to collect and store software data related to software quality assessment. This module can specifically collect software data from external data sources (such as databases, log files, user feedback systems, etc.) and perform preliminary cleaning and preprocessing. As an example, the data acquisition module may include a database connection module, data acquisition scripts, data cleaning tools, etc.

[0197] The network module is used to communicate with external data sources and remote servers through a network interface. Specifically, it can be used to send data requests, receive data responses, and handle various protocols and formats in network communication. As an example, the network module may include an HTTP client library, a WebSocket communication module, and API calling tools.

[0198] The human-computer interaction (HCI) module receives user input parameters, requests, and feedback. Specifically, it processes various information input by the user through the user interface, including quality assessment parameter settings, user request descriptions, and feedback. The HCI module may consist of a form processing module, a command-line parameter parser, and user feedback collection tools.

[0199] The system construction module is used to automatically build a software quality assessment system based on large-scale models and traditional statistical methods. Specifically, this module can use large-scale models to extract features from the collected quality data, obtain the numerical features of the quality data, and further analyze the potential relationships between various data items in the quality data based on software attributes and the numerical features of the quality data, thereby extracting potential quality indicators and related quality indicators. The system construction module can also combine traditional statistical methods (such as factor analysis, information entropy, etc.) and large-scale models to jointly configure the expected value range and weight of each indicator, ultimately forming a complete quality assessment system.

[0200] In Figure 15 When the software quality assessment equipment shown is applied to software quality assessment, an illustrative process can be as follows (1)-(5). It should be noted that the following is merely a summary description to facilitate a clear understanding of the implementation principles of the embodiments of this application; specific implementation methods can be found in the foregoing. Figure 3 and Figure 9 The relevant methods are not detailed here. Specifically:

[0201] (1) User input.

[0202] Users input parameters, requirements, and feedback for quality assessment through the user interface, and the user input module receives and processes this information.

[0203] (2) Data collection.

[0204] The data acquisition module obtains software data from external data sources through network interfaces and network modules, and performs preliminary cleaning and preprocessing on the software data to obtain quality data.

[0205] (3) Feature extraction and analysis.

[0206] The system construction module uses a large model to extract and analyze features from quality data, identify potential patterns and relationships in the data, and obtain potential quality indicators and related quality indicators.

[0207] (4) Configuration of expected indicator value range and indicator weight.

[0208] By combining large-scale models and traditional statistical methods, the expected range of each quality indicator is calculated. Based on information entropy theory and factor analysis, the indicator weights of potential quality indicators and related quality indicators are calculated.

[0209] (5) Output of quality assessment results and optimization of the quality assessment system.

[0210] By using a large model to validate and optimize the quality assessment system, a high-quality quality assessment system can be obtained. Furthermore, the final quality assessment system can be output and provided to relevant programs or personnel for quality assessment of similar software.

[0211] The software quality assessment device provided in this application utilizes a large model to automatically extract key quality indicators from complex software data, reducing manual intervention and thus improving the objectivity of quality indicators. Furthermore, the large model possesses powerful pattern recognition capabilities. By combining the large model with relevant data at various stages of the quality assessment system construction, the depth and breadth of the quality assessment can be enhanced, improving the scientific rigor of the results. In summary, by combining a large model with traditional statistical methods, the software quality assessment device comprehensively leverages the rigor of traditional statistical methods with the efficient processing capabilities of the large model for complex data, resulting in more comprehensive and accurate quality assessment results.

[0212] Based on the aforementioned method embodiments, this application also proposes a software quality assessment device, which can be mounted on a computer device and used to implement the above-mentioned methods. Figure 3 or Figure 9 The described method embodiments include some or all of the functionalities. Specifically, a schematic structure of the device can be found in [reference needed]. Figure 16 .like Figure 16 The device may include a data acquisition unit 1601, an indicator prediction unit 1602, an indicator selection unit 1603, and a quality assessment unit 1604, wherein:

[0213] The data acquisition unit 1601 is used to collect quality data of the target software within a historical time period and extract the numerical features of the quality data. The target software refers to software that has a quality assessment requirement.

[0214] The indicator prediction unit 1602 is used to predict indicators based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data, so as to obtain reference indicators for quality evaluation of the target software.

[0215] The indicator selection unit 1603 is used to perform correlation analysis on each reference indicator and select quality indicators from the reference indicators based on the correlation analysis results. The correlation between any two quality indicators is less than the correlation threshold.

[0216] The quality assessment unit 1604 is used to configure the expected value range of the quality indicators, and to perform software quality assessment on the target software based on the quality indicators and the expected value range, so as to obtain the quality assessment result of the target software.

[0217] In one implementation, the quality data includes at least two data items, and the numerical characteristics of the quality data include the numerical characteristics of each data item; when the indicator prediction unit 1602 performs indicator prediction based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data to obtain reference indicators for quality assessment of the target software, it may specifically perform the following:

[0218] Based on the software attributes and the quality data, quality keywords are extracted to obtain the quality keywords corresponding to the quality data. Each quality keyword is used to describe a quality assessment dimension.

[0219] Based on the data items and their numerical characteristics, a quality fluctuation analysis is performed to obtain the quality fluctuation index of each data item. The quality fluctuation index is used to indicate the probability that the corresponding data item will cause quality fluctuation.

[0220] The large language model is invoked to generate reference indicators for quality assessment of the target software based on the quality assessment dimensions described by the quality keywords and the target data items whose quality fluctuation index is greater than the index threshold.

[0221] In another embodiment, the number of quality indicators is at least two; the quality assessment unit 1604 performs a software quality assessment on the target software based on the range of the quality indicators and the expected indicator values, and obtains the quality assessment result of the target software, which can be specifically executed as follows:

[0222] Based on the at least two quality indicators, data dimensionality reduction is performed to obtain potential quality indicators;

[0223] Obtain the correlation coefficient between each quality indicator and the potential quality indicator, and use the quality indicators with a correlation coefficient greater than a preset threshold as the associated quality indicators of the potential quality indicators;

[0224] The potential quality indicators, the associated quality indicators of the potential quality indicators, and the expected value range of each associated quality indicator are used as evaluation parameters in the software quality evaluation task. Based on the evaluation parameters, the target software is evaluated to obtain the quality evaluation result of the target software.

[0225] In another embodiment, the evaluation parameters further include the indicator weights of potential quality indicators, and the quality evaluation unit 1604 can also be used to perform:

[0226] Obtain the information retention rate of the potential quality indicators for the at least two quality indicators;

[0227] Based on the information retention rate, the indicator weights of the potential quality indicators are configured, and the information retention rate is positively correlated with the indicator weights of the potential quality indicators.

[0228] In another embodiment, the evaluation parameters further include the weights of the associated quality indicators; the quality data includes at least two data items and historical data values ​​for each data item within the historical time period, and the associated quality indicator is related to at least one data item in the quality data; the quality evaluation unit 1604 can also be used to perform:

[0229] Based on the data items related to the associated quality indicator in the quality data and their respective historical data values, the historical data values ​​of the associated quality indicator are determined.

[0230] Based on the number of associated quality indicators of the potential quality indicators and the historical data values ​​of the associated quality indicators, the information entropy of the associated quality indicators is determined.

[0231] Based on the information entropy of the associated quality indicators, the indicator weights of the associated quality indicators are configured, and the indicator weights of the associated quality indicators are negatively correlated with the information entropy.

[0232] In another embodiment, when the quality assessment unit 1604 performs a software quality assessment on the target software based on the assessment parameters and obtains the quality assessment result of the target software, it may specifically perform the following:

[0233] Obtain the current index values ​​of the target software under each associated quality index;

[0234] For each of the associated quality indicators, the quality indicator score is determined based on its current indicator value, expected indicator value range, and indicator weight, and the total quality indicator score corresponding to each associated quality indicator of the potential quality indicator is obtained.

[0235] The product of the potential quality indicators and the total score of the quality indicators is used to determine the software quality score of the target software, and a quality assessment result containing the software quality score is generated.

[0236] In another implementation, when the quality assessment unit 1604 determines the quality indicator score for each of the associated quality indicators based on its corresponding current indicator value, expected indicator value range, and indicator weight, it may specifically perform the following:

[0237] The expected value range of the associated quality indicators is divided into at least one sub-range, and a quality indicator score is assigned to each sub-range.

[0238] From the at least one sub-interval, determine the target sub-interval to which the current index value of the associated quality index belongs;

[0239] The product of the quality index score corresponding to the target sub-interval and the index weight of the associated quality index is determined as the quality index score of the associated quality index.

[0240] In another embodiment, the software quality assessment apparatus further includes an assessment optimization unit 1605, which, after obtaining the quality assessment result of the target software, can be used to perform:

[0241] Collect new quality data of the target software within a target time period, wherein the target time period includes the time period following the historical time period;

[0242] Based on the new quality data, an update strategy for the associated quality indicators is generated, and an intelligent analysis model is invoked to perform a feasibility analysis on the update strategy based on the existing quality data and the new quality data, thereby obtaining the credibility of the update strategy.

[0243] The associated quality indicators are updated based on an update strategy that prioritizes indicators with a credibility level higher than a credibility threshold, so that the updated associated quality indicators can be used to evaluate the software quality of the target software; wherein, the indicator update includes at least one of indicator weight update, expected indicator value range update, and indicator content update.

[0244] In another implementation, when the indicator selection unit 1603 selects a quality indicator from the various reference indicators based on the correlation analysis results, it may specifically perform the following:

[0245] The large language model is invoked to perform indicator analysis on each reference indicator based on the correlation analysis results and the quality data, and the indicator analysis results are used to indicate one or two of the indicator importance and indicator optimization strategies.

[0246] Based on the results of the indicator analysis and the correlation analysis, the quality indicator is selected from the various reference indicators.

[0247] In another embodiment, the quality data includes at least two data items and historical data values ​​for each data item within a historical time period, and the quality indicator is related to at least one data item in the quality data; when configuring the expected indicator value range of the quality indicator, the quality assessment unit 1604 can specifically perform the following:

[0248] Based on the historical data values ​​corresponding to the data items related to the quality indicators in the quality data, the numerical distribution characteristics of the quality indicators are determined.

[0249] Based on the numerical distribution characteristics, the numerical distribution density interval of the quality indicator is determined, and based on the numerical distribution density interval, the expected value range of the quality indicator is determined.

[0250] In another embodiment, when the data acquisition unit 1601 is used to acquire quality data of numerical type corresponding to the target software to be evaluated within a historical time period, it can specifically be used to perform:

[0251] Obtain numerical software data of the target software within a historical time period, wherein the software data includes at least one of code modification data, operational data, and test data;

[0252] The software data is subjected to outlier detection based on distribution hypothesis to obtain a first outlier dataset, and the first outlier dataset is subjected to outlier detection based on machine learning to obtain a second outlier dataset.

[0253] Obtain the data source of each abnormal data in the first abnormal dataset and the second abnormal dataset;

[0254] The intelligent analysis model is invoked to optimize the software data based on the first abnormal dataset, the second abnormal dataset, and the data sources of each abnormal dataset, to obtain the quality data.

[0255] In one embodiment, Figure 16Each unit in the illustrated device can be individually or entirely combined into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effects of the embodiments of this application. In other words, the above units are based on logical functional division. In practical applications, the function of one unit can be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of this application, Figure 16 The device shown may also include other units, and in practical applications, these functions may also be implemented with the assistance of other units, and may be implemented by multiple units working together.

[0256] According to another embodiment of this application, the following can be executed by running on a computing device including processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). Figure 3 or Figure 9 The computer program (including program code) for each step involved in the corresponding method shown, to construct such... Figure 16 The apparatus shown. A computer program may be recorded on, for example, a computer-readable storage medium, loaded onto the apparatus via the computer-readable storage medium, and executed therein.

[0257] Based on the descriptions of the above method and apparatus embodiments, this application also provides a computer device. Specifically, please refer to... Figure 17 , Figure 17 This application provides a schematic diagram of the structure of a computer device, as shown in the embodiment of the present application. Figure 17 As shown, the computer device may include a processor 1701, a memory 1702, and a communication interface 1703, and the processor 1701, the memory 1702, and the communication interface 1703 may be connected by a bus or other means.

[0258] The processor 1701 (or Central Processing Unit, CPU) is the computing and control core of a computer device. It can parse various instructions and process various data within the computer device.

[0259] Memory 1702 is a memory device in a computer device used to store programs and data. It is understood that memory 1702 here can include both the computer device's internal memory and, of course, extended memory supported by the computer device.

[0260] The communication interface 1703 may optionally include a standard wired interface or a wireless interface (such as Wi-Fi, mobile communication interface, etc.), and can be used to send and receive data under the control of the processor 1701; the communication interface 1703 can also be used for the transmission and interaction of data within the computer device.

[0261] In a specific embodiment, the processor 1701 may load and execute one or more computer programs stored in the memory 1702 to implement the steps in the method described in the above embodiments.

[0262] This application embodiment also provides a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understood that the computer-readable storage medium here can include both built-in storage media in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the processing system of the computer device. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by the processor 1701, which may be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here may be high-speed RAM memory or non-volatile memory, such as at least one disk storage device; optionally, it may also be at least one readable storage medium located remotely from the aforementioned processor.

[0263] This application also provides a computer program product, which includes computer instructions, and the processor of a computer device can execute the above-described method embodiments by loading the computer instructions.

[0264] Based on the same inventive concept, the principles and beneficial effects of the devices, computer equipment, computer-readable storage media and computer program products provided in the embodiments of this application are similar to those described in the foregoing corresponding method embodiments. Therefore, the corresponding method implementation principles and beneficial effects can be referred to, which will not be repeated here for the sake of brevity.

[0265] It should be further noted that the steps in the methods of this application embodiment can be adjusted, merged, and deleted according to actual needs, and the modules in the device of this application embodiment can be merged, divided, and deleted according to actual needs. Furthermore, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. It should also be particularly emphasized that when the above embodiments of this application are applied to specific products or technologies, the data acquisition involved in each specific implementation of this application requires the permission or consent of the relevant parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0266] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art will understand that all or part of the processes for implementing the above embodiments and equivalent variations made in accordance with the claims of this application are still within the scope of this application.

Claims

1. A software quality assessment method, characterized in that, include: Collect quality data of the target software over a historical period and extract the numerical features of the quality data. The target software refers to software for which quality assessment is required. Based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data, index prediction is performed to obtain reference indicators for quality assessment of the target software. A correlation analysis was performed on each reference indicator, and a quality indicator was selected from the reference indicators based on the correlation analysis results. The correlation between any two quality indicators was less than the correlation threshold. Configure the expected value range of the quality indicators, and perform software quality assessment on the target software based on the quality indicators and the expected value range to obtain the quality assessment result of the target software.

2. The method according to claim 1, characterized in that, The quality data includes at least two data items, and the numerical characteristics of the quality data include the numerical characteristics of each data item; the step of predicting indicators based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data to obtain reference indicators for quality assessment of the target software includes: Based on the software attributes and the quality data, quality keywords are extracted to obtain the quality keywords corresponding to the quality data. Each quality keyword is used to describe a quality assessment dimension. Based on the data items and their numerical characteristics, a quality fluctuation analysis is performed to obtain the quality fluctuation index of each data item. The quality fluctuation index is used to indicate the probability that the corresponding data item will cause quality fluctuation. The large language model is invoked to generate reference indicators for quality assessment of the target software based on the quality assessment dimensions described by the quality keywords and the target data items whose quality fluctuation index is greater than the index threshold.

3. The method according to claim 1, characterized in that, The number of quality indicators is at least two; the process of evaluating the target software based on the quality indicators and the expected indicator value range to obtain the quality evaluation result of the target software includes: Based on the at least two quality indicators, data dimensionality reduction is performed to obtain potential quality indicators; Obtain the correlation coefficient between each quality indicator and the potential quality indicator, and use the quality indicators with a correlation coefficient greater than a preset threshold as the associated quality indicators of the potential quality indicators; The potential quality indicators, the associated quality indicators of the potential quality indicators, and the expected value range of each associated quality indicator are used as evaluation parameters in the software quality assessment task. Based on the evaluation parameters, the target software is evaluated to obtain the quality assessment result of the target software.

4. The method according to claim 3, characterized in that, The evaluation parameters also include the weights of potential quality indicators, and the method further includes: Obtain the information retention rate of the potential quality indicators for the at least two quality indicators; Based on the information retention rate, the indicator weights of the potential quality indicators are configured, and the information retention rate is positively correlated with the indicator weights of the potential quality indicators.

5. The method according to claim 3, characterized in that, The evaluation parameters also include the weights of the associated quality indicators; the quality data includes at least two data items and historical data values ​​for each data item within the historical time period, and the associated quality indicators are related to at least one data item in the quality data; the method further includes: Based on the data items related to the associated quality indicator in the quality data and their respective historical data values, the historical data values ​​of the associated quality indicator are determined. Based on the number of associated quality indicators of the potential quality indicators and the historical data values ​​of the associated quality indicators, the information entropy of the associated quality indicators is determined. Based on the information entropy of the associated quality indicators, the indicator weights of the associated quality indicators are configured, and the indicator weights of the associated quality indicators are negatively correlated with the information entropy.

6. The method according to claim 3, characterized in that, The process of performing a software quality assessment on the target software based on the evaluation parameters to obtain the quality assessment result of the target software includes: Obtain the current index values ​​of the target software under each associated quality index; For each of the associated quality indicators, the quality indicator score is determined based on its current indicator value, expected indicator value range, and indicator weight, and the total quality indicator score corresponding to each associated quality indicator of the potential quality indicator is obtained. The product of the potential quality indicators and the total score of the quality indicators is used to determine the software quality score of the target software, and a quality assessment result containing the software quality score is generated.

7. The method according to claim 6, characterized in that, The step of determining the quality indicator score for each of the associated quality indicators based on its current indicator value, expected indicator value range, and indicator weight includes: The expected value range of the associated quality indicators is divided into at least one sub-range, and a quality indicator score is assigned to each sub-range. From the at least one sub-interval, determine the target sub-interval to which the current index value of the associated quality index belongs; The product of the quality index score corresponding to the target sub-interval and the index weight of the associated quality index is determined as the quality index score of the associated quality index.

8. The method according to any one of claims 3-7, characterized in that, After obtaining the quality assessment results of the target software, the method further includes: Collect new quality data of the target software within a target time period, wherein the target time period includes the time period following the historical time period; Based on the new quality data, an update strategy for the associated quality indicators is generated, and an intelligent analysis model is invoked to perform a feasibility analysis on the update strategy based on the existing quality data and the new quality data, thereby obtaining the credibility of the update strategy. The associated quality indicators are updated based on an update strategy where the confidence level is higher than the confidence level threshold, so that the updated associated quality indicators can be used to evaluate the software quality of the target software; wherein, the indicator update includes at least one of indicator weight update, expected indicator value range update, and indicator content update.

9. The method according to claim 1, characterized in that, The selection of quality indicators from the reference indicators based on the correlation analysis results includes: The large language model is invoked to perform indicator analysis on each reference indicator based on the correlation analysis results and the quality data, and the indicator analysis results are used to indicate one or two of the indicator importance and indicator optimization strategies. Based on the results of the indicator analysis and the correlation analysis, the quality indicator is selected from the various reference indicators.

10. The method according to claim 1, characterized in that, The quality data includes at least two data items and historical data values ​​for each data item within a historical time period; the quality indicator is related to at least one data item in the quality data; configuring the expected value range of the quality indicator includes: Based on the historical data values ​​corresponding to the data items related to the quality indicators in the quality data, the numerical distribution characteristics of the quality indicators are determined. Based on the numerical distribution characteristics, the numerical distribution density interval of the quality indicator is determined, and based on the numerical distribution density interval, the expected value range of the quality indicator is determined.

11. The method according to claim 1, characterized in that, The collected quality data, in numerical form, of the target software to be evaluated over a historical period includes: Obtain numerical software data of the target software within a historical time period, wherein the software data includes at least one of code modification data, operational data, and test data; The software data is subjected to outlier detection based on distribution hypothesis to obtain a first outlier dataset, and the first outlier dataset is subjected to outlier detection based on machine learning to obtain a second outlier dataset. Obtain the data source of each abnormal data in the first abnormal dataset and the second abnormal dataset; The intelligent analysis model is invoked to optimize the software data based on the first abnormal dataset, the second abnormal dataset, and the data sources of each abnormal dataset, to obtain the quality data.

12. A software quality assessment device, characterized in that, include: The data acquisition unit is used to collect quality data of the target software within a historical time period and extract the numerical characteristics of the quality data. The target software refers to software that has a quality assessment requirement. The indicator prediction unit is used to predict indicators based on the software attributes of the target software, the numerical characteristics of the quality data, and the quality data, so as to obtain reference indicators for quality evaluation of the target software. The indicator selection unit is used to perform correlation analysis on each reference indicator and select quality indicators from the reference indicators based on the correlation analysis results. The correlation between any two quality indicators is less than the correlation threshold. A quality assessment unit is used to configure the expected value range of the quality indicators, and to perform software quality assessment on the target software based on the quality indicators and the expected value range, so as to obtain the quality assessment result of the target software.

13. A computer device, characterized in that, include: A memory, wherein a computer program is stored; A processor for loading the computer program to implement the method as described in any one of claims 1-11.

14. A computer-readable storage medium, characterized in that, The computer storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1-11.

15. A computer program product, characterized in that, The computer program product includes computer instructions, and the processor of the computer device reads the computer instructions and executes the method as described in any one of claims 1-11.

Images