CN116594913B

CN116594913B - Intelligent software automatic test method

Info

Publication number: CN116594913B
Application number: CN202310867520.1A
Authority: CN
Inventors: 孙天岳; 彭鑫
Original assignee: Qingdao University
Current assignee: Qingdao University
Priority date: 2023-07-15
Filing date: 2023-07-15
Publication date: 2023-09-19
Anticipated expiration: 2043-07-15
Also published as: CN116594913A

Abstract

The invention relates to the technical field of software testing, in particular to an intelligent software automatic testing method. The method comprises the following steps: constructing and configuring a software testing environment through an automatic script and a cloud service technology, so as to acquire testing environment index data; performing depth prediction calculation on the test environment index data so as to obtain reference test case data; monitoring deployment and monitoring data acquisition are carried out on the edge computing nodes according to the reference test case data, so that real-time monitoring data are obtained; performing enhanced dynamic adjustment according to the real-time monitoring data so as to obtain adjustment strategy data; generating multi-mode test data through cloud service calculation according to the adjustment strategy data, so as to obtain the multi-mode test data; and performing self-adaptive feedback optimization according to the real-time monitoring data and the multi-mode test data, thereby obtaining intelligent software test feedback optimization data. The invention can improve the test efficiency and quality and enhance the stability and adaptability of software.

Description

Intelligent software automatic test method

Technical Field

The invention relates to the technical field of software testing, in particular to an intelligent software automatic testing method.

Background

Software testing is a process that is performed to discover errors and problems in the software development process. Its objective is to verify and confirm whether the software product meets the prescribed requirements and is able to function properly under the predetermined conditions and circumstances. The test procedure typically includes the steps of designing test cases, performing tests, finding and reporting errors, and tracking and managing errors.

Intelligent software automation testing refers to a method that utilizes Artificial Intelligence (AI) technology to improve the efficiency and quality of software testing. The method can automatically generate test cases, automatically execute tests, and automatically analyze and solve problems. It generally includes AI technology using deep learning, machine learning, pattern recognition, natural language processing, and advanced computing technology using cloud computing and edge computing. Existing intelligent software automated testing typically requires a significant amount of computing and memory resources, and thus may increase the consumption of hardware devices and energy.

Disclosure of Invention

The invention provides an intelligent software automatic test method for solving at least one technical problem.

The application provides an intelligent software automatic test method, which comprises the following steps:

step S1: constructing and configuring a software testing environment through an automatic script and a cloud service technology, so as to acquire testing environment index data;

step S2: performing depth prediction calculation on the test environment index data so as to obtain reference test case data;

step S3: monitoring deployment and monitoring data acquisition are carried out on the edge computing nodes according to the reference test case data, so that real-time monitoring data are obtained;

step S4: performing enhanced dynamic adjustment according to the real-time monitoring data so as to obtain adjustment strategy data;

step S5: generating multi-mode test data through cloud service calculation according to the adjustment strategy data, so as to obtain the multi-mode test data;

step S6: and performing self-adaptive feedback optimization according to the real-time monitoring data and the multi-mode test data, thereby obtaining intelligent software test feedback optimization data.

The application uses the automatic script and cloud service technology, and can quickly construct and configure the test environment, thereby greatly shortening the time for preparing the test environment. In addition, the efficiency of test execution and optimization can be further improved through the self-adaptive feedback optimization of real-time monitoring data and multi-mode test data. The method not only can obtain the reference test case data through the depth prediction calculation, but also can carry out the enhanced dynamic adjustment according to the real-time monitoring data, and generate the multi-mode test data through the cloud service calculation, thereby comprehensively detecting and verifying the performance and the function of the software and improving the comprehensiveness and the accuracy of the test. By means of real-time monitoring and enhanced dynamic adjustment, the method can find and solve the problem in the first time, and stability and reliability of software are enhanced. The self-adaptive feedback optimization can be performed according to the real-time monitoring data and the multi-mode testing data, so that the software can be better adapted to various environments and conditions, and the adaptability and the flexibility of the software are improved. By means of an automatic test mode, time and energy of manual test can be reduced, and manpower resources are saved. By using cloud services and edge computing nodes, optimized resource allocation and management can be realized, so that computing and storage resources are fully utilized.

Preferably, step S1 is specifically:

step S11: acquiring intelligent software testing requirement data;

step S12: performing environment demand assessment on intelligent software test demand data so as to obtain test environment assessment data;

step S13: generating resource configuration according to the test environment evaluation data, thereby obtaining test resource configuration strategy data;

step S14: performing test environment configuration deployment through a preset cloud service test environment configuration script by using test resource configuration strategy data, so as to obtain configuration test script data;

step S15: performing test environment verification by using configuration test script data so as to obtain test environment verification report data;

step S16: generating test environment monitoring strategy data according to the test environment verification report data;

step S17: and generating the test environment index according to the test environment monitoring strategy data, thereby acquiring the test environment index data.

According to the invention, the software testing environment is constructed and configured through the automatic script and the cloud service technology, so that the time and the workload for manually configuring and deploying the testing environment are greatly reduced, and the testing efficiency is improved. The accuracy and stability of the test environment can be ensured through the steps of depth prediction calculation, monitoring data acquisition, dynamic adjustment, multi-mode test data generation and self-adaptive feedback optimization, so that the reliability of the test result is ensured. The real-time monitoring data and the multi-mode testing data are considered, so that the testing process can be dynamically adjusted and optimized according to the real-time software running condition, and the flexibility of the testing process is improved. The method takes a series of complete steps from acquisition of test requirements, evaluation of environment requirements, resource allocation, deployment of environment configuration, environment verification and finally generation of an environment monitoring strategy into consideration, so that a complete software test environment configuration flow is formed, and the test integrity and accuracy are improved. The most advanced artificial intelligence technology and the Internet technology are utilized to conduct real-time monitoring and data acquisition, the testing process is dynamically adjusted and optimized, and the intellectualization of the testing process is achieved.

Preferably, in step S15, the test environment verification is quantitatively verified by a test environment verification calculation formula, where the test environment verification calculation formula specifically is:

；

;

verifying reporting data for test environment, < >>For testing environmental parameter items->For testing environmental safety parameter items, +.>For testing environmental performance parameter items +_>Is a base constant term,/->For configuring the data format of the test script data +.>For order item, ++>Is->Individual test environment verification items,/->For the first weight index +.>Is->Individual test environmental parameter items,/->For the second weight index->Is->A test environmental security parameter item, +.>For the third weight index +.>Is->And testing environmental performance parameter items.

The present invention constructs a test environment verification calculation formula that quantitatively evaluates a test environmentVerifying the result, thereby improving the test efficiency and quality; considering a plurality of parameters and indexes of the test environment, thereby improving the comprehensiveness and accuracy of the test; the constant parameters and the weight indexes are adjusted, so that different test requirements and scenes can be flexibly adapted; by logarithmic functionReflecting the complexity and difficulty of the test environment verification result; by limit function- >Reflecting the stability and convergence of the test environment verification result; by trigonometric function->、/>And +.>Reflecting the periodicity and volatility of the test environment verification result; by derivative function->Reflecting the change rate and sensitivity of the test environment verification result; />Is a test environment parameter item, such as CPU usage rate and memory usage rate, which affects the basic configuration and running state of the test environment; />The security parameter items of the test environment, such as encryption mode and firewall setting, influence the confidentiality and reliability of the test environment; />Is a test environment performance parameter item, such as response time, throughput, affecting the efficiency and load capacity of the test environment; />Is a base constant term, and affects the growth speed and curve shape of the logarithmic function; />The method is to configure the data format of test script data, and influence the calculation precision and range of the limit function; derivative function->By->And->Interaction, mean->When changing, the value of the derivative function reflects the change rate of the logarithmic function; the present invention provides an accurate and flexible computing method for verification and evaluation of software testing environments to provide reliable data support.

Preferably, step S2 is specifically:

Step S21: carrying out data normalization processing on the test environment index data so as to obtain the test environment index normalization data;

step S22: feature selection is carried out on the normalized data of the test environment indexes, so that the selection data of the test environment indexes are obtained;

step S23: constructing a neural network model for the test environment index selection data, so as to construct test environment index prediction model data;

step S24: checking and calculating the test environment index prediction model data so as to obtain test model verification report data;

step S25: and carrying out prediction calculation on the test environment index prediction model data by using the test model verification report data, thereby obtaining the reference test case data.

Through normalization processing and feature selection, the invention can enable the understanding of the test environment to be more accurate and deep, which is helpful for improving the accuracy of the test. By using the neural network model to predict, the user can know which test cases are likely to fail in advance, so that the test can be performed more specifically, and the test efficiency is improved. By predicting the neural network model, more comprehensive test cases can be generated, so that the test coverage is improved, and the fact that any possible errors are not missed is ensured. The test model verification report can provide model prediction accuracy and effect for us, so that the model is optimized and adjusted, and the test quality is improved.

Preferably, the test environment index selection data includes first test environment index selection data and second test environment index selection data, and step S22 specifically includes:

performing first feature selection on the normalized test environment index data, thereby obtaining first test environment index selection data;

and performing second characteristic selection on the normalized test environment index data, thereby obtaining second test environment index selection data.

According to the invention, through carrying out two different feature selections on the same normalized data, various characteristics of a test environment can be more comprehensively considered and utilized, and the coverage and accuracy of the test are further enhanced. Through twice feature selection, the most critical and representative features can be selected according to different test requirements and targets, and the precision and quality of the test case are improved. By selecting proper characteristics for testing, unnecessary test cases can be reduced, resource waste is avoided, and testing efficiency is improved. The first test environment index selection data and the second test environment index selection data are respectively acquired, diversified input is provided for subsequent neural network model construction, more possible software problems can be captured, and accordingly the comprehensiveness and the deep property of the test are enhanced.

Preferably, the test calculation in step S24 is performed by a test model test calculation formula, wherein the test model test calculation formula is specifically:

;

validating report data for a test model,/->For the first test weight item, +.>For testing environmental index deviation term, < ->For error checking weight term, +.>For the second test weight item, +.>To adjust the proportional check weight term, +.>Predicting the total number of data of the model data for the test environmental indicator, < >>For the third test weight item, +.>For testing environmental index constant term, < ->For periodic adjustment of weight index +.>For the fourth test weight item, +.>For test time data, ++>For power control item->For testing the predicted value of the environmental index, < >>The real value of the environmental index is tested.

The invention constructs a test model test calculation formula which comprises various error test methods, such as proportional-based error test and time-based dynamic error test, and can ensure comprehensive evaluation and test of the prediction capability of the model from different angles and different levels. The various weight and adjustment terms in the formula allow for personalized adjustments to specific test environments and specific predictive tasks, which enhances the flexibility and adaptability of the formula in coping with different test tasks. First check weight term And->Indicating that the effect of model prediction error is adjusted based on the square root relationship; likewise, the second test weight term->And->Indicating that another influence factor is adjusted based on the logarithmic relationship. Furthermore, the->The effect of (2) is to indicate the time factor in the test procedure, whereas +.>Then it is a power control term that can be influenced by adjusting itTo->Is a function of the degree of influence of (a). According to the invention, different weight items, adjustment proportion and test time factors are introduced, so that the accuracy, deviation and stability of the model are better evaluated.

Preferably, step S3 is specifically:

step S31: acquiring edge computing node data, and carrying out edge node optimal evaluation on the edge computing node data by utilizing the reference test case data so as to acquire edge node evaluation data;

step S32: performing test monitoring strategy generation according to the edge node evaluation data so as to obtain edge test monitoring strategy data;

step S33: monitoring deployment is carried out on the edge computing nodes according to the edge test monitoring policy data, so that edge node deployment data are obtained;

step S34: generating a monitoring strategy according to the edge node deployment data, thereby acquiring monitoring data acquisition strategy data;

Step S35: and acquiring the monitoring data according to the monitoring data acquisition strategy data, thereby acquiring real-time monitoring data.

According to the method, the edge computing node data are acquired and evaluated, so that the advantages of edge computing, such as high data processing speed and low delay, can be better utilized. This has significant advantages for software testing tasks that require real-time feedback. The method not only monitors and deploys on the edge computing nodes, but also can generate and implement a monitoring data acquisition strategy, which is beneficial to acquiring and processing monitoring data in real time and improving the monitoring efficiency. According to the method, corresponding test monitoring strategies and monitoring data acquisition strategies can be generated according to the edge node evaluation data and the edge node deployment data, so that the test process can be flexibly adjusted according to actual conditions, and the pertinence and the effectiveness of the test are enhanced. By carrying out the optimal evaluation on the edge computing nodes, the method can effectively find the nodes which are most suitable for testing, optimize the use of resources and save the cost. Through collecting real-time monitoring data, problems in the testing process can be found and adjusted in time, and the testing efficiency and the testing effect are improved.

Preferably, step S4 is specifically:

Step S41: performing data preprocessing according to the real-time monitoring data, thereby obtaining preprocessed real-time monitoring data;

step S42: training the reinforcement model of the preprocessed real-time monitoring data through a preset reinforcement learning algorithm, so as to acquire a real-time monitoring reinforcement model;

step S43: performing optimization test adjustment on the real-time monitoring reinforcement model so as to obtain simulation test report data;

step S44: performing strategy dynamic adjustment according to the simulation test report data so as to obtain dynamic adjustment strategy data;

step S45: performing real-time verification according to the dynamic adjustment strategy data, thereby obtaining real-time strategy verification report data;

step S46: and carrying out strategy optimization on the dynamic adjustment strategy data by utilizing the real-time strategy verification report data so as to acquire the adjustment strategy data.

The preprocessing method of the step S41 can reduce insignificant noise and redundant information and optimize the format and structure of data, thereby providing high-quality data for subsequent analysis and study. In step S42, model training is performed using a preset reinforcement learning algorithm. Reinforcement learning is an algorithm that learns by exploring the environment and obtaining feedback, which can effectively address the problem of high dynamics and uncertainty. By simulating test report data, strategy dynamic adjustment can be performed, which means that our test strategy can be adjusted according to real-time conditions, and the adaptability and flexibility of the test are improved. In step S45 and step S46, our test strategy can be continuously revised and improved through real-time verification and strategy optimization, so as to obtain better performance in the actual environment.

Preferably, step S5 is specifically:

step S51: performing policy analysis on the adjustment policy data so as to obtain policy analysis data;

step S52: performing data definition generation according to the strategy analysis data so as to obtain modal definition data;

step S53: performing edge node database configuration according to the mode definition data so as to obtain data source configuration report data;

step S54: and generating multi-mode test data according to the data source configuration report data, so as to obtain the multi-mode test data.

The policy resolution in step S51 in the present invention helps to understand and interpret the dynamic adjustment policies, thereby providing key information on how to optimize and configure the data. By generating modality-defining data, it is possible to accurately determine which types of data need to be collected and organize the data together in step S52, which may provide the possibility for more complex multi-modality testing. In step S53, the edge node database configuration may ensure that the storage and processing of data can take place closer to the data source, thereby reducing the delay and cost of data transfer. In step S54, multi-modal test data is generated that allows for simultaneous consideration and utilization of data of multiple modalities, such as text, images, and sounds, for a more comprehensive and thorough understanding during the test.

Preferably, step S6 is specifically:

step S61: performing feature extraction according to the real-time monitoring data and the multi-mode test data, so as to obtain feedback feature data;

step S62: performing self-adaptive optimization according to the feedback characteristic data, so as to obtain optimization target data;

step S63: generating an optimization strategy according to the optimization target data, so as to obtain optimization strategy data;

step S64: and implementing and evaluating according to the optimized measurement data, thereby obtaining intelligent software test feedback optimized data.

According to the method, through feature extraction and self-adaptive optimization of the real-time monitoring data and the multi-mode testing data, various conditions in the testing process and the characteristics and requirements of each specific application or environment can be deeply understood, so that more accurate and personalized optimization is realized. The method can generate the optimization strategy and also can evaluate and feed back in real time, which means that the method can dynamically adjust and optimize the test strategy according to the real-time test condition and result, so that the test process is more flexible and efficient. By setting the optimization target data and generating the optimization strategy, the optimization target of the test is determined and tracked more accurately, so that higher test quality and effect are realized. Through real-time evaluation and feedback, continuous learning and improvement are performed, the efficiency and effect of the test are continuously improved, and the processing capacity of various problems and challenges is improved.

The invention has the beneficial effects that: the invention adopts an automatic script, a cloud service technology and edge calculation, automatically creates and configures a test environment without manual intervention, generates a test case, collects and processes monitoring data, generates multi-mode test data and performs test optimization. The labor cost is reduced, and the testing efficiency and consistency are improved. Through depth prediction calculation and reinforcement learning, possible problems and challenges are predicted, and a test strategy is dynamically adjusted to optimize a test effect so as to discover and repair the problems earlier, thereby improving the quality and reliability of software. The multi-mode test data generation is supported, and the data of multiple modes, such as text, images and sound, can be considered and utilized simultaneously, so that more comprehensive and deep understanding is obtained in the test process, and the comprehensiveness and accuracy of the test are improved. The self-adaptive feedback optimization is carried out through the real-time monitoring data and the multi-mode test data, so that the problems in the test process can be timely found and solved, the test strategy and environment are optimized, and the test efficiency and effect are further improved. By utilizing cloud service and edge calculation, the elastic expansion of resources can be realized, and various test requirements can be met. Meanwhile, by carrying out data processing at a place closer to a data source, delay and cost of data transmission can be reduced, and speed and efficiency of data processing can be improved.

Drawings

Other features, objects and advantages of the application will become more apparent upon reading of the detailed description of a non-limiting implementation, made with reference to the accompanying drawings in which:

FIG. 1 is a flow chart illustrating steps of an intelligent software automated test method of an embodiment;

FIG. 2 shows a step flow diagram of step S1 of an embodiment;

FIG. 3 shows a step flow diagram of step S2 of an embodiment;

FIG. 4 shows a step flow diagram of step S3 of an embodiment;

FIG. 5 shows a step flow diagram of step S4 of an embodiment;

FIG. 6 shows a step flow diagram of step S5 of an embodiment;

fig. 7 shows a step flow diagram of step S6 of an embodiment.

Detailed Description

The following is a clear and complete description of the technical method of the present patent in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, are intended to fall within the scope of the present application.

Furthermore, the drawings are merely schematic illustrations of the present application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.

It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.

Referring to fig. 1 to 7, the application provides an intelligent software automatic test method, which comprises the following steps:

specifically, for example, a software testing environment containing elements of an operating system, a database and middleware is constructed by using publicly available automation scripts and combining with a general cloud service technology. Various key metrics such as CPU utilization, memory usage, network throughput are then collected from the environment.

specifically, for example, deep learning technology (such as convolutional neural network and long-term and short-term memory network) is adopted to analyze and predict the test environment index data collected in the step S1, and reference test case data is generated so as to simulate various possible software use situations.

specifically, the monitoring module is deployed on an edge computing node (e.g., smart phone, embedded device) using, for example, benchmark test case data. This module periodically gathers data regarding aspects of device performance, network status, user interaction.

specifically, for example, with real-time monitoring data, a reinforcement learning algorithm (such as Q-learning, deep Q network) is used to dynamically adjust the test strategy to generate adjustment strategy data to accommodate software variations and uncertainties.

specifically, the multi-modal test data is generated through cloud service calculation, for example, according to the adjustment policy data obtained in step S4. The data may include multiple modes of text, images, speech to cover the various functions and interaction modes of the software.

Specifically, for example, real-time monitoring data and multi-mode test data are comprehensively utilized, a machine learning method (such as logistic regression and support vector machine) is utilized to perform self-adaptive feedback optimization, and final intelligent software test feedback optimization data are generated. These data will be used to guide subsequent software development and optimization efforts.

The invention uses the automatic script and cloud service technology, and can quickly construct and configure the test environment, thereby greatly shortening the time for preparing the test environment. In addition, the efficiency of test execution and optimization can be further improved through the self-adaptive feedback optimization of real-time monitoring data and multi-mode test data. The method not only can obtain the reference test case data through the depth prediction calculation, but also can carry out the enhanced dynamic adjustment according to the real-time monitoring data, and generate the multi-mode test data through the cloud service calculation, thereby comprehensively detecting and verifying the performance and the function of the software and improving the comprehensiveness and the accuracy of the test. By means of real-time monitoring and enhanced dynamic adjustment, the method can find and solve the problem in the first time, and stability and reliability of software are enhanced. The self-adaptive feedback optimization can be performed according to the real-time monitoring data and the multi-mode testing data, so that the software can be better adapted to various environments and conditions, and the adaptability and the flexibility of the software are improved. By means of an automatic test mode, time and energy of manual test can be reduced, and manpower resources are saved. By using cloud services and edge computing nodes, optimized resource allocation and management can be realized, so that computing and storage resources are fully utilized.

Preferably, step S1 is specifically:

step S11: acquiring intelligent software testing requirement data;

specifically, for example, test requirement data of intelligent software is collected, including requirements of functional test, performance test, security test and compatibility test and corresponding test scenes thereof.

specifically, the demand assessment of the test environment is performed, for example, based on the collected test demand data. This involves categorizing and prioritizing the various requirements and determining the key characteristics and resources that the test environment should possess. The following requirements are classified: the requirements are classified, typically by the nature of the requirements, such as functional requirements, performance requirements, security requirements. At the same time, there is a need to subdivide the requirements, e.g. the functional requirements may be subdivided into core functional requirements, additional functional requirements. Demand prioritization: the demands are ordered according to their importance. Ordering is typically performed using methods such as the MoSCoW method (Must have, shoold have, could have, and Won't have). This step will help the team to determine which requirements are most important and which requirements are achievable in a deferred manner.

specifically, an appropriate resource allocation policy is formulated, for example, based on the results of the test environment evaluation. This includes selecting the appropriate type and number of computing nodes, storage resources, network resources, and corresponding operating systems and middleware.

specifically, for example, a pre-written cloud service testing environment configuration script is used to deploy the testing environment according to a resource configuration policy. This includes creating, configuring and connecting computing nodes, installing and setting storage and network resources, and configuring operating systems and middleware.

specifically, for example, after deployment is completed, the test environment is verified. This includes checking whether the resource is properly deployed, whether the service is functioning properly, and whether performance is expected. The verification results may generate test environment verification report data.

specifically, for example, according to the test environment verification report data, a monitoring policy of the test environment is formulated. This includes determining key metrics to monitor, such as CPU utilization, memory usage, network throughput, and formulating corresponding data collection and reporting mechanisms.

Specifically, the generation of the test environment index is performed, for example, in accordance with a test environment monitoring policy. This includes periodically collecting, processing and analyzing the monitored data to generate corresponding test environment index data.

；

;

The invention constructs a test environment verification calculation formula, which quantitatively evaluates the verification result of the test environment, thereby improving the test efficiency and quality; considering a plurality of parameters and indexes of the test environment, thereby improving the comprehensiveness and accuracy of the test; the constant parameters and the weight indexes are adjusted, so that different test requirements and scenes can be flexibly adapted; by logarithmic functionReflecting the complexity of the test environment validation resultsAnd difficulty; by limit function- >Reflecting the stability and convergence of the test environment verification result; by trigonometric function->、/>And +.>Reflecting the periodicity and volatility of the test environment verification result; by derivative function->Reflecting the change rate and sensitivity of the test environment verification result; />Is a test environment parameter item, such as CPU usage rate and memory usage rate, which affects the basic configuration and running state of the test environment; />The security parameter items of the test environment, such as encryption mode and firewall setting, influence the confidentiality and reliability of the test environment; />Is a test environment performance parameter item, such as response time, throughput, affecting the efficiency and load capacity of the test environment; />Is a base constant term, and affects the growth speed and curve shape of the logarithmic function; />The method is to configure the data format of test script data, and influence the calculation precision and range of the limit function; derivative function->By->And->Interaction, mean->When changing, the value of the derivative function reflects the change rate of the logarithmic function; the present invention provides an accurate and flexible computing method for verification and evaluation of software testing environments to provide reliable data support.

Preferably, step S2 is specifically:

specifically, for example, the collected test environment index data is normalized, for example, each item of data is converted into a unified scale range by a maximum and minimum normalization or z-score normalization method, so that the influence of the dimension among the data is reduced, and the processed data is more suitable for subsequent feature selection and model construction.

specifically, for example, feature selection is performed on normalized test environment index data, such as a method of using a pearson correlation coefficient method and a feature selection method based on a tree model, so as to screen out a key index having the greatest influence on a prediction model.

specifically, for example, data is selected for the screened test environment indexes, a neural network is used for constructing a prediction model, for example, a feedforward neural network or a convolution neural network is adopted for constructing the model, and the mapping relation between the test environment indexes and the reference test cases is learned.

specifically, for example, the established test environment index prediction model is checked, such as the model is checked and calculated by using the indexes of the mean square error and the root mean square error, so as to obtain test model verification report data.

Specifically, for example, the test model verification report data is used to perform prediction calculation on the test environment index prediction model, for example, a group of test environment index data is input, and the prediction model is used to perform prediction calculation to obtain corresponding reference test case data.

specifically, for example, the first feature selection is performed on the normalized data of the test environment index, and an information gain method is adopted. The method is characterized in that the characteristic selection is carried out based on information entropy, and the characteristic with larger information gain value is selected as an important characteristic by calculating the information gain value of each characteristic, so that first test environment index selection data is obtained.

Specifically, for example, the second feature selection is performed on the normalized data of the test environment index, and a feature selection method based on random forests is adopted. The random forest is an integrated learning model, the importance of each feature in the random forest can be evaluated, and the feature with higher importance is selected as a key feature, so that second test environment index selection data are obtained.

;

validating report data for a test model,/->For the first test weight item, +.>For testing environmental index deviation term, < ->For error checking weight term, +.>For the second test weight item, +.>To adjust the proportional check weight term, +. >Predicting the total number of data of the model data for the test environmental indicator, < >>For the third test weight item, +.>For testing environmental index constant term, < ->For periodic adjustment of weight index +.>For the fourth test weight item, +.>For test time data, ++>For power control item->For testing the predicted value of the environmental index, < >>The real value of the environmental index is tested.

The invention constructs a test model test calculation formula which comprises various error test methods, such as proportional-based error test and time-based dynamic error test, and can ensure comprehensive evaluation and test of the prediction capability of the model from different angles and different levels. The various weight and adjustment terms in the formula allow for personalized adjustments to specific test environments and specific predictive tasks, which enhances the flexibility and adaptability of the formula in coping with different test tasks. First check weight termAnd->Indicating that the effect of model prediction error is adjusted based on the square root relationship; likewise, the second test weight term->And->Indicating that another influence factor is adjusted based on the logarithmic relationship. Furthermore, the->The effect of (2) is to indicate the time factor in the test procedure, whereas +. >Then it is the power control term, which can be influenced by adjusting it>Is a function of the degree of influence of (a). According to the invention, different weight items, adjustment proportion and test time factors are introduced, so that the accuracy, deviation and stability of the model are better evaluated.

Preferably, step S3 is specifically:

specifically, for example, edge computing node data including information of computing power, storage capacity, network connection speed of the node is acquired. Edge computing node data is then evaluated using the baseline test case data (e.g., expected test workload, expected data processing speed) to determine edge nodes that are most suitable for performing the test tasks, generating edge node evaluation data.

specifically, for example, according to the edge node evaluation data, a test monitoring policy suitable for the edge node is generated in consideration of the computing power, the storage capacity and the network connection speed of the edge node, and edge test monitoring policy data is obtained.

specifically, for example, according to the edge test monitoring policy data, determining which monitoring tasks are deployed on the edge computing node, for example, CPU usage monitoring, memory usage monitoring, and network transmission speed monitoring, so as to deploy corresponding monitoring tasks on the edge computing node, and obtain edge node deployment data.

specifically, for example, according to the edge node deployment data, the type, the acquisition frequency and the acquisition mode of the monitoring data to be acquired are determined, and a corresponding monitoring strategy is generated to obtain monitoring data acquisition strategy data.

Specifically, for example, according to the monitoring data acquisition policy data, the monitoring task is started to be executed on the edge computing node, and real-time monitoring data such as CPU utilization, memory utilization and network transmission speed are acquired.

Preferably, step S4 is specifically:

Specifically, for example, the acquired real-time monitoring data may contain noise and abnormal values. Therefore, the step adopts data preprocessing means such as data cleaning and normalization processing, so as to acquire the preprocessed real-time monitoring data.

specifically, for example, a pre-set reinforcement learning algorithm (e.g., deep Q network, actor-Critic method) is used to perform model training on the pre-processed real-time monitoring data to optimize the test strategy. Parameters of the model are continuously adjusted in the training process so as to maximize the obtained rewards, and therefore a real-time monitoring strengthening model is generated.

specifically, after the real-time monitoring reinforcement model is generated, for example, the simulated test environment can be used for verification and optimization to generate simulated test report data. This process can help to understand the performance and effect of the model in order to optimize it.

In particular, from simulation test report data, for example, problems that may exist with the model in some aspects, such as accuracy, stability, etc., may be appreciated. Based on these issues, parameters or policies of the model may be dynamically adjusted, generating dynamically adjusted policy data.

specifically, the dynamic adjustment policy data is verified in real time, for example, to confirm whether the policy adjustment is valid. This step may generate real-time policy validation report data for further optimization.

Specifically, the dynamic adjustment policy is further optimized, for example, according to the real-time policy verification report data, for example, some parameters may be fine-tuned, or application conditions of some policies may be changed, and finally adjustment policy data is generated.

Preferably, step S5 is specifically:

specifically, for example, policy resolution is a process of interpreting the adjustment policy data in detail. This step may involve a structured description of the policy, e.g., determining which parts are conditions, which parts are operations, and how the parts are related to each other. The result of this process is policy resolution data.

specifically, for example, after parsing the policy, the relevant data needs to be defined. For example, defining characteristics of edge compute nodes to be tested, or defining software characteristics to be tested. This step generates modality definition data.

specifically, the database of edge nodes is configured, for example, according to modality definition data. This may include creating a new data table, defining data fields, or setting a data index. After this process is completed, data source configuration report data is generated.

Specifically, for example, after the data source is configured, the generation of the multi-mode test data can be performed by using cloud service calculation according to the configuration report data. These test data may include data of various modalities, such as text data, picture data, audio data. After this process is completed, multi-modal test data is obtained.

Preferably, step S6 is specifically:

specifically, for example, feature extraction is to obtain representative characteristics from input real-time monitoring data and multi-mode test data through a series of calculations, where the characteristics or features can better reflect essential attributes of the data. The result of this step is feedback characteristic data.

specifically, the adaptive optimization is performed, for example, based on the extracted feedback feature data. Specific optimization methods may include various search algorithms, such as genetic algorithms, simulated annealing algorithms, or machine learning algorithms, such as neural networks, decision trees, to find optimal or solutions that meet certain conditions, the result of which is optimization objective data.

specifically, an optimization strategy is generated, for example, from the optimization target data. An optimization strategy is an action scheme that is performed under conditions that meet optimization objectives, including a specific series of implementation steps and operations. The result of this process is optimization policy data.

Specifically, for example, the optimization strategy data is implemented and evaluated. The purpose of the evaluation is to check whether the optimization strategy is valid and whether the result of the optimization is expected. The evaluation may include comparing the differences before and after the optimization, checking whether the result of the optimization meets expectations. After the process is completed, intelligent software test feedback optimization data can be obtained.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent software automatic test method is characterized by comprising the following steps:

step S2, including:

step S25: predicting and calculating the test environment index prediction model data by using the test model verification report data so as to obtain reference test case data;

step S4, including:

performing data preprocessing according to the real-time monitoring data, thereby obtaining preprocessed real-time monitoring data;

training the reinforcement model of the preprocessed real-time monitoring data through a preset reinforcement learning algorithm, so as to acquire a real-time monitoring reinforcement model;

performing optimization test adjustment on the real-time monitoring reinforcement model so as to obtain simulation test report data;

performing strategy dynamic adjustment according to the simulation test report data so as to obtain dynamic adjustment strategy data;

Performing real-time verification according to the dynamic adjustment strategy data, thereby obtaining real-time strategy verification report data;

performing policy optimization on the dynamic adjustment policy data by using the real-time policy verification report data so as to acquire the adjustment policy data;

2. The method according to claim 1, wherein step S1 is specifically:

step S11: acquiring intelligent software testing requirement data;

3. The method according to claim 2, wherein the test environment verification in step S15 is quantitatively verified by a test environment verification calculation formula, wherein the test environment verification calculation formula is specifically:

；

;

verifying reporting data for test environment, < >>For testing environmental parameter items->For testing environmental safety parameter items, +.>For testing environmental performance parameter items +_>Is a base constant term,/->For configuring the data format of the test script data +.>In order to be able to select the order item,is->Individual test environment verification items,/->For the first weight index +.>Is->Individual test environmental parameter items,/->For the second weight index->Is->A test environmental security parameter item, +.>For the third weight index +.>Is->And testing environmental performance parameter items.

4. The method according to claim 1, wherein the test environment indicator selection data includes first test environment indicator selection data and second test environment indicator selection data, and step S22 is specifically:

5. The method according to claim 1, wherein step S3 is specifically:

acquiring edge computing node data, and carrying out edge node optimal evaluation on the edge computing node data by utilizing the reference test case data so as to acquire edge node evaluation data;

performing test monitoring strategy generation according to the edge node evaluation data so as to obtain edge test monitoring strategy data;

monitoring deployment is carried out on the edge computing nodes according to the edge test monitoring policy data, so that edge node deployment data are obtained;

generating a monitoring strategy according to the edge node deployment data, thereby acquiring monitoring data acquisition strategy data;

and acquiring the monitoring data according to the monitoring data acquisition strategy data, thereby acquiring real-time monitoring data.

6. The method according to claim 1, wherein step S5 is specifically:

performing policy analysis on the adjustment policy data so as to obtain policy analysis data;

Performing data definition generation according to the strategy analysis data so as to obtain modal definition data;

performing edge node database configuration according to the mode definition data so as to obtain data source configuration report data;

and generating multi-mode test data according to the data source configuration report data, so as to obtain the multi-mode test data.

7. The method according to claim 1, wherein step S6 is specifically:

performing feature extraction according to the real-time monitoring data and the multi-mode test data, so as to obtain feedback feature data;

performing self-adaptive optimization according to the feedback characteristic data, so as to obtain optimization target data;

generating an optimization strategy according to the optimization target data, so as to obtain optimization strategy data;

and implementing and evaluating according to the optimized measurement data, thereby obtaining intelligent software test feedback optimized data.