CN115239122A - Method and device recommended by testers of digital grid software project - Google Patents

Abstract

The invention discloses a method and a device for recommending a tester of a digital grid software project. The method includes: determining the type of test tasks according to the test requirements of the software project, including common test tasks and important test tasks; for common test tasks, analyzing and testing The characteristics of the system can effectively analyze the tester's ability and recommend a group of reliable testers; for important test tasks, based on the three-dimensional characteristics of the tester's test environment, test ability and domain knowledge, maximize the tester's error detection probability, The relevance of the test task, the diversity of personnel, and the goal of minimizing the cost of testing are used to effectively recommend testers. The present invention recommends suitable testers for the test tasks, improves the defect detection rate, and shortens the task completion period by combining the types of tasks and the characteristics of testers.

Description

Method and device recommended by testers of digital grid software project

technical field

The invention relates to a method and a device for recommending a tester of a digital power grid software project, which are applied in the field of power grid software testing.

Background technique

At present, an important sign leading the digital transformation of the power grid is the in-depth application of new technologies such as the Internet, artificial intelligence, big data, and the Internet of Things based on cloud platforms. The construction of the digital power grid will make the production and operation of the future power grid highly dependent on networking and informatization. With the in-depth development of power grid informatization construction and the in-depth application of informatization testing, test environments, test types, test problems, and test data all show a trend of diversification and complexity. The traditional testing mode faces great challenges, and the testing of some major software projects usually consumes nearly half of the testing resources (manpower, equipment, etc.), which is unacceptable.

Faced with the challenges that testing resources cannot meet the rapidly growing testing demands, and testing methods cannot quickly respond to the new requirements of business changes, a number of new software testing tools have been developed to conduct more and faster testing, but the development of new Tools require additional investment and often require highly skilled but scarce testers, increasing labor costs. Therefore, much of the expert work in grid informatization has focused on reducing the cost of testing resources for testing tasks. How to maximize the input gain of each tester based on the ability of existing personnel is of great significance to the construction of power grid information software projects.

SUMMARY OF THE INVENTION

Purpose of the invention: The purpose of the present invention is to provide a method and device for recommending testers of digital grid software projects, to maximize the input gain of each tester based on the capabilities of existing personnel, and to effectively recommend reliable testers to complete the test task.

Technical solution: In the first aspect, a method recommended by testers of a digital grid software project includes the following steps:

Determine the type of test tasks according to the test requirements of the software project, including common test tasks and important test tasks;

For common test tasks, the functions to be tested are obtained by analyzing the characteristics of the software system to be tested, the relevant data is extracted from the repository, the dependencies between the functions are calculated based on the extracted data, and a software function dependency tree is constructed. Test parallelism of system functions, the subsystem is a subsystem that can be tested in parallel through different test tasks, establishes a ranking list of suitable candidates based on medal counters, and comprehensively calculates the ranking of each candidate for the system test and Contribution, giving tester recommendation results for each subsystem;

For important test tasks, the software, hardware and environmental attributes that affect the test results are obtained by analyzing the context of the test task operation, and the characteristics of the test environment are constructed; the tester's test capability characteristics are obtained based on the tester's historical test results; based on the power grid test performed on the tester The domain testing experience acquired by the task establishes the domain knowledge characteristics of the testers; based on the characteristics of the test environment, the characteristics of the test ability and the characteristics of the domain knowledge, to maximize the error detection probability of the testers, the correlation with the test task, and the diversity of the testers The target model is established and solved with the goal of improving the performance and minimizing the test cost, and the matching tester is recommended based on the solution result.

Preferably, the test requirements include test volume and completion time. When the corresponding test requirements of a software project are not higher than a preset threshold, it is regarded as a common test task, otherwise, it is regarded as an important test task.

In a second aspect, a device recommended by a tester of a digital grid software project includes:

The test task type determination module determines the test task type according to the test requirements of the software project, including common test tasks and important test tasks;

Common test task recommendation module, for common test tasks, obtain the functions to be tested by analyzing the characteristics of the software system to be tested, extract relevant data from the repository, and calculate the dependencies between functions based on the extracted data to build a software function dependency tree , based on the dependency relationship to identify the test parallelism of the function of the subsystem, which is a subsystem that can be tested in parallel through different test tasks, establish a ranking list of suitable candidates based on the medal counter, and comprehensively calculate each candidate by For the ranking and contribution of system testing, the tester recommendation results of each subsystem are given;

The important test task recommendation module, for important test tasks, obtains the software, hardware and environmental attributes that affect the test results by analyzing the context of the test task operation, and constructs the test environment characteristics; Establish the domain knowledge characteristics of the testers based on the domain testing experience obtained by the testers performing the power grid test tasks; based on the test environment characteristics, test capability characteristics and domain knowledge characteristics, to maximize the error detection probability of the testers and the correlation with the test task. The target model is established and solved with the goal of minimizing the cost of testing, the diversity of testers, and the matching testers are recommended based on the solution results.

In a third aspect, the present invention also provides a computer device, comprising:

one or more processors;

memory; and

One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs when executed by the processors implement the above Steps in the method recommended by digital grid software project testers.

In a fourth aspect, the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the steps of the above-mentioned method recommended by a digital grid software project tester.

Beneficial effect: the present invention maximizes the input gain of each tester in the face of different test tasks. For common test tasks, use the existing project team to check the knowledge coverage of test tasks by testers, and display which group of developers has the highest joint coverage for selection. For large and important testing tasks, by extracting the testing environment, testing capabilities and domain knowledge, an optimization model is established to maximize the tester's error detection probability, the correlation with the test task, the diversity of testers, and minimize the test cost. Effectively recommend testers, improve the defect detection rate, and shorten the task completion cycle. The invention combines the types of tasks and the characteristics of the tester team, recommends suitable testers for the test tasks, detects more software defects with fewer personnel, and effectively improves the overall efficiency of the test.

Description of drawings

Fig. 1 is a flow chart of a method for recommending a digital grid software project tester of the present invention;

Fig. 2 is the flow chart of the personnel recommendation method for common test tasks of the present invention;

FIG. 3 is a flow chart of a method for recommending personnel for important testing tasks according to the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

Referring to Fig. 1, the digital grid software project tester recommendation method of the present invention comprises the following steps:

Step S10: Determine the test task type according to the test requirements of the software project, including common test tasks and important test tasks.

The test task requirements are obtained through the test service management platform. When the system software developed by the manufacturer needs to be tested, the relevant information of the system to be tested will be submitted on the test service management platform, including the test task requirements. In the embodiment of the present invention, the test requirements include the test volume and the completion time. The urgency of the project can be calculated according to the test volume and completion time. Generally speaking, when the corresponding test requirements of a software project and/or are derived based on the test requirements When the urgency is not higher than the preset threshold, it is regarded as a common test task, otherwise it is regarded as an important test task.

According to the embodiment of the present invention, the measurement volume mainly includes project type, project level, project urgency, and the number of cases to be tested, and the project type mainly includes security development projects, test development projects, tool development projects, and other development projects; The level includes general projects, first-class key projects and second-class key projects; project urgency = (deadline for submitting test reports - general test time)/average test cycle, where general test time is normal according to the difficulty of the function point. The supporting test time, such as ordinary functional performance test 30min/h, will take more time for more difficult tests. The average test period is the average of the test periods for the project type; the number of test cases is given according to the total number of requirements. The number of use cases that need to be tested.

According to an embodiment of the present invention, in the process of testing for various project types (that is, security development, test development, tool development and other development projects), the classification of test tasks is mainly through:

For general projects, the urgency of the project is >1, and the number of test cases to be tested is less than 1/2 of the same type of project, it is divided into ordinary test tasks:

For important projects, if the urgency of the project is less than 1, and the number of test cases to be tested is higher than 1/2 of the same type of project, it is an important test task.

By classifying test tasks, testers suitable for corresponding task characteristics can be recommended for different types of tasks, which is helpful to improve the overall test efficiency.

Step S20 , in the face of common test tasks, by effectively analyzing the system change history, functional dependencies, etc. in the multi-version iteration of the test, the most suitable testers are recommended to form the most suitable team in the test task collaboration in terms of the tester's ability and knowledge coverage.

Obtain the functions to be tested by analyzing the characteristics of the software system to be tested, and extract the test data submitted together from the repository. First, apply for test resources, and then calculate the dependencies between the functions to build a software function dependency tree. Based on the dependencies Identify the test parallelism of the functions of the subsystems that can be tested in parallel through different test tasks, build a ranking list of suitable candidates based on the medal counter, and comprehensively calculate the contribution of each candidate to the system test. Rankings and contributions, giving tester recommendation results for each subsystem.

Referring to FIG. 2, in one example, a method for recommending a person to a common test task includes:

Step S21: In the face of a system to be tested in the common test task, relevant data is extracted from the repository by submitting the test requirements. The data here mainly includes the requirements specification of the system, the user operation manual, the environment configuration table, and the function dependency relationship. Analysis table, factory test report, etc., firstly, apply for the resources required for testing according to the information in the environment configuration table. The resources of the equipment mainly include database server and application server; secondly, comprehensively consider the relevant data extracted, and use the test function of the software to be tested as the test function. The root node uses the software function dependency file to analyze and find all the dependencies of the software functions. The dependencies here include third-party software and function calls in software packages to form a software dependency tree. The leaf nodes of the tree represent the dependencies and functions of software functions. The corresponding version and status of the completion. The version here is mainly the number of system changes in the software iteration process, and the status is mainly the test pass status and difficulty level. At the same time, continue to traverse the leaf nodes of the tree through the above data, find the dependencies that depend on the software function, and record them as the child nodes of the current node. Repeat the above steps to find out the software function dependencies of all nodes and sub-nodes until the leaf nodes no longer depend on other functions, forming a complete software function dependency tree with the software function module to be tested as the root node.

Step S22: Check the constructed software dependency tree, and divide the subsystems that can be tested separately based on the test requirements. For one of the subsystems, the test version of the original test system is 1.0, the number of change history is 0, and the difficulty coefficient is 0. 0. For the retested subsystem, check the system version and status in its version iteration. If the number of change history is greater than 1 and its test status fails, its difficulty level is increased by 1;

Step S23: Based on the ranking list of suitable candidates based on the medal counter, by counting the medals for the test system for which the test of the personnel in the test group has been completed, one complete subsystem is tested at one time, and the medal count of the tester is increased by 1, based on the test Personnel perform a ranking of medal counts, match testers through the difficulty level mapping of different subsystems, and give tester recommendations for each subsystem.

In step S30 , in the face of important testing tasks, based on the three-dimensional characteristics of the power grid tester's testing environment, testing capability and domain knowledge, a multi-objective optimization model is established to effectively recommend the tester.

For large-scale testing tasks, a large number of testing resources are used to perform many small tasks through public testing, which can reduce costs and effectively ensure that the tasks are completed within the specified time. However, in the process of allocation, how to optimize the personnel participating in the test needs to comprehensively consider the environment, ability and background skills.

Referring to FIG. 3, in one example, a human recommendation method for important testing tasks includes:

S31, extracting test environment features.

The characteristics of the test environment mainly refer to the components that run the test task in its specific context and affect the test results, mainly including the attributes of the hardware device model, software operating system, and network environment owned by the test workers. These properties can be shared across different test tasks, since they may be required in different projects to reproduce defects in the test application.

The hardware device model is mainly based on the device memory and device storage capacity. When the basic performance of the device is high, it can effectively improve the efficiency of the test; the software operating system is mainly aimed at the system needs of the test environment and the system used for production applications later. , if the operating system is inconsistent, it may lead to inaccuracies based in part on security and performance testing. The network environment is the guarantee for realizing communication. Indicators such as throughput (I/O) and round-trip delay (RTT) have an important impact on the interaction performance between devices. There is no doubt that a good network environment will improve the test efficiency. Therefore, this must be considered. factor.

Among them, for the hardware device model and software operating system, the tester uploads the basic information after registering with the test service management and control platform, and can obtain these basic attribute characteristics. For the network environment, it can be obtained based on the communication test between devices. For example, the network I/O capability can be obtained by dividing the data size of the test-related files transmitted between the devices in the actual test cycle by the upload time, and/or by obtaining the time RTT required by the sender to transmit to the receiver, as the network I/O capability. Environment properties.

S32, extracting test capability features.

A test capability is characterized as the capability abstracted from the tester's historical test results. Although capability is a relatively abstract concept, it can also be reflected from multiple aspects of related results. The present invention uses the following attributes to describe the capabilities of testers, mainly including the number of projects involved, the number of test reports submitted, the number of test reports submitted, and the Number of bug reports, percentage of bug reports submitted, degree to which testers repeat bug reports. A test report is a test result submitted by a test worker after a test task is completed. It contains the report ID, the worker ID (ie who submitted the report), the task ID (ie which task was performed), how the test was executed, and what happened during the test Description of the situation, error labels, and duplicate labels. Specifically, labels are assigned by the project manager to indicate whether a report contains an "error" (ie, a wrong label), and whether the report is a "duplicate" of another report (ie, a duplicate label).

The quantified digital statistics are used as a tester's ability value vector. In one embodiment, the percentage of bug reports submitted and the degree to which testers repeat bug reports are calculated by the following formula:

The percentage of bug reports submitted by testers = the number of bug reports submitted by testers / the number of test reports submitted.

The degree to which testers duplicate bug reports = the duplicate index of testers / the number of bug reports submitted by testers.

S33, extracting domain knowledge features.

The characteristic of domain knowledge refers to the experience of testing information system in the power grid domain obtained by the tester by performing the testing task of the power grid. The applications being tested usually come from different business domains of the power grid, which require testers with knowledge of a specific business domain to better explore and test the business-related functions. The tester's familiar domain knowledge is represented by using "descriptive terms" extracted from the tester's historical submissions and represented as a vector.

In one example, domain knowledge feature descriptive terms are obtained as follows:

A list of descriptive terms is first constructed from all tasks in the training dataset, followed by tokenization and removal of stop words to reduce noise. Terms are sorted according to the reported number of occurrences of a term (ie, document frequency df), and then a certain percentage (eg, 5%) of terms with the lowest document frequency are filtered out. Under the 5% threshold, the recommendation effect can reach a good and stable value. Secondly, by extracting words from the tester's historical submission report, and mapping these words with a list of descriptive terms, that is, a comparison search, so as to obtain descriptive terms that represent the tester's domain knowledge.

S34, establishing a multi-objective optimization model.

In the face of large-scale power grid test tasks entrusted to testers, they need to be completed within the specified time. It is necessary to effectively recommend a group of testers to complete this test task together. Not all testers are equally good at finding bugs. Inappropriate testers may miss bugs or report duplicate bugs. This group of testers depends on each other. Different testers face different levels of defect discovery. Recommending a suitable group of testers for a testing task can be helpful. It is used to detect more software defects with fewer staff, effectively improving the overall efficiency of testing.

Because the purpose of software tester recommendation is to help test as many bugs as possible with fewer testers and complete the testing task. First, the tester with the highest probability of defect detection should be recommended because it can potentially improve defect detection performance. Next, the test tasks for the power grid information system are basically user-driven and more complex, so the correlation between the staff and the task should be considered, and the testers who maximize the expertise related to the test task should be found, because this part of the staff has more Business background knowledge can increase the probability of defect detection. Second, selecting a population of workers with different characteristics will help detect more defects and reduce duplication of reporting, as different workers may explore different areas of the application being tested. The cost of testing should also be considered.

Therefore, in the embodiments of the present invention, the goal is to maximize the defect detection probability of the tester, the correlation with the test task, the diversity of the staff, and to minimize the test cost.

According to an embodiment of the present invention, aiming at the goal of maximizing the defect detection probability, a machine learning model is established to learn the defect detection probability of each tester by determining the features. The feature extraction has an important impact on the identification of the model. According to the exploration and verification of the present invention, the defect detection features include: (1) The general consensus in the industry is that the tester's ability is closely related to the defect detection probability. Relevant attributes are treated as features in machine learning models. (2) The work experience of the tester also has a great influence on the execution of the test work, and the experience is formed through the accumulation of history. Therefore, in the present invention, the test is better simulated by further considering the time-related factors Worker's past experience. Extract its work in the past 2 weeks, 1 month, and 2 months through its ability attributes. In this way, the original one attribute can produce four features in the machine learning model. (3) Another time-related feature is adopted in the machine learning model, that is, the time interval between the tester’s last submission to the test task release, in days. The longer this interval, the less likely the tester will be involved in the task.

For the above features, by using a logistic regression machine learning model. Based on a logistic regression model trained on the training dataset, given a task in the test dataset, the model's probability of defect detection for all candidate testers can be obtained. For a set of candidate test workers, by summing their defect detection probabilities on a given test task, the sum is considered as the defect detection probability for the test task.

The present invention uses a logistic regression algorithm for classification. According to the above description, the feature of defect detection includes 5 feature indexes, which are represented by points in a 5-dimensional space. Through the sigmoid function, each set of input data x ⁽ⁱ⁾ can be mapped to a number between 0 and 1. If the function value is greater than 0.5, it is judged to belong to 1, otherwise it belongs to 0. In addition, the function requires undetermined parameters. By using sample training, this parameter can accurately predict the data in the training set.

According to an embodiment of the present invention, for the relevance objective, the relevance between the candidate testers and the testing task needs to be measured. Relevance is represented by using the similarity between the tester's domain knowledge and the test task. The present invention performs calculation based on the cosine similarity between the descriptive terms of the tester's domain knowledge and the descriptive terms of the test task requirements. A larger similarity value indicates that the tester's domain knowledge is more closely related to the test task.

Given a specific large-scale testing task, to obtain the correlation of a set of candidate testers, first combine the domain knowledge of all selected testers into a unified vector, and then calculate the cosine similarity of the testers based on the requirements of the test task degree vector. Specifically include:

(1) Obtain the test task requirements through the test service management platform, and construct a descriptive term list based on the test task requirements contained in the test task requirements;

(2) Calculate the cosine similarity between the descriptive terms of the candidate tester's domain knowledge and the descriptive terms of the test task requirements; Combine several keywords into a set, calculate the word frequency of the two descriptive terms for the words in this set, generate their respective word frequency vectors, and then calculate the cosine similarity of the two vectors;

(3) Sorting based on the calculated cosine similarity value, the larger the value, the more similar, and the stronger the correlation.

Maximizing tester relevancy according to embodiments of the present invention aims to find testers who are familiar with the testing task, in addition to that, the nature of software testing requires different testers to help explore parts of the application and reduce duplication of testing Report. Therefore, maximizing tester diversity aims to find testers with different backgrounds. Although these two goals seem to conflict with each other, the goal of the present invention is to strike a balance between maximizing relevance and maximizing diversity through a multi-objective optimization framework. To explore attribute diversity, it was measured using a count-based approach and counted how many distinct attribute values appeared in a selected set of testers.

As mentioned above, testers are characterized by three aspects: test environment, competence and domain knowledge. For the capability dimension, it is unreasonable to consider diversity. Therefore, the diversity is calculated based on the other two dimensions. Specifically, for the test environment, by calculating how many different operating systems, network environments, etc. are included in a group of testers. For domain knowledge, by counting how many different terms appear in the worker's domain knowledge. According to the order of magnitude difference between the attributes of the test environment and domain knowledge, the diversity is calculated through the attributes of the test environment and domain knowledge, and finally the final diversity value is obtained by using the weight parameter.

Specifically, in the embodiment of the present invention, the diversity of testers=0.5*the diversity of test environments+0.5*the diversity of domain knowledge attributes. Among them, the diversity of the test environment depends on the difference of the test environment equipped by a group of testers, as long as there are differences in the operating system, network environment, etc., it is a set of different environments. Diversity of test environments = number of distinct environments/number of environments for a set of total testers. Diversity of domain knowledge attributes = how many different terms appear in domain knowledge/total number of domain knowledge terms. It should be understood that the respective weights of 0.5 in the tester diversity calculation formula are only examples, and other weights may be used in other embodiments. The present invention tries different weights, and relatively good and stable performance can be obtained when the weight value is 0.5.

In accordance with embodiments of the present invention, cost is an unavoidable goal when recommending testers for system testing tasks in order to minimize testing costs. The most important cost in testing is the reward to the tester. Assuming that all participating testers are paid equally in order to expedite a task, the cost of a selected set of testers can be measured. Depending on the testing needs, the corresponding number of testers is multiplied by the reward to get the cost of testing.

The final objective function is expressed as:

f(x)=[f ₁ (x), f ₂ (x), f ₃ (x), f ₄ (x)]

where f ₁ (x), f ₂ (x), f ₃ (x) and f ₄ (x) respectively represent the above sub-objective functions, namely to maximize the defect detection probability of testers, the correlation with test tasks, Diversity of staff and minimization of testing costs are the goals.

S35. Solve the multi-objective model, and obtain a recommended tester.

Multi-objective optimization is difficult to achieve the best results for all objectives at the same time. For example, in order to maximize the probability of false detection, more testers may be required to work, thus sacrificing to minimize the cost of testing. The present invention seeks a Pareto frontier solution. The objective is optimized by using the Non-Dominated Sorting Genetic Algorithm-II (NSGA-II). In the tester recommendation scenario, the Pareto frontier represents the best trade-off among the four objectives identified by NSGA-II. The tester can then examine the Pareto front to find the best between a tester choice that balances false detection probability, correlation, diversity, and test cost, or a tester choice that maximizes three of the objectives penalizing the remaining objectives Eclectic.

The realization of the above-mentioned multi-objective optimization mainly includes four steps, namely decoding, initialization, genetic operator and fitness function.

1) Unencode: by encoding each test worker as a binary variable. The value is 1 if a test job is selected; otherwise, the value is zero. Solutions are represented as a vector of binary variables whose length is equal to the number of candidate test workers. The solution space of a test worker's recommendation problem is the set of all possible combinations, whether or not each test worker is selected.

2) Initialization. The initial population is randomly initialized, that is, K solutions (K is the size of the initial population) are randomly selected among all possible solutions, and K is set to 200.

3) Genetic operator. For the evolution of binary coding of solutions, using standard operators, single-point crossover, mutation is used to generate the next generation. With league as a selection operator, where two solutions are chosen randomly, the fitter of the two will survive in the next population.

4) Fitness function. Since the objective of the present invention is to optimize the four considered objectives, each candidate solution is evaluated by the described objective function. For false detection probability, correlation, and diversity, the larger these values, the faster the solution converges, while the test cost benefits from smaller values.

Based on the same inventive concept as the method embodiment, the present invention also provides a device for recommending a digital grid software project tester, the device comprising:

The test task type determination module determines the test task type according to the test requirements of the software project, including common test tasks and important test tasks;

Common test task recommendation module, for common test tasks, obtain the functions to be tested by analyzing the characteristics of the software system to be tested, extract relevant data from the repository, and calculate the dependencies between functions based on the extracted data to build a software function dependency tree , based on the dependency relationship to identify the test parallelism of the function of the subsystem, which is a subsystem that can be tested in parallel through different test tasks, establish a ranking list of suitable candidates based on the medal counter, and comprehensively calculate each candidate by For the ranking and contribution of system testing, the tester recommendation results of each subsystem are given;

The important test task recommendation module, for important test tasks, obtains the software, hardware and environmental attributes that affect the test results by analyzing the context of the test task operation, and constructs the test environment characteristics; Establish the domain knowledge characteristics of the testers based on the domain testing experience obtained by the testers performing the power grid test tasks; based on the test environment characteristics, test capability characteristics and domain knowledge characteristics, to maximize the error detection probability of the testers and the correlation with the test task. The target model is established and solved with the goal of minimizing the cost of testing, the diversity of testers, and the matching testers are recommended based on the solution results.

In an embodiment of the present invention, the test requirements include test volume and completion time, wherein the measurement volume mainly includes project type, project level, project urgency, and the number of cases to be tested, and the test task type determination module specifically includes:

The project urgency determination unit is used to calculate the urgency of the project according to the test volume and completion time. The project types mainly include security development projects, test development projects, tool development projects, and other development projects; the project level includes general development projects. Projects, first-class key projects and second-class key projects; the urgency of the project = (the deadline for submitting the test report - the general test time)/average test cycle, where the general test time is the normal test according to the difficulty of the function point. time, the average test cycle is the average value of the test cycle for the project type; the number of use cases to be tested is the number of use cases that need to be tested based on the total number of requirements;

and a test task classification unit, which is used to classify test tasks. When the corresponding test requirements of a software project and/or the urgency derived based on the test requirements are not higher than the preset threshold, it is regarded as a common test task, otherwise it is regarded as an important test Task. According to the embodiment of the present invention, for various project types, the classification of test tasks is as follows: For general projects, if the urgency of the project is >1, and the number of test cases to be tested is less than 1/2 of the same type of projects, it is classified as general test Task: For important projects, if the urgency of the project is less than 1, and the number of test cases to be tested is higher than 1/2 of the same type of project, it is an important test task.

According to an embodiment of the present invention, the general test task recommendation module includes:

The data extraction unit is used to extract relevant data from the repository based on test requirements, including system requirements specification, user operation manual, environment configuration table, function dependency analysis table, and factory test report; Building a software function dependency tree, including: taking the test function of the software to be tested as the root node, using the software function dependency relationship file to analyze, and finding all the dependencies of the software functions, the dependencies include third-party software, software packages The function calls within the software form a software dependency tree. The leaf nodes of the tree represent the dependencies of the software functions and the corresponding versions and states of the completion of the functions. The versions are mainly the historical number of system changes in the software iteration process, and the state is the test passed. Status and difficulty level;

The data analysis unit is used to divide the subsystems that can be tested separately based on the test requirements according to the built software dependency tree. For one of the subsystems, the test version of the original test system is 1.0, the number of change history is 0, and the difficulty is 0. The coefficient is 0. For the retested subsystem, check the system version and status in its version iteration. If the number of change history is greater than 1 and its test status fails, its difficulty level is increased by 1;

And the personnel recommendation unit, which is used to build a ranking list of suitable candidates based on the medal counter, by counting the medals for the test system for which the personnel tests in the test group have been completed, and passing 1 complete subsystem at one time, the tester's medal count is added. 1. Based on a ranking of the medal counts by the testers, the testers are matched through the difficulty level mapping of different subsystems, and the tester recommendation for each subsystem is given.

In an embodiment of the present invention, the important test task recommendation module includes: a test environment feature extraction unit, a test capability feature extraction unit, a domain knowledge feature extraction unit, a multi-objective optimization model building unit, and a solving unit; wherein the test environment feature extraction unit The unit is used to obtain the software, hardware and environmental attributes that affect the test results by analyzing the context of the test task operation, and construct the test environment characteristics; the test capability feature extraction unit is used to obtain the tester's test capability characteristics based on the tester's historical test results; domain knowledge The feature extraction unit is used to establish the domain knowledge characteristics of the testers based on the domain testing experience obtained by the testers performing the power grid testing tasks; the multi-objective optimization model building unit is used to establish the characteristics of the test environment The target model is established with the goal of maximizing the error detection probability of the tester, the correlation with the test task, the diversity of the tester, and minimizing the test cost; the solving unit is used to solve the target model, and based on the solution result, it recommends matching Testers.

According to an embodiment of the present invention, the characteristics of the test environment include the hardware device model, software operating system and network environment owned by the test staff;

The test capability characteristics of the testers include the number of projects involved, the number of test reports submitted, the number of bug reports submitted, the percentage of bug reports submitted, and the degree to which testers repeat bug reports, wherein,

The percentage of bug reports submitted by testers = the number of bug reports submitted by testers / the number of test reports submitted;

The degree to which testers duplicate bug reports = the duplicate index of testers / the number of bug reports submitted by testers.

In an embodiment of the present invention, the domain knowledge features of the tester include descriptive terms of the tester's domain knowledge, and the acquisition method includes: constructing a descriptive term list according to all tasks in the training data set, performing word segmentation and deleting stop words , sort the entries according to the number of reports that an entry appears, filter out a certain percentage of entries with the lowest document frequency, extract words from the tester's historical submission report, and map these words to the list of descriptive terms, This yields descriptive terms that represent the tester's domain knowledge.

In the embodiment of the present invention, the calculation of the correlation between the tester and the test task includes:

(1) Obtain the test task requirements through the test service management platform, and construct a descriptive term list based on the test task requirements contained in the test task requirements;

(2) Calculate the cosine similarity between the descriptive terms of the candidate tester's domain knowledge and the descriptive terms of the test task requirements; Combine several keywords into a set, calculate the word frequency of the two descriptive terms for the words in this set, generate their respective word frequency vectors, and then calculate the cosine similarity of the two vectors;

(3) Sorting based on the calculated cosine similarity value, the larger the value, the more similar, and the stronger the correlation.

In an embodiment of the present invention, the method for obtaining the error detection probability of the tester includes:

(1) Extracting features related to defect detection, including: all ability-related attributes of the tester; based on the tester's ability attribute extraction of his work in the past 2 weeks, 1 month, 2 months, and the tester's last submission The time interval until the test task is released;

(2) For the extracted features, use a logistic regression model based on training on the training dataset to give the defect detection probabilities of all candidate testers for a task in a given test dataset;

(3) For a set of candidate test workers, by summing their defect detection probabilities on a given test task, the sum is regarded as the defect detection probability of the test task.

In the embodiment of the present invention, the diversity of testers=α*the diversity of the test environment+β*the diversity of domain knowledge attributes. Among them, α and β are the corresponding weights respectively. The diversity of the test environment depends on the difference of the test environment equipped by a group of testers. As long as there are differences in the operating system, network environment, etc., it is a group of different environments. Diversity of test environments = number of distinct environments/number of environments for a set of total testers. Diversity of domain knowledge attributes = how many different terms appear in domain knowledge/total number of domain knowledge terms. The present invention tries different weights, and relatively good and stable performance can be obtained when the weight value is 0.5.

In an embodiment of the present invention, the solving unit optimizes the objective by using the Non-Dominated Sorting Genetic Algorithm-II (NSGA-II). In the tester recommendation scenario, the Pareto frontier represents the best trade-off among the four objectives identified by NSGA-II. Testers can examine the Pareto front to find the best compromise between a tester choice that balances false detection probability, correlation, diversity, and test cost, or a tester choice that maximizes three of the objectives penalizing the remaining objectives.

The present invention also provides a computer device comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the memory Executed by one or more processors, the program, when executed by the processors, implements the steps of the digital grid software project tester-recommended method as described above.

The present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the steps of the above-mentioned method recommended by a digital grid software project tester.

As will be appreciated by those skilled in the art, the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the present invention can still be Modifications or equivalent replacements are made to the specific embodiments of the present invention, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention shall be included within the protection scope of the claims of the present invention.

Claims (10)

1. A recommendation method for a tester of a digital power grid software project is characterized by comprising the following steps:

determining the types of test tasks according to the test requirements of the software project, wherein the types of the test tasks comprise common test tasks and important test tasks;

for a common test task, acquiring functions to be tested by analyzing the characteristics of a software system to be tested, extracting relevant data from a storage library, calculating the dependency relationship among the functions based on the extracted data to construct a software function dependency relationship tree, identifying the test parallelism of subsystem functions based on the dependency relationship, establishing a proper candidate ranking list based on a medal counter for a subsystem capable of performing parallel test through different test tasks, and giving a tester recommendation result of each subsystem by comprehensively calculating the ranking and contribution of each candidate to the system test;

for important test tasks, software, hardware and environmental attributes influencing test results are obtained by analyzing the context of the running of the test tasks, and test environmental characteristics are constructed; acquiring the testing capability characteristics of the tester based on the historical testing result of the tester; establishing field knowledge characteristics of the testers based on field test experience obtained by executing the power grid test tasks on the testers; based on the test environment characteristics, the test capability characteristics and the field knowledge characteristics, an objective model is built and solved with the aim of maximizing the error detection probability of the testers, the correlation with the test tasks, the diversity of the testers and minimizing the test cost, and the matched testers are recommended based on the solution result.

2. The method of claim 1, wherein the testing requirements include a testing amount and a completion time, and the urgency of the item is derived from the testing amount and the completion time, and the corresponding testing requirements and/or the urgency derived based on the testing requirements of the software item are used as a general testing task when not higher than a preset threshold, and are used as an important testing task otherwise.

3. The method of claim 1, wherein the relevant data extracted from the repository includes a requirement specification of the system, a user operation manual, an environment configuration table, a function dependency analysis table, and a factory test report;

the method for constructing the software function dependency relationship tree based on the extracted data and the dependency relationship among the functions comprises the following steps: the method comprises the steps of taking a test function of software to be detected as a root node, analyzing by utilizing a software function dependency relation file, and searching all dependencies of the software function, wherein the dependencies comprise third-party software and function calls in a software package to form a software dependency relation tree, leaf nodes of the tree represent the dependencies of the software function and corresponding versions and states of function completion, the versions mainly comprise system change history numbers in a software iteration process, and the states are test passing states and difficulty levels.

4. The method of claim 1, wherein the test environment characteristics include hardware device model, software operating system, and network environment owned by a test worker;

the testing capability characteristics of the tester include the number of participating items, the number of test reports submitted, the number of error reports submitted, the percentage of error reports submitted, and the degree to which the tester repeats the error reports, wherein,

the percentage of error reports submitted by the tester = number of error reports submitted by the tester/number of test reports submitted;

the degree to which the tester repeats the error report = the repeat index of the tester/the number of error reports submitted by the tester.

5. The method of claim 1, wherein the domain knowledge characteristics of the test person comprise descriptive terms of the domain knowledge of the test person, and the obtaining method comprises: the method comprises the steps of constructing a descriptive term list according to all tasks in a training data set, segmenting words, deleting stop words, sequencing the entries according to the number of reports of the occurrence of one entry, filtering out the entries with a certain proportion of the lowest document frequency, extracting words from historical submission reports of testers, and mapping the words and the descriptive term list to obtain descriptive terms representing the field knowledge of the testers.

6. The method of claim 5, wherein the calculation of the relevance of the test person to the test task comprises:

(1) Acquiring a test task requirement through a test service management platform, and constructing a descriptive term list based on the requirement of a test task contained in the test task requirement;

(2) Calculating cosine similarity between descriptive terms of field knowledge of candidate testers and descriptive terms of requirements of the test tasks; taking out a plurality of keywords from descriptive terms of field knowledge of candidate testing personnel and descriptive terms of testing task requirements respectively, combining the keywords into a set, calculating the word frequency of the descriptive terms of the candidate testing personnel on the words in the set, generating respective word frequency vectors, and calculating to obtain the cosine similarity of the two vectors;

(3) And sorting is carried out based on the calculated cosine similarity, wherein the larger the value is, the more similar the value is, and the stronger the correlation is.

7. The method of claim 1, wherein the method for obtaining the false detection probability of the tester comprises:

(1) Extracting defect detection related features, comprising: all capability-related attributes of the tester; extracting the working conditions of the testers in the past 2 weeks, 1 month and 2 months based on the capability attributes of the testers, and the time interval from the last submission of the testers to the release of the test tasks;

(2) For the extracted features, using a logistic regression model trained on a training data set to give defect detection probabilities for all candidate testers for a task in a given test data set;

(3) For a group of candidate test workers, the sum is considered as the probability of defect detection for a test task by adding their probability of defect detection over a given test task.

8. A digital power grid software project tester recommendation device is characterized by comprising: .

The testing task type determining module is used for determining the type of a testing task according to the testing requirements of the software project, wherein the testing task type comprises a common testing task and an important testing task;

the system comprises a common test task recommending module, a test task recommending module and a data processing module, wherein for a common test task, functions to be tested are obtained by analyzing the characteristics of a software system to be tested, relevant data are extracted from a storage library, a software function dependency relationship tree is constructed by calculating the dependency relationship among the functions based on the extracted data, the test parallelism of subsystem functions is identified based on the dependency relationship, the subsystems are subsystems capable of performing parallel test through different test tasks, a proper candidate ranking list based on a medal counter is established, and the recommendation result of testers of each subsystem is given by comprehensively calculating the ranking and contribution of each candidate to the system test;

the important test task recommending module is used for analyzing the context of the running of the test task to obtain software, hardware and environmental attributes influencing the test result for the important test task and constructing test environmental characteristics; acquiring the testing capability characteristics of the tester based on the historical testing result of the tester; establishing field knowledge characteristics of the testers based on field test experience obtained by executing the power grid test tasks on the testers; based on the test environment characteristics, the test capability characteristics and the field knowledge characteristics, an objective model is built and solved with the aim of maximizing the error detection probability of the testers, the correlation with the test tasks, the diversity of the testers and minimizing the test cost, and the matched testers are recommended based on the solution result.

9. A computer device, comprising:

one or more processors;

a memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs when executed by the processors implement the steps of the digital power grid software project tester recommendation method of any one of claims 1-7.

10. A computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the digital grid software project tester recommendation method according to any one of the claims 1-7.

Images