Low-code intelligent test generation system based on AI

Abstract

The invention relates to the technical field of software testing, in particular to an AI-based low-code intelligent test generation system which comprises a demand analysis module, a test path generation module, a test script conversion module, an execution environment management module and a result analysis module. A test path of the test path generation module is supplied to the test script conversion module and the result analysis module, a test script generated by the test script conversion module is sent to the execution environment management module, and an execution result of the execution environment management module is fed back to the result analysis module; and an optimization suggestion generated by the result analysis module is sent to the test path generation module and the test script conversion module. Through cooperation of multiple modules, full-process automation of low-code testing is achieved, testing paths and scripts are dynamically optimized, testing efficiency and coverage pertinence are improved, manual intervention and the error rate are reduced, and dynamic adaptation and execution stability of a testing scheme are ensured.

Description

Technical Field

[0001] This invention relates to the field of software testing technology, and in particular to an AI-based low-code intelligent test generation system. Background Technology

[0002] Low-code development technology emerged in the early 21st century, evolving from early visual programming tools to modern integrated development platforms. It has gradually enabled application development through drag-and-drop component configuration and parameter setting, significantly lowering the development threshold. With the penetration of artificial intelligence technology, intelligent testing technology has also developed. From traditional manual test case writing to automated testing tools enabling script recording and playback, and further to the introduction of machine learning algorithms to optimize test case generation, a technological trend combining low-code development and intelligent testing has emerged, driving improvements in software testing efficiency.

[0003] Despite the increasing convergence of low-code development and intelligent testing technologies, existing technologies still have limitations in practical applications. During the requirements analysis phase, the semantic understanding of visual components and configuration information in low-code platforms is not precise enough, easily leading to deviations between test objectives and actual requirements. Test path generation often relies on fixed rules or simple coverage models, making it difficult to dynamically adjust priorities based on feature parameters, resulting in unreasonable allocation of test resources. Test script conversion often encounters compatibility issues due to syntax differences between low-code platforms, and there is a lack of effective syntax validation mechanisms. Furthermore, the feedback loop between test execution and result analysis is incomplete, making continuous optimization of test strategies difficult and affecting the comprehensiveness and efficiency of low-code application testing. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned problems and provide an AI-based low-code intelligent test generation system. To achieve the above objective, this invention adopts the following technical solution:

[0005] An AI-based low-code intelligent test generation system includes a requirement parsing module, a test path generation module, a test script conversion module, an execution environment management module, and a result analysis module. The requirement parsing module processes low-code requirements and transmits the results to the test path generation module. The test paths generated by the test path generation module are provided to both the test script conversion module and the result analysis module. The test scripts generated by the test script conversion module are sent to the execution environment management module. The execution results from the execution environment management module are fed back to the result analysis module. Optimization suggestions generated by the result analysis module are sent to both the test path generation module and the test script conversion module.

[0006] Furthermore, the requirement parsing module includes a low-code requirement extraction unit, a semantic mapping unit, and a feature parameter extraction unit. The low-code requirement extraction unit extracts visual components and configuration information from the low-code development environment. The extracted information is converted into a structured requirement description by the semantic mapping unit. The structured requirement description is processed by the feature parameter extraction unit to obtain test-related feature parameters.

[0007] Furthermore, the feature parameter extraction unit uses a feature importance calculation method, with the following formula:

[0008]

[0009] Among them, I k V represents the importance of the k-th feature parameter, m is the number of samples, and V ik Let k be the feature value of the i-th sample. Let σ be the mean of the k-th feature. k Let be the standard deviation of the k-th feature.

[0010] Furthermore, the test path generation module includes a path planning unit, a constraint adaptation unit, and a path priority ranking unit. The path planning unit generates a set of potential test paths based on the processing results of the requirement analysis module. The constraint adaptation unit matches input constraints to each test path. The path priority ranking unit ranks the test paths by combining the path coverage and the importance of the feature parameters obtained by the feature parameter extraction unit.

[0011] Furthermore, the path priority sorting unit uses a priority calculation formula:

[0012]

[0013] Among them, P j Let C be the priority of the j-th test path. j Let I be the coverage of the j-th path. k R represents the importance of the k-th feature parameter. jk Let ω1 and ω2 be the correlation between the j-th path and the k-th feature parameter, where ω1 and ω2 are weight coefficients and ω1 + ω2 = 1.

[0014] Furthermore, the test script conversion module includes a template library calling unit, a parameter filling unit, and a script syntax verification unit. The template library calling unit calls the corresponding low-code test script template according to the test path type generated by the test path generation module. The parameter filling unit fills the constraints and input parameters into the test script template. The script syntax verification unit performs syntax compliance verification on the filled script.

[0015] Furthermore, the execution environment management module includes an environment configuration unit, a resource scheduling unit, and a script execution unit. The environment configuration unit configures a low-code runtime environment according to the test script requirements generated by the test script conversion module. The resource scheduling unit allocates computing and storage resources to the runtime environment. The script execution unit executes the test scripts according to the priority order determined by the test path generation module and records the execution status.

[0016] Furthermore, the result analysis module includes an execution result acquisition unit, an anomaly location unit, and an optimization suggestion generation unit. The execution result acquisition unit collects the execution results and performance indicators of the script execution unit in the execution environment management module. The anomaly location unit locates components or logic nodes with anomalies in the low-level code based on the execution results. The optimization suggestion generation unit generates adjustment suggestions for the test path generation module and the test script conversion module based on the location results of the anomaly location unit.

[0017] Furthermore, the anomaly localization unit adopts an anomaly correlation degree calculation method, the formula of which is:

[0018]

[0019] Among them, L p Let N be the degree of abnormal correlation of the p-th component. p N represents the number of exceptions involving the p-th component. t T represents the total number of anomalies. p Let be the execution time of the p-th component, s be the total number of components, and T be the execution time of the p-th component. q Let q be the execution time of the q-th component.

[0020] Furthermore, the script syntax verification unit uses a syntax matching degree calculation method, with the following formula:

[0021]

[0022] Where M is the syntax matching degree, M c To verify the number of valid syntax rules, M t This represents the total number of syntax rules.

[0023] The advantages of this invention are:

[0024] 1. This invention achieves full automation of low-code testing from requirement analysis to execution optimization through closed-loop collaboration of the requirement analysis module, test path generation module, test script conversion module, execution environment management module, and result analysis module. Combined with the optimization suggestions generated by the result analysis module, the test path and script are dynamically adjusted. This reduces manual intervention, improves test iteration efficiency, and ensures dynamic adaptation of test plans to low-code requirements.

[0025] 2. This invention uses a feature parameter extraction unit to calculate feature importance using a formula, and combines it with a path priority ranking unit based on a quantitative ranking model of coverage and parameter importance. This enables precise screening and dynamic priority adjustment of test paths, reduces redundant test paths, prioritizes test scenarios corresponding to high-importance features, and improves the rationality of test resource allocation and the relevance of test coverage.

[0026] 3. This invention achieves automated and compliant generation and efficient execution of low-code test scripts by combining template calling, parameter filling, and matching degree calculation of the syntax verification unit in the test script conversion module with resource scheduling and priority execution mechanism in the execution environment management module. This reduces the script syntax error rate, ensures the compatibility between the test environment and script requirements, and improves the stability and efficiency of test execution. Attached Figure Description

[0027] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application.

[0028] In the attached diagram:

[0029] Figure 1 This is a system framework diagram of an AI-based low-code intelligent test generation system in Example 1. Detailed Implementation

[0030] The present invention will now be described in detail and specifically through specific embodiments to enable a better understanding of the invention. However, the following embodiments do not limit the scope of protection of the present invention.

[0031] Example 1

[0032] like Figure 1 As shown, an AI-based low-code intelligent test generation system includes a requirement parsing module, a test path generation module, a test script conversion module, an execution environment management module, and a result analysis module. The requirement parsing module processes low-code requirements and then transmits the results to the test path generation module. The test paths generated by the test path generation module are provided to both the test script conversion module and the result analysis module. The test scripts generated by the test script conversion module are sent to the execution environment management module. The execution results from the execution environment management module are fed back to the result analysis module. Optimization suggestions generated by the result analysis module are sent to both the test path generation module and the test script conversion module.

[0033] In a specific embodiment, the system is deployed on a server equipped with an Intel Xeon E5-2690 v4 processor and runs CentOS 7.9. The Mendix 9.18 low-code platform is used as the development environment. The requirements analysis module uses a Python 3.9 script to scan the .mpr files of the Mendix project to extract requirements. The test path generation module calls the LSTM model trained with the TensorFlow 2.8 library to generate paths. The test script conversion module converts the paths into Selenium 4.4 scripts and outputs them to the VMware ESXi 7.0 virtual machine of the execution environment management module. The result analysis module generates optimization suggestions using a Java 11 program and sends them to both the test path generation module and the test script conversion module.

[0034] Furthermore, the requirement parsing module includes a low-code requirement extraction unit, a semantic mapping unit, and a feature parameter extraction unit. The low-code requirement extraction unit extracts visual components and configuration information from the low-code development environment. The extracted information is converted into a structured requirement description by the semantic mapping unit. The structured requirement description is processed by the feature parameter extraction unit to obtain test-related feature parameters.

[0035] In a specific embodiment, the low-code requirement extraction unit reads the dataGrid component of the "Customer Information Management" application within the platform using the Mendix SDK and extracts its configuration information "maxRows=100" and "editable=true". The semantic mapping unit converts this information into a structured requirement description that "the customer information table can display a maximum of 100 rows of data and supports editing". The feature parameter extraction unit extracts three feature parameters from this description: "maximum number of rows", "editable status", and "data field type", where "maximum number of rows" is 100 and "editable status" is a boolean value.

[0036] Furthermore, the feature parameter extraction unit uses a feature importance calculation method, with the following formula:

[0037]

[0038] Among them, I k V represents the importance of the k-th feature parameter, m is the number of samples, and V ik Let k be the feature value of the i-th sample. Let σ be the mean of the k-th feature. k Let be the standard deviation of the k-th feature.

[0039] In a specific embodiment, the feature parameter extraction unit calculates the importance of "maximum number of rows": selects m=5 sample values ​​of 20, 50, 80, 100, and 150 respectively, and calculates the mean. Standard deviation σ k =47.43. Substitute into the formula:

[0040] Furthermore, the test path generation module includes a path planning unit, a constraint adaptation unit, and a path priority ranking unit. The path planning unit generates a set of potential test paths based on the processing results of the requirement analysis module. The constraint adaptation unit matches input constraints to each test path. The path priority ranking unit ranks the test paths by combining the path coverage and the importance of the feature parameters obtained by the feature parameter extraction unit.

[0041] In a specific embodiment, the path planning unit generates three test paths based on the business process of the "Customer Information Management" application: "User Login → Data Entry → Table Display", "User Login → Data Filtering → Table Display", and "User Login → Data Deletion → Table Display". The constraint adaptation unit matches the first path with the constraints "username length ≥ 6" and "entered data field not empty" and matches the second path with the constraint "filtering condition is text type". The path priority sorting unit calls the feature parameter importance calculated in claim 3 to sort the three paths.

[0042] Furthermore, the path priority sorting unit uses a priority calculation formula:

[0043]

[0044] Among them, P j Let C be the priority of the j-th test path. j Let I be the coverage of the j-th path. k R represents the importance of the k-th feature parameter. jk Let ω1 and ω2 be the correlation between the j-th path and the k-th feature parameter, where ω1 and ω2 are weight coefficients and ω1 + ω2 = 1.

[0045] In a specific embodiment, the priority of the path "user login → data entry → table display" is calculated as follows: C j =0.8 (covering 80% of business nodes), I k The maximum number of rows is 0.76, and the editable state is 0.62. jk The values ​​are 0.9 and 0.7 respectively, ω1 = 0.5, ω2 = 0.5. Substituting into the formula: P j =0.5×0.8+0.5×(0.76×0.9+0.62×0.7)=0.4+0.5×(0.684+0.434)=0.4+0.5×1.118=0.4+0.559=1.0 (after normalization).

[0046] Furthermore, the test script conversion module includes a template library calling unit, a parameter filling unit, and a script syntax verification unit. The template library calling unit calls the corresponding low-code test script template according to the test path type generated by the test path generation module. The parameter filling unit fills the constraints and input parameters into the test script template. The script syntax verification unit performs syntax compliance verification on the filled script.

[0047] In a specific embodiment, the template library calling unit calls the "Selenium login test template" from the MySQL 8.0 database containing 100 preset templates according to the "user login" path type; the parameter filling unit fills in "username=test001", "password=Abc123" and "login URL" into the template; the script syntax verification unit checks 10 rules such as "element positioning syntax" and "assertion statement format", of which 8 are passed, and the calculated M=8 / 10=0.8, and the script is judged to be qualified.

[0048] Furthermore, the execution environment management module includes an environment configuration unit, a resource scheduling unit, and a script execution unit. The environment configuration unit configures a low-code runtime environment according to the test script requirements generated by the test script conversion module. The resource scheduling unit allocates computing and storage resources to the runtime environment. The script execution unit executes the test scripts according to the priority order determined by the test path generation module and records the execution status.

[0049] In a specific embodiment, the environment configuration unit creates a container containing a Chrome 112 browser and Python 3.9 based on Docker 20.10 and configures an Ubuntu 20.04 system environment; the resource scheduling unit allocates a 2-core Intel Xeon E5-2690v4 CPU and 4GB DDR4 memory to the container through Kubernetes 1.24; the script execution unit executes the script first according to the path with priority 1.0, records the status every 30 seconds during the execution process, and finally records "execution successful, time taken 45 seconds".

[0050] Furthermore, the result analysis module includes an execution result acquisition unit, an anomaly location unit, and an optimization suggestion generation unit. The execution result acquisition unit collects the execution results and performance indicators of the script execution unit in the execution environment management module. The anomaly location unit locates components or logic nodes with anomalies in the low-level code based on the execution results. The optimization suggestion generation unit generates adjustment suggestions for the test path generation module and the test script conversion module based on the location results of the anomaly location unit.

[0051] In a specific embodiment, the execution result acquisition unit collects the "user login timeout" result through Jmeter 5.5, with a response time of 2.3 seconds and an error code of "504". The anomaly location unit compares the normal execution log and locates the anomaly in the token generation function of the "login verification component". The optimization suggestion generation unit generates adjustment suggestions based on the result, such as "add login timeout parameters (3 seconds, 5 seconds) test" and "check the timeout handling logic of the token generation function".

[0052] Furthermore, the anomaly localization unit adopts an anomaly correlation degree calculation method, the formula of which is:

[0053]

[0054] Among them, L p Let N be the degree of abnormal correlation of the p-th component. p N represents the number of exceptions involving the p-th component. t T represents the total number of anomalies. p Let be the execution time of the p-th component, s be the total number of components, and T be the execution time of the p-th component. q Let q be the execution time of the q-th component.

[0055] In a specific embodiment, the anomaly localization unit statistically obtains N. p =3 (number of login verification component errors), N t =5 (total number of anomalies), T p =15 seconds (duration of this component's execution) (Total duration of 5 components), substitute into the formula:

[0056] Furthermore, the script syntax verification unit uses a syntax matching degree calculation method, with the following formula:

[0057]

[0058] Where M is the syntax matching degree, M c To verify the number of valid syntax rules, M t This represents the total number of syntax rules.

[0059] In a specific embodiment, when testing the "Data Entry" script, the script syntax verification unit checks 10 rules, including "Form Submission Syntax" and "Data Type Conversion". c =9,M t =10. Substitute into the formula: The script's syntax matching was rated as excellent.

[0060] The specific embodiments of the present invention have been described in detail above, but they are merely examples, and the present invention is not equivalent to the specific embodiments described above. For those skilled in the art, any equivalent modifications and substitutions to the present invention are also within the scope of the present invention. Therefore, all equivalent transformations and modifications made without departing from the spirit and scope of the present invention should be covered within the scope of the present invention.

Claims

1. An AI-based low-code intelligent test generation system, characterized by, The system comprises a requirement analysis module, a test path generation module, a test script conversion module, an execution environment management module and a result analysis module, the requirement analysis module processes low-code requirements and transmits the results to the test path generation module, the test path generated by the test path generation module is provided to the test script conversion module and the result analysis module, the test script generated by the test script conversion module is sent to the execution environment management module, the execution result of the execution environment management module is fed back to the result analysis module, and the optimization suggestions generated by the result analysis module are sent to the test path generation module and the test script conversion module.

2. The AI-based low-code intelligent test generation system of claim 1, wherein, The requirement analysis module comprises a low-code requirement extraction unit, a semantic mapping unit and a feature parameter extraction unit, the low-code requirement extraction unit extracts visual components and configuration information from a low-code development environment, the extracted information is converted into a structured requirement description by the semantic mapping unit, and the structured requirement description is processed by the feature parameter extraction unit to obtain test-related feature parameters.

3. The AI-based low-code intelligent test generation system of claim 2, wherein, The feature parameter extraction unit adopts a feature importance calculation method, and the formula is: where I k is the importance of the kth feature parameter, m is the number of samples, V ik is the kth feature value of the ith sample, is the mean of the kth feature, σ k is the standard deviation of the kth feature.

4. The AI-based low-code intelligent test generation system of claim 3, wherein, The test path generation module comprises a path planning unit, a constraint condition adaptation unit and a path priority sorting unit, the path planning unit generates a set of potential test paths based on the processing result of the requirement analysis module, the constraint condition adaptation unit matches input constraint conditions for each test path, and the path priority sorting unit sorts the test paths in combination with the path coverage range and the feature parameter importance obtained by the feature parameter extraction unit.

5. The AI-based low-code intelligent test generation system of claim 4, wherein, The path priority sorting unit adopts a priority calculation formula: wherein P j is the priority of the jth test path, C j is the coverage of the jth path, I k is the importance of the kth feature parameter, R jk is the relevance of the jth path to the kth feature parameter, and ω1and ω2are weight coefficients and ω1+ ω2= 1.

6. The AI-based low-code intelligent test generation system of claim 5, wherein, The test script conversion module comprises a template library calling unit, a parameter filling unit and a script syntax verification unit, the template library calling unit calls a corresponding low-code test script template according to the test path type generated by the test path generation module, the parameter filling unit fills the constraint conditions and input parameters into the test script template, and the script syntax verification unit performs syntax compliance verification on the filled script.

7. The AI-based low-code intelligent test generation system of claim 6, wherein, The execution environment management module comprises an environment configuration unit, a resource scheduling unit and a script execution unit, the environment configuration unit configures a low-code running environment according to the test script generated by the test script conversion module, the resource scheduling unit allocates computing resources and storage resources for the running environment, and the script execution unit executes the test script in the priority order determined by the test path generation module and records the execution state.

8. The AI-based low-code intelligent test generation system of claim 7, wherein, The result analysis module comprises an execution result collection unit, an exception positioning unit and an optimization suggestion generation unit, the execution result collection unit collects the execution results and performance indicators of the script execution unit in the execution environment management module, the exception positioning unit locates the abnormal components or logic nodes in the low code according to the execution results, and the optimization suggestion generation unit generates adjustment suggestions for the test path generation module and the test script conversion module based on the positioning results of the exception positioning unit.

9. The AI-based low-code intelligent test generation system of claim 8, wherein, The exception positioning unit adopts an exception correlation degree calculation method, and the formula is: wherein, L p is the abnormal correlation degree of the pth component, N p is the number of abnormalities involving the pth component, N t is the total number of abnormalities, T p is the execution duration of the pth component, s is the total number of components, T q is the execution duration of the qth component.

10. The AI-based low-code intelligent test generation system of claim 9, wherein, The script syntax checking unit adopts a syntax matching degree calculation method, and the formula is: Wherein, M is the syntax matching degree, M c is the number of syntax rules passed by verification, M t is the total number of syntax rules.

Images