An AI-based test performance analysis and prediction system and method

Abstract

The application relates to the technical field of software test management automation, in particular to a test performance analysis and prediction system and method based on AI. The system comprises an intelligent data acquisition module, an AI analysis engine module, a multidimensional evaluation module, a performance prediction module, a visual interactive module and an intelligent report generation module, and is additionally provided with an abnormality detection and early warning module, and each module cooperates to realize data full-link processing. The method is a six-step standardized process, intelligent data acquisition preprocessing is completed through dynamic mapping, a large language model is called to carry out multidimensional performance analysis, the evaluation weight is dynamically adjusted in combination with the project stage to generate a comprehensive score, performance trend prediction and abnormal inflection point identification are realized by relying on a Prophet or LSTM algorithm, and visual interactive exploration and multi-format report automatic generation are simultaneously supported. The application realizes full-process automation of test performance analysis and prediction, improves the scientificity and accuracy of test performance management, and provides comprehensive data support for related decision-making.

Description

Technical Field

[0001] This invention relates to the field of software test management automation technology, specifically to an AI-based test performance analysis and prediction system and method. Background Technology

[0002] Currently, software testing performance management methods on the market are still mainly based on traditional manual models, supplemented by simple test management systems for data recording. They have not yet formed an intelligent, end-to-end analysis and prediction system, revealing numerous problems in practical applications and failing to meet the evolving needs of modern software testing. In the data collection phase, different companies use different test management systems with heterogeneous database structures. Traditional methods cannot automatically connect to these heterogeneous databases, requiring manual querying and exporting to extract defect data and related information. Furthermore, the cleaning and formatting of raw data also rely on manual work, consuming significant manpower and time, and being highly susceptible to data omissions and errors due to human intervention, resulting in a lack of completeness and consistency in the data sources for subsequent analysis.

[0003] In the performance analysis phase, traditional methods rely solely on manual statistical analysis of collected basic data. The analysis is limited to fundamental metrics such as test case execution count and defect discovery, failing to provide in-depth analysis across multiple dimensions, including response timeliness, defect quality, and team collaboration. Furthermore, it struggles to uncover underlying performance issues and underlying logic. Simultaneously, the capabilities of general artificial intelligence (AI) technologies have not been effectively integrated into test performance analysis scenarios. The semantic understanding and data analysis capabilities of large language models cannot be leveraged to generate business-guided insights and optimization suggestions, leaving performance analysis merely at the data statistics level and unable to provide practical support for optimizing testing operations.

[0004] In the performance evaluation process, there is a lack of standardized, multi-dimensional evaluation systems. Most companies use a single evaluation indicator with fixed weights, failing to dynamically adjust them according to different lifecycle stages of the project, such as requirements review, integration testing, and system testing. This leads to a disconnect between evaluation results and the actual testing focus of the project. Furthermore, the evaluation process is heavily influenced by subjective human experience, lacking unified quantitative scoring rules. Performance results between different testing projects and teams lack horizontal comparability, and longitudinal analysis of results for the same evaluation object at different times is difficult, compromising the objectivity and scientific rigor of the evaluation.

[0005] In the performance trend prediction stage, traditional management methods lack effective technical means and algorithmic support. They rely solely on managers' historical experience to subjectively judge future performance trends, failing to build scientific predictive models based on comprehensive historical performance data. This results in an inability to accurately generate future performance trends or identify potential anomalies during performance development. Consequently, test performance management remains in a reactive state of post-event analysis and problem remediation, hindering the development of proactive prevention strategies and making it easy for performance declines to go undetected, impacting test project progress and product quality.

[0006] Overall, the various stages of existing software testing performance management are disconnected, lacking an automated and integrated execution process. Data collection, analysis, evaluation, prediction, and visualization all require manual intervention, resulting in low overall management efficiency. Furthermore, the system suffers from low levels of intelligence, insufficient analytical depth, a lack of objectivity in evaluation, and a lack of scientific rigor in prediction, failing to meet the current demands for refined and efficient management of software testing projects. Therefore, developing an AI-based test performance analysis and prediction system and method capable of intelligent data collection, multi-dimensional in-depth analysis, scientific quantitative evaluation, accurate trend prediction, intuitive visualization, and intelligent report generation has become a pressing technical challenge in the field of software testing management. Summary of the Invention

[0007] The purpose of this invention is to provide an AI-based test performance analysis and prediction system and method to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: An AI-based test performance analysis and prediction system includes: The intelligent data acquisition module is configured to automatically identify and connect to the database of the test management system, extract defect data and related information through a dynamic mapping table structure, and perform cleaning and formatting preprocessing on the raw data. The AI ​​analysis engine module communicates with the intelligent data acquisition module and is configured to perform multi-dimensional performance analysis on preprocessed defect data by calling the large language model API and combining it with preset structured prompt word templates, generating structured analysis results containing key insights and optimization suggestions. The multi-dimensional evaluation module is configured to receive the analysis results from the AI ​​analysis engine module. Based on an evaluation system that includes response timeliness, defect quality, test coverage, and team collaboration, it calculates scores for each dimension and dynamically adjusts the weights according to the project stage to generate a comprehensive performance score. The performance prediction module is configured to acquire historical performance data, build and train a prediction model based on time series analysis or machine learning algorithms, and generate future performance trend prediction results and information on potential abnormal inflection points. The visualization and interaction module is configured to provide a web-based interactive interface, enabling database connection configuration, analysis parameter settings, and dynamic display of comprehensive performance scores, scores for each dimension, and performance trend prediction results in the form of charts and tables. It also supports interactive exploration through data drill-down and dimension filtering. The intelligent report generation module is configured to automatically integrate comprehensive performance scores, scores for each dimension, performance trend prediction results, and optimization suggestions to generate a structured performance analysis report, and supports export in multiple formats.

[0009] As a preferred option, the intelligent data acquisition module is further configured for: Establish a connection through the database connection library, and support two configuration methods: direct connection and connection string; By querying the database system tables, the data table structure and field mapping relationships of the test management system are automatically identified; Based on the identification results, a data extraction model is dynamically constructed, and parameterized queries are executed to obtain defect data, user information, and product information.

[0010] As a preferred option, the AI ​​analytics engine module is further configured for: Construct structured prompts that include system role settings, analysis dimension definitions, and output format constraints; The aggregated performance data is populated into structured prompts, and the large language model API is invoked via an HTTP request. Parse the JSON response returned by the API, extract multi-dimensional analysis results, key performance insights and improvement suggestions, and deserialize them into structured data for use by subsequent modules.

[0011] As a preferred option, the multi-dimensional evaluation module is further configured for: Iterate through the performance data and calculate the score for each preset dimension. Obtain information on the current stage of the project and dynamically adjust the weight of each dimension in the overall performance score based on preset rules; The weighted average method is used to calculate the overall performance score and generate score data for each dimension for radar chart visualization.

[0012] As a preferred embodiment, the performance forecasting module is further configured to: Historical performance data is aggregated according to a preset time granularity and converted into a time series format; Train a time series forecasting model using the Prophet algorithm or LSTM algorithm to generate forecast data that includes future performance trends; Calculate the confidence interval of the prediction results and use a mutation point detection algorithm to identify potential abnormal inflection points in the performance trend.

[0013] As a preferred option, the visual interaction module is further configured for: Provides a dynamic configuration interface for database connection parameters, API keys, and prediction model parameters; The chart component displays performance trend forecasts and confidence intervals, while the table component displays performance details and personnel information. Interactive components enable time range adjustment and evaluation dimension filtering, while expandable components provide drill-down access to detailed defect information.

[0014] As a preferred option, the system also includes an anomaly detection and early warning module, configured for: Based on statistical process control or machine learning anomaly detection algorithms, monitor fluctuations in performance indicators in real time; When an indicator is detected to exceed a preset threshold or to experience abnormal fluctuations, an early warning notification is automatically triggered, and root cause analysis suggestions based on data correlation analysis are provided.

[0015] An AI-based method for test performance analysis and prediction, applied to the system, includes the following steps: Step S1: Intelligent data acquisition and preprocessing: Automatically identify the database structure of the test management system through dynamic mapping, extract defect data, and clean and format it; Step S2: AI Large Model Interaction and Performance Analysis: Construct structured prompt words, call the large language model API to perform multi-dimensional analysis of performance data, and parse the returned structured analysis results; Step S3: Multi-dimensional assessment and weight calculation: Calculate the scores of each dimension based on the preset assessment system, and dynamically adjust the weights according to the project stage to generate a comprehensive performance score; Step S4: Time Series Analysis and Trend Prediction: Train a time series prediction model based on historical performance data to generate future performance trend predictions and information on abnormal inflection points; Step S5: Visualization and Interactive Exploration: Dynamically display analysis results and predicted trends in the web interface, and provide interactive functions such as data drill-down and filtering; Step S6: Intelligent Report Generation and Management: Automatically integrates analytical data and improvement suggestions to generate structured performance reports and supports export in multiple formats.

[0016] As a preferred embodiment, step S2 specifically includes: Construct a structured prompt word template that includes system roles, analysis dimensions, and output format constraints; The preprocessed performance data is populated into the prompt word template and sent to the large language model via API request; Receive and parse the JSON response returned by the API, and extract structured multi-dimensional analysis conclusions and optimization suggestions in natural language description.

[0017] As a preferred embodiment, step S4 specifically includes: Use the Pandas library to aggregate historical performance data by day or week and convert it into a time series format; Use the Prophet or LSTM algorithm library to train the model on the time series data; Based on the trained model, generate performance trend predictions for a future preset time range and calculate the confidence interval of the prediction results; The mutation point detection algorithm is used to identify potential performance inflection points within the prediction period.

[0018] As can be seen from the technical solutions provided by the present invention above, the AI-based test performance analysis and prediction system and method provided by the present invention have the following beneficial effects: It enables intelligent collection and standardized processing of test performance data. By relying on dynamic mapping table structure technology, it breaks down the heterogeneous database barriers of different test management systems, completes database docking in a non-intrusive manner, automatically and accurately extracts defect data and related information, and completes the cleaning and formatting preprocessing of raw data. It ensures the integrity, consistency and availability of data from the source, replaces the traditional manual data processing mode, greatly improves data processing efficiency, and effectively avoids errors and deviations caused by manual operation. By integrating a large language model, we can achieve multi-dimensional and in-depth analysis of test performance. By constructing structured prompt word templates, we can adapt the capabilities of a general large language model to the specific scenarios of test performance analysis. Combined with API full-link management technology, we can achieve stable interaction with the large model. We can extract key insights and optimization suggestions with business guidance from preprocessed performance data, break through the dimensional limitations and cognitive bottlenecks of traditional manual performance analysis, and upgrade performance analysis from simple numerical statistics to in-depth business judgment, providing scientific and targeted analytical basis for test performance management decisions. An objective and standardized multi-dimensional performance evaluation system has been established. Unified scoring rules and judgment standards have been formulated around the four core dimensions of response timeliness, defect quality, test coverage, and team collaboration. At the same time, the weight ratio of each dimension is dynamically adjusted in combination with the project life cycle stage. A comprehensive performance score is generated by weighted average method, which completely eliminates the subjective judgment problem of traditional manual performance evaluation. This allows for horizontal comparison of the performance results of different test projects and different test teams, and vertical comparison of the results of the same evaluation object at different times, so as to achieve quantitative and objective evaluation of test performance. It enables forward-looking and accurate prediction of test performance trends, integrates all historical performance data accumulated in the system, and converts it into a standardized time series format through time series analysis technology. It supports intelligent selection and adaptation training of Prophet and LSTM algorithms. It can not only generate performance trend data within a preset time range in the future, but also calculate the confidence interval of the prediction results to quantify the reference value. At the same time, it accurately identifies potential performance anomaly inflection points through mutation point detection algorithm, breaking the limitation of traditional performance analysis that only focuses on the current situation. It promotes the transformation of test performance management from post-analysis to pre-prevention and control, and facilitates the formulation of optimization strategies in advance to avoid the risk of performance decline. We have created an efficient, visually interactive, and intelligent report output system. Through a web interface, we can dynamically adjust configuration information such as database connection parameters and API keys. We can intuitively display core data such as comprehensive performance scoring dimensions, performance trend predictions, etc. in various forms such as charts and tables. At the same time, we support interactive exploration operations such as adjusting time range, filtering dimensions, and drilling down data, which lowers the operational threshold of performance management. The intelligent report generation module can automatically integrate full-process analysis data and optimization suggestions to generate structured performance reports and support export in multiple formats, improving the efficiency of performance analysis results delivery and the standardization of archiving. This system enables real-time monitoring and intelligent early warning of test performance anomalies. By integrating statistical process control and machine learning anomaly detection algorithms into the anomaly detection and early warning module, it provides 24 / 7 real-time monitoring of all-dimensional performance indicators. When an indicator exceeds a preset threshold or exhibits abnormal fluctuations, it automatically triggers multiple forms of early warning notifications. Simultaneously, it provides root cause analysis suggestions through data correlation analysis, enabling timely detection and accurate location of performance anomalies. This significantly improves the efficiency and effectiveness of handling performance anomalies and ensures the stability of test performance. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the structure of an AI-based test performance analysis and prediction system according to the present invention; Figure 2 This is a schematic diagram of the steps of an AI-based test performance analysis and prediction method according to the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0021] To better understand the above technical solutions, the following will provide a detailed description of the technical solutions in conjunction with the accompanying drawings and specific embodiments.

[0022] like Figure 1-2As shown, this embodiment of the invention provides an AI-based test performance analysis and prediction system, including: The intelligent data acquisition module is configured to automatically identify and connect to the database of the test management system, extract defect data and related information through a dynamic mapping table structure, and perform cleaning and formatting preprocessing on the raw data. The AI ​​analysis engine module communicates with the intelligent data acquisition module and is configured to perform multi-dimensional performance analysis on preprocessed defect data by calling the large language model API and combining it with preset structured prompt word templates, generating structured analysis results containing key insights and optimization suggestions. The multi-dimensional evaluation module is configured to receive the analysis results from the AI ​​analysis engine module. Based on an evaluation system that includes response timeliness, defect quality, test coverage, and team collaboration, it calculates scores for each dimension and dynamically adjusts the weights according to the project stage to generate a comprehensive performance score. The performance prediction module is configured to acquire historical performance data, build and train a prediction model based on time series analysis or machine learning algorithms, and generate future performance trend prediction results and information on potential abnormal inflection points. The visualization and interaction module is configured to provide a web-based interactive interface, enabling database connection configuration, analysis parameter settings, and dynamic display of comprehensive performance scores, scores for each dimension, and performance trend prediction results in the form of charts and tables. It also supports interactive exploration through data drill-down and dimension filtering. The intelligent report generation module is configured to automatically integrate comprehensive performance scores, scores for each dimension, performance trend prediction results, and optimization suggestions to generate a structured performance analysis report, and supports export in multiple formats.

[0023] In this embodiment, the intelligent data acquisition module is the data foundation and source entry point of the AI-based test performance analysis and prediction system. Through automatic adaptation to the test management system database and full-link data processing, it provides high-quality and standardized basic data support for subsequent full-process performance analysis, evaluation and prediction. The intelligent data acquisition module serves as the front-end data hub of the entire system, primarily responsible for three core tasks: adapting and connecting to the test management system database, accurately extracting target data, and standardizing the preprocessing of raw data. It breaks down data barriers between different test management systems, enabling non-intrusive data integration with heterogeneous databases. It automates the acquisition and standardization of core performance-related data such as defect data, user information, and product information, ensuring the integrity, consistency, and availability of data input to subsequent modules. This provides a solid data foundation for downstream processes such as AI analysis, multi-dimensional evaluation, and performance prediction. The intelligent data acquisition module includes: Structured prompt word building unit: Prompt word template system construction: Construct standardized prompt word templates that include three core parts: system role setting, analysis dimension definition, and output format constraints. Clarify the exclusive role of the large language model in the test performance analysis scenario, define fixed analysis dimensions such as response timeliness, defect quality, test coverage, and team collaboration, and standardize the structure and content boundaries of the output content to ensure that the output content of the large language model fits the business needs of test performance analysis and avoids invalid information output. Dynamic template adaptation and content filling: Based on the differences in project type, project stage, and analysis object, the prompt word template is dynamically adjusted to adapt to the analysis focus under different business scenarios; the aggregated performance data output by the intelligent data collection module is standardized and filled according to the rules set in the template to ensure that the content input to the large language model has a complete business context and accurate data support, and to ensure the relevance and accuracy of the analysis results; Template compliance and validity verification: The prompts for completing the content filling are verified for compliance and completeness. The verification is conducted to check whether the role settings are clear, whether the analysis dimensions are complete, whether the data filling is accurate, and whether the format constraints are clear. Prompts that fail the verification are automatically corrected and optimized to ensure that the prompts input into the large language model meet the specifications and ensure the quality of the analysis output from the source. Large Model API Interaction Scheduling Unit: Standardized API Request Encapsulation: Based on the preset large language model API interface specification, the content of the verification prompt is encapsulated in a standardized HTTP request, and the corresponding interface authentication information, request parameters, timeout control rules and retry mechanism are configured to ensure that the request content conforms to the calling specification of the large language model interface and to ensure the stability and success rate of the interface call. API Call and Status Management: Initiate call requests to the target large language model API and monitor the entire link status of the API call in real time, including core indicators such as request response time, API return status, and link connectivity; when an API call timeout, call failure, or return exception is detected, automatically perform retry operations and switch to a backup API interface channel to ensure the continuous stability of the large model interaction process; Call traffic and permission control: Dynamically control the call frequency and traffic of large language model APIs, adapt to the interface rate limiting rules of different large model service providers, and avoid interface blocking issues caused by excessive calls; at the same time, unified management and permission verification of API keys ensure the security and compliance of interface calls; Response parsing and structured transformation unit: Response content legality verification: Upon receiving the response content returned by the large language model API, the content is first verified for legality, completeness and compliance. It checks whether the returned content conforms to the preset output format constraints, whether it covers all set analysis dimensions, and whether there is invalid content or illegal information. For content that fails the verification, a second analysis request is automatically triggered. Core Information Extraction and Analysis: Deeply analyze the verified response content, extract three core contents according to preset rules: multi-dimensional analysis results, key performance insights, and improvement suggestions. Remove invalid format information and redundant content from the response content, and accurately locate the core analysis conclusions related to the business. Structured data deserialization: The extracted core analysis content is deserialized according to the system's preset data structure standards, and converted into standardized structured data that can be directly read and called by downstream multi-dimensional evaluation modules, visualization interaction modules, and intelligent report generation modules. This completes the conversion of unstructured natural language content into the system's standardized data format, enabling seamless flow of analysis content within the system.

[0024] In this embodiment, the AI ​​analysis engine module is the core intelligent hub of the AI-based test performance analysis and prediction system. It is the core carrier for realizing test performance from data statistics to in-depth insights. Through the deep semantic understanding and data analysis capabilities of the large language model, it completes multi-dimensional business mining of standardized test performance data, generates guiding analysis insights and optimization suggestions, and provides core analysis support for subsequent multi-dimensional evaluation, performance prediction, intelligent report generation and other links. The AI ​​analysis engine module communicates in real time with the intelligent data acquisition module. It primarily receives standardized pre-processed performance data and is responsible for three core tasks: building a structured prompt word system, scheduling API interactions with the large language model, and parsing and converting returned results into structures. It deeply integrates standardized test performance data with the deep analysis capabilities of the large language model, breaking through the dimensional limitations and efficiency bottlenecks of traditional manual performance analysis. It performs multi-dimensional in-depth analysis of test performance data, accurately uncovering performance problems, root causes, and optimization directions behind the data, generating standardized and applicable structured analysis results. This provides core analytical content and decision-making basis for downstream multi-dimensional evaluation, visualization, and intelligent report generation. The AI ​​analysis engine module includes: Structured prompt word building unit: The system of prompt word templates is built and constructed. It includes three core parts: system role setting, analysis dimension definition, and output format constraints. The standardized prompt word templates clarify the exclusive role of the large language model in the test performance analysis scenario, define fixed analysis dimensions such as response timeliness, defect quality, test coverage, and team collaboration, and standardize the structure and content boundaries of the output content to ensure that the output content of the large language model fits the business needs of test performance analysis and avoids the output of invalid information. Dynamic template adaptation and content filling: Based on the differences in project type, project stage, and analysis object, the prompt word template is dynamically adjusted to adapt to the analysis focus under different business scenarios; the aggregated performance data output by the intelligent data collection module is standardized and filled according to the rules set in the template to ensure that the content input to the large language model has a complete business context and accurate data support, and to ensure the relevance and accuracy of the analysis results; Template compliance and validity verification: The prompts for completing the content filling are verified for compliance and completeness. The verification is conducted to check whether the role settings are clear, whether the analysis dimensions are complete, whether the data filling is accurate, and whether the format constraints are clear. Prompts that fail the verification are automatically corrected and optimized to ensure that the prompts input into the large language model meet the specifications and ensure the quality of the analysis output from the source. Large Model API Interaction Scheduling Unit: Standardized API Request Encapsulation: Based on the preset large language model API interface specification, the content of the verification prompt is encapsulated in a standardized HTTP request, and the corresponding interface authentication information, request parameters, timeout control rules and retry mechanism are configured to ensure that the request content conforms to the calling specification of the large language model interface and to ensure the stability and success rate of the interface call. API Call and Status Management: Initiate call requests to the target large language model API and monitor the entire link status of the API call in real time, including core indicators such as request response time, API return status, and link connectivity; when an API call timeout, call failure, or return exception is detected, automatically perform retry operations and switch to a backup API interface channel to ensure the continuous stability of the large model interaction process; Call traffic and permission control: Dynamically control the call frequency and traffic of large language model APIs, adapt to the interface rate limiting rules of different large model service providers, and avoid interface blocking issues caused by excessive calls; at the same time, unified management and permission verification of API keys ensure the security and compliance of interface calls; Response parsing and structured transformation unit: Response content legality verification: When receiving the response content returned by the large language model API, the first step is to verify the legality, completeness and compliance of the content. It checks whether the returned content conforms to the preset output format constraints, whether it covers all the set analysis dimensions, and whether there is any invalid content or illegal information. For content that fails the verification, a second analysis request is automatically triggered. Core Information Extraction and Analysis: Deeply analyze the verified response content, extract three core contents according to preset rules: multi-dimensional analysis results, key performance insights, and improvement suggestions. Remove invalid format information and redundant content from the response content, and accurately locate the core analysis conclusions related to the business. Structured data deserialization: The extracted core analysis content is deserialized according to the system's preset data structure standards, and converted into standardized structured data that can be directly read and called by downstream multi-dimensional evaluation modules, visualization interaction modules, and intelligent report generation modules. This completes the conversion of unstructured natural language content into the system's standardized data format, enabling seamless flow of analysis content within the system.

[0025] In this embodiment, the multi-dimensional evaluation module is the core of the quantitative evaluation of the AI-based test performance analysis and prediction system. It is the key carrier for realizing the transformation of test performance from qualitative analysis to quantitative evaluation. Through a standardized evaluation system and a dynamic weight adaptation mechanism, it transforms the analysis results output by the AI ​​analysis engine into quantifiable and comparable dimensional scores and comprehensive performance scores, providing core quantitative evaluation basis for subsequent performance prediction visualization and intelligent report generation. The multi-dimensional evaluation module, at its core, inherits the structured analysis results from the AI ​​analysis engine module and is responsible for four core tasks: calculating dimension scores, dynamically adapting weights to the overall score, and generating visualized data output. It constructs a standardized evaluation system based on four core dimensions: response timeliness, defect quality, test coverage, and team collaboration. It traverses all performance data to accurately calculate scores for each dimension, dynamically adjusts the weighting of each dimension according to the project's lifecycle stage, and generates an objective and fair overall performance score using a weighted average method, simultaneously generating standardized data adapted for visualization. This module breaks through the limitations of traditional subjective and singular performance evaluations, achieving multi-dimensional, all-scenario adaptable quantitative evaluation of the performance of testing teams and projects. It provides standardized quantitative data support for the implementation of the system's core functions. The multi-dimensional evaluation module includes: Assessment System Management Unit: Core Dimension System Construction: Construct a standardized evaluation system centered on response timeliness, defect quality, test coverage, and team collaboration. Organize performance evaluation criteria covering the entire testing process around the four core dimensions, so that performance evaluation can comprehensively reflect the execution quality and efficiency of testing work, forming a multi-dimensional and all-round performance evaluation framework. Standardized management of evaluation criteria: Develop unified scoring rules and judgment criteria for the four core dimensions, clarify the score range corresponding to different performance under each dimension, define the core data basis and judgment boundary for score calculation, eliminate the result deviation caused by inconsistent evaluation criteria, and ensure that the score calculation of all evaluation objects follows the same criteria. The system is flexible and adaptable: It supports fine-tuning of the scoring rules for each dimension according to the personalized needs of enterprise test management and the business characteristics of different projects. It can also supplement the evaluation criteria under the dimensions according to the actual management focus, so that the evaluation system can adapt to the test performance management requirements of enterprises of different industries and sizes. Dimensional score calculation unit: Performance data collection and verification: Receive structured analysis results from the AI ​​analysis engine module in real time, synchronously retrieve standardized raw performance data processed by the intelligent data acquisition module, complete the full collection of performance data corresponding to each evaluation dimension, and verify the completeness and accuracy of the collected data, remove invalid data, and fill in missing data to ensure the reliability of the data source for score calculation. Single-dimensional score calculation: According to the preset standardized scoring rules, the entire collection of effective performance data is traversed, and the four core dimensions of response timeliness, defect quality, test coverage and team collaboration are calculated one by one to obtain the independent score results of each dimension. At the same time, the original data support and calculation process corresponding to the score of each dimension are recorded simultaneously. Score Result Reasonableness Verification: Perform compliance verification on the calculated scores of each dimension, check whether the scores are within the preset scoring range, trace the source of abnormal scores that exceed the range, investigate and correct problems in the calculation rules or data sources, and finally generate accurate final scores for each dimension. Dynamic weight adaptation unit: Project phase information acquisition and identification: Real-time connection to relevant project management data ports to accurately obtain information on the current life cycle phase of the project, complete accurate identification and feature matching of project phases, covering all project life cycle nodes such as requirement review, use case design, integration testing, system testing, and deployment regression; Stage weighting rule management: Preset dimension weighting rules corresponding to each stage of the project lifecycle, set differentiated weighting percentages according to the core test management objectives of different stages, fit the actual test work focus of the project, and make the weighting reflect the performance evaluation focus of each stage. Dynamic weight adjustment and compliance verification: Based on the current stage of the identified project, the corresponding weight allocation rules are automatically matched to dynamically adjust the weight ratio of each evaluation dimension in the comprehensive performance score. At the same time, the weights of all dimensions after adjustment are normalized and verified to ensure that the sum of the weights of each dimension conforms to the evaluation logic and to ensure the objectivity and rigor of the comprehensive score calculation. Comprehensive score generation and data output unit: Overall performance score calculation: Based on the final scores of each core dimension and the dynamically adjusted dimension weights, the weighted average method is used to complete the final calculation of the overall performance score, generating the overall performance quantification results of the corresponding test project or test team, which comprehensively reflects the overall performance level of the testing work; Visualized data standardization processing: The final scores of each core dimension and the overall performance score are converted into a standardized data format that is compatible with visualization components such as radar charts, generating a structured dataset that can be directly called by the visualization interaction module, ensuring that the performance data can be smoothly displayed and interacted with on the front-end interface; Scoring results compilation and synchronization: The scores of each dimension, the comprehensive performance score and the corresponding calculation details are compiled to form a standardized quantitative evaluation result set, which is synchronously transmitted to the performance prediction module, the visualization interaction module and the intelligent report generation module, providing core quantitative data support for subsequent performance trend prediction, visualization display and report generation.

[0026] In this embodiment, the performance prediction module is the core of the AI-based test performance analysis and prediction system. It is a key module for realizing the quantitative assessment of test performance from the current status to the prediction of future development trends. It receives the quantitative performance data from the multi-dimensional assessment module and integrates the full amount of historical performance data accumulated by the system. Through professional time series analysis and machine learning algorithms, it builds and trains a prediction model to generate information on future performance trends and potential abnormal inflection points. It provides forward-looking data support for the pre-decision-making of test performance management and is an important hub connecting the assessment of the current performance status and the prevention and control of future risks. The performance prediction module primarily receives quantitative data such as comprehensive performance scores and scores for each dimension from the multi-dimensional evaluation module. It also retrieves all historical performance data accumulated in the system. Its main tasks include five core operations: historical performance data aggregation and format conversion, prediction model construction and training, future performance trend generation, identification of potential abnormal inflection points, and calculation of confidence intervals for prediction results. The module standardizes and aggregates historical performance data according to a preset time granularity, converts it into a time series format suitable for algorithm training, and uses the Prophet or LSTM algorithm to build and train a time series prediction model. This generates performance trend prediction data within a preset future time range and calculates the confidence intervals of the prediction results to quantify prediction reliability. It also identifies potential abnormal inflection points in performance trends through a mutation point detection algorithm, accurately locating the time nodes and corresponding indicators where performance risks may occur. This module breaks through the limitations of traditional performance analysis, which only focuses on current status assessment, enabling forward-looking prediction of test performance. It provides scientific data basis for the testing team to formulate optimization strategies and prevent performance risks in advance, promoting the transformation of performance management from post-event analysis to pre-event prediction. The performance prediction module includes: Historical performance data management unit: Historical data collection: Real-time retrieval of historical and current quantitative performance data from the multi-dimensional evaluation module, covering scores for each dimension and comprehensive performance scores. At the same time, it integrates the raw performance data and related information from the intelligent data acquisition module to complete the systematic collection of all historical performance data, ensuring the integrity of the data source for model training. Data classification and storage: Collected historical performance data is classified and organized according to project, testing team, and time dimension to establish a standardized performance data warehouse, realize orderly storage and fast retrieval of data, and at the same time, retrieval indexes are established according to data type to improve data retrieval efficiency; Data updates and maintenance: Real-time synchronization of newly added performance evaluation data and original performance data in the system; regular cleaning and verification of historical data; removal of invalid, abnormal, and duplicate data; timely completion of missing data; ensuring the quality of data sources for predictive model training. Time-series data processing unit: Time granularity adaptation: Based on forecasting needs, project characteristics and business scenarios, different time granularities such as daily and weekly are preset. The system can automatically match or manually select the corresponding granularity, and aggregate the cleaned historical performance data according to the selected granularity to make the data meet the basic requirements of time series analysis. Time series format conversion: Convert the aggregated historical performance data into a standardized time series format, sort out the correspondence between the time dimension of the data and the performance indicator dimension, clarify the index and feature columns of the time series, and form a standardized time series dataset that can be directly called for predictive model training; Scientific partitioning of the dataset: The processed time series dataset is divided into a training set and a test set according to a preset ratio. The training set is used for building and training the prediction model, and the test set is used for verifying the effect and evaluating the accuracy of the model after training, so as to ensure the scientific and reasonable nature of the model training process. Prediction model construction and training unit: Intelligent algorithm selection and adaptation: The system provides two algorithm options, Prophet and LSTM. It can automatically analyze the characteristics of time series datasets (including trends, seasonality, nonlinear correlation, etc.) and the time range requirements of prediction tasks, and intelligently match the appropriate algorithm. It also supports manual selection of algorithms based on actual business scenarios, fully adapting to the performance prediction needs of different projects. Flexible model parameter configuration: It provides standardized model parameter configuration items, supports adjusting algorithm parameters according to time series data characteristics and specific prediction targets. At the same time, the system will save high-quality parameter configuration schemes for different business scenarios to form a parameter configuration library, which can be directly called later, greatly improving the efficiency of model building. Model training and accuracy verification: The selected algorithm and configured parameters are systematically trained using the training set to capture the time variation patterns and intrinsic correlations in historical performance data; the trained model is verified using the test set to evaluate the prediction accuracy of the model from multiple dimensions; models whose verification results do not meet the preset accuracy are readjusted and retrained until the prediction accuracy requirements are met. Performance Trend Prediction and Inflection Point Identification Unit: Future performance trend generation: Using a trained and validated prediction model, the test performance is predicted within a preset time range, generating complete future performance trend data including scores for each dimension and a comprehensive performance score, which intuitively reflects the future trend of various performance indicators. Potential anomaly inflection point identification: The mutation point detection algorithm is used to conduct a comprehensive and detailed analysis of the generated future performance trend data, monitor the magnitude and rate of change of performance data in real time, identify potential anomaly inflection points in the trend, and accurately locate the time node of the inflection point and the corresponding performance indicator. Confidence interval calculation for prediction results: Confidence interval is calculated for each performance trend prediction result generated. By combining the fluctuation characteristics of historical data and model training error, the error range and reliability of the prediction results are quantified, so that managers can clearly understand the actual reference value of the prediction results. Prediction result verification unit: Forecast data rationality verification: The generated performance trend forecast results are compared and analyzed from multiple dimensions with the changing patterns of historical performance data and the business characteristics of the current stage of the project to verify whether the forecast results are consistent with the actual business logic. For forecast results that deviate significantly from the actual business, the system automatically re-triggers the model training and trend forecast process. Multi-model cross-validation: If multiple algorithms are used to build prediction models for the same prediction task, the prediction results of different models are cross-validated, and the common parts of the results are extracted as the final prediction results, thereby further improving the accuracy and reliability of the prediction results. Standardization of prediction results: The final performance trend prediction data, potential abnormal inflection point information, and confidence interval data are standardized to form a structured prediction result set, unifying the data format and transmission standards, so that downstream modules can directly call and display them.

[0027] In this embodiment, the visualization interaction module is the core of the human-computer interaction of the AI-based test performance analysis and prediction system. It builds a professional web interactive interface to realize the flexible configuration of system parameters and the intuitive display of performance data in multiple forms. At the same time, it supports diverse interactive data exploration operations, opens up the information interaction link between the system and users, and ensures that managers can efficiently obtain performance analysis and prediction information and accurately carry out in-depth analysis of test performance. The visualization interaction module is primarily responsible for providing a dynamic parameter configuration interface for the entire system process. It receives performance analysis, evaluation, and prediction data transmitted from various backend modules and displays the data dynamically in the form of charts and tables. It also enables interactive operations such as adjusting the time range, filtering evaluation dimensions, and drilling down into the data. This module connects the backend data processing with the frontend user operations, allowing users to easily complete various configurations such as database connections, API keys, and prediction model parameters. Users can intuitively grasp core information such as comprehensive performance scores, scores for each dimension, and performance trend predictions. Furthermore, it allows for personalized interactive exploration of performance data based on business needs, providing efficient and intuitive human-computer interaction support for test performance management decision-making. The visualization interaction module includes: Parameter configuration management unit: Configuration interface setup: Build a standardized web dynamic configuration interface. The interface is divided into database connection configuration area, API key configuration area, and prediction model parameter configuration area according to function. Each configuration area organizes the corresponding configuration items according to system requirements, forming a clear configuration hierarchy and operation logic to adapt to the operating habits of different users. Configuration item entry and verification: Users can manually enter various parameters in the corresponding configuration area. The system performs real-time legality verification on the entered parameters, checks whether the parameter format meets the system requirements and whether the parameter content is complete and valid, and provides clear prompts for parameters that fail the verification to guide users to correct them. Configuration information submission and synchronization: After the user completes the parameter configuration and confirms it, the system will standardize and encapsulate the configuration information and transmit it synchronously to the corresponding backend module through the communication link. At the same time, the system will save the historical version of the configuration information locally, support the modification, rollback and repeated use of the configuration information, and improve the efficiency of parameter configuration. Multi-format visualization display units: Data reception and format adaptation: Real-time reception of structured data transmitted from the multi-dimensional assessment module and performance prediction module, including comprehensive performance score, scores of each dimension, performance trend prediction data, confidence interval data, performance detail data, etc. The received data is processed for front-end format adaptation and converted into a data format that can be directly called by chart and table components. Chart Component Display: Use professional chart components to visualize performance data. Use trend charts to show performance trend prediction results and confidence intervals, intuitively reflecting the future trend of performance indicators. Use radar charts to show the scores of each dimension, clearly showing the differences in performance across different dimensions. All charts support real-time dynamic updates based on data. Table component display: Performance details and personnel details are displayed in a standardized table. The table is divided into columns according to data type, and supports data sorting and pagination to ensure that users can accurately view the specific information of each performance data item. At the same time, the table data and chart data are linked in real time to achieve consistency in data display. Interactive Data Exploration Unit: Time range adjustment: Provides an interactive time range adjustment function. Users can select the start and end time, or directly select a preset time period. The system will re-filter and aggregate the performance data according to the time range selected by the user, and synchronously update the content displayed in the charts and tables to adapt to the performance analysis needs of different time dimensions. Evaluation Dimension Filtering: Set up filtering options for evaluation dimensions, covering all core evaluation dimensions such as response timeliness, defect quality, test coverage, and team collaboration. Users can select dimensions individually or in combination, and the system will display the performance data of the corresponding dimensions based on the filtering results, focusing on the performance analysis priorities that users care about. Data drill-down viewing: Develop an expandable drill-down viewing component. When viewing performance summary data, users can trigger the drill-down function by clicking to view more granular performance data layer by layer, until they can view specific defect details. This enables deep penetration from summary data to detailed data, making it easier to accurately locate the root cause of performance problems.

[0028] In this embodiment, the intelligent report generation module is the core output of the AI-based test performance analysis and prediction system. It automatically integrates the performance analysis, evaluation, prediction data and optimization suggestions from various modules of the system to generate a standardized structured performance analysis report. It also supports multi-format export and systematic management, becoming an important carrier for the implementation of test performance management results and information transmission, ensuring that the performance analysis results can serve management decisions and internal reporting in a standardized and clear form. The intelligent report generation module is primarily responsible for the automatic integration of all performance-related data. Based on preset standardized templates, it generates structured performance analysis reports containing core performance data, multi-dimensional analysis conclusions, future trend predictions, and targeted optimization suggestions. It also supports exporting reports in multiple formats and enables report storage, classification, and retrieval management. This module receives all data from the visualization interaction module, multi-dimensional evaluation module, performance prediction module, and AI analysis engine module, connecting the final link from performance data processing to output. It presents the analysis and prediction results of test performance in a standardized document format, adapting to various business scenarios such as internal reporting, performance management, and historical review, improving the efficiency of performance information transmission and implementation. The intelligent report generation module includes: Report Data Integration Unit: Multi-source data integration: Establish standardized data interfaces with each core module of the system, and integrate in real time with the comprehensive performance score and score data of each dimension of the multi-dimensional evaluation module, the trend prediction results and abnormal inflection point information of the performance prediction module, and the multi-dimensional analysis conclusions and optimization suggestions of the AI ​​analysis engine module. At the same time, synchronize the core performance details data filtered by the user in the visualization interaction module to achieve one-stop collection of all performance-related data. Data cleaning and verification: Consistency verification and redundant data cleaning are performed on the collected multi-source data, the correlation and matching between data in different modules are checked, duplicate and invalid information is removed, and missing data required for report generation is supplemented to ensure that the integrated data is accurate and complete, and to provide reliable data source support for report generation; Data restructuring: Following the business logic of the performance analysis report, the cleaned full data is restructured into sections such as core performance indicators, multi-dimensional analysis results, trend prediction information, and optimization suggestions. At the same time, the relationships between the data in each section are established so that the data organization format fits the content presentation logic of the report. Report Template Management Unit: Standardized template development: Build standardized report templates adapted to test performance analysis scenarios. The templates are divided into report chapters according to a fixed structure, covering report overview, core performance data display, multi-dimensional performance analysis, future performance trend prediction, performance optimization suggestions, data detail appendix, etc. At the same time, the layout format and data display format of each chapter are unified to ensure the standardization of the report; Template customization: Users can personalize the standardized templates according to their actual business needs. They can add or delete report chapters, modify the data display format, and adjust the layout. At the same time, the system will save the user's personalized configuration scheme to form a custom template library, which can be easily reused later to adapt to the report generation needs of different enterprises and projects. Template version management: Systematically manage the version of all report templates, record information such as template creation time, modified content, and modifiers, support template version rollback and comparison, avoid report format chaos caused by template modifications, and ensure the stability and standardization of template use; Structured report generation unit: Automatic data population: Based on the selected report template, the structured and reorganized performance data is automatically populated into the corresponding positions in the template, realizing the quantitative display of core performance indicators, the textual presentation of multi-dimensional analysis results, the embedding of trend prediction information in charts, and the point-by-point summary of optimization suggestions. The main content of the report can be generated without manual intervention. Report content verification: Verify the completeness and logical rationality of the generated preliminary report, check whether each chapter of the report has been filled with data, whether the data matches the analysis conclusions, and whether the trend predictions and optimization suggestions are relevant to the business. Automatically correct any problems found during the verification to ensure that the report content is complete and logically sound. Report content optimization: The text content in the report was optimized based on natural language processing technology. The logic of the expression of the conclusions and optimization suggestions was sorted out to make the text content concise, clear and well-organized. At the same time, the charts in the report were standardized to ensure that the chart styles are consistent and the data is displayed intuitively, thereby improving the overall readability of the report. Multi-format export unit: Multi-format adaptation and conversion: A standardized format conversion engine is built to support the conversion of generated structured performance analysis reports into commonly used document formats to meet the needs of different use cases. During the conversion process, the text content, chart information and layout structure of the report will be completely preserved, ensuring that the report content is not lost and the format is not disordered. Customizable export parameters: Users can personalize the export parameters of the report, select specific chapters of the report to export, adjust the page settings of the exported file, set the export resolution of charts, and batch export multiple performance reports with different templates or different data ranges, improving the flexibility and efficiency of report export. Export file verification: Perform integrity verification on the exported file after format conversion, check whether the text, charts and data in the file are consistent with the original report, and whether the file can be opened and edited normally. Files that fail the verification are re-converted to ensure the usability of the exported file. Report Management and Retrieval Unit: Systematic Report Storage: All generated performance analysis reports are systematically stored, categorized and archived by project name, generation time, report type, and other dimensions. At the same time, the reports are linked with the corresponding raw performance data for storage, enabling one-click linkage viewing of reports and data sources, which facilitates subsequent data analysis and traceability. Multi-condition search query: Build an efficient report search engine that supports users to combine multiple conditions such as project name, generation time, report template type, and core performance indicator range to quickly locate the required performance analysis report. At the same time, it supports keyword search of report content to facilitate accurate retrieval of specific information in the report. Report access control: Fine-grained management of viewing, downloading, and modifying permissions for reports. Different operation permissions are assigned according to the user's system role to ensure the information security of performance reports. At the same time, user operation logs for reports are recorded, including viewing time, download records, and modification content, so as to achieve full traceability of report operations.

[0029] In this embodiment, the system also includes an anomaly detection and early warning module. This module is the core of the AI-based test performance analysis and prediction system for risk prevention and control. Through real-time monitoring of all-dimensional performance indicators and anomaly identification using multi-algorithm fusion, it achieves timely detection, automatic early warning, and root cause analysis of performance anomalies. It serves as a crucial guarantee for proactive risk prevention in test performance management, ensuring that managers can quickly respond to performance anomaly issues and take timely optimization measures to avoid performance decline risks. The anomaly detection and early warning module includes: Performance indicator monitoring unit: Full access to indicators: Establish standardized data interfaces with the multi-dimensional evaluation module, performance prediction module, and visualization interaction module to achieve one-stop access to all performance indicators such as performance scores of each core dimension, comprehensive performance scores, performance trend data, and defect detail related data, covering all core dimensions and related data of test performance evaluation; Real-time data acquisition: The performance indicator data of each interface is collected in real time according to the preset collection frequency to ensure that the collected data is consistent with the latest data of each module of the system. At the same time, the data transmission status during the collection process is monitored to ensure the continuity and integrity of data collection. Indicator Status Monitoring: Continuously monitor the status of all collected performance indicators, track the changes, fluctuations and rates of change of each indicator in real time, and dynamically compare the real-time status of the indicators with the preset normal status benchmark to provide a real-time and accurate data source for subsequent anomaly detection. Multi-algorithm fusion detection unit: Anomaly detection algorithm scheduling: The system has two built-in anomaly detection algorithms: statistical process control and machine learning. The system can automatically schedule the appropriate detection algorithm according to the type of performance indicators, data characteristics and business scenarios. It also supports managers to manually select detection algorithms or configure multi-algorithm fusion detection strategies according to actual needs, so as to improve the adaptability of anomaly detection. Anomaly identification: The selected detection algorithm is used to perform in-depth analysis of real-time data of performance indicators, identify obvious anomalies where the indicator values ​​exceed the preset threshold, and at the same time, uncover hidden anomalies such as irregular fluctuations and trend changes in the indicator data, accurately determine whether the performance indicators are in an abnormal state, and clarify the type and specific manifestation of the anomaly. Detection result verification: Multi-level verification is performed on the abnormal results identified by the algorithm. Combining the historical change patterns of the indicators, the business characteristics of the current stage of the project, and the status of other related indicators, the authenticity and validity of the abnormal results are verified, and false anomalies caused by data collection errors and random fluctuations are eliminated to ensure the accuracy of the anomaly detection results. Intelligent early warning triggering unit: Early warning rule configuration: Provides flexible early warning rule configuration function, allowing managers to set differentiated early warning trigger conditions, early warning push format and push targets according to the type of anomaly, the level of anomaly and the importance of the indicators involved. At the same time, early warning rules can be modified, enabled and disabled in real time to adapt to different test performance management needs. Multi-form early warning push: When a real and valid performance anomaly is detected, the module will automatically trigger an early warning notification, supporting multiple push formats such as system in-site messages, emails, and SMS, accurately pushing abnormal information, including abnormal indicators, abnormal performance, and occurrence time, to preset management personnel to ensure that abnormal information can be received in a timely manner; Warning status tracking: Track the status of each triggered warning notification throughout its entire lifecycle, recording the trigger time, target audience, and reception status. It also allows administrators to mark warnings as read, in progress, or resolved in the system, enabling visualized control over the warning handling process. Root cause analysis suggestion unit: Correlation data mining: When an anomaly in performance is detected, the module will automatically retrieve all data related to the anomaly indicator, including historical change data of the indicator, performance data of related dimensions, defect details, project stage information, etc. Through data correlation analysis, the module will mine the inherent relationship between the anomaly indicator and other data and identify the relevant factors that may lead to the anomaly. Root cause analysis: Based on the results of data correlation analysis and combined with the business logic of software test management, analyze the deep-seated causes and logical chains of performance anomalies, identify the core triggers of the anomalies, and distinguish whether the performance anomalies are caused by operational problems of the test team, test process problems, project phase adjustment problems, or other external factors. Optimization Direction Suggestions: Based on the identified root causes of anomalies, combined with optimization suggestions for the AI ​​analysis engine module and best practices for test performance management, targeted root cause handling suggestions and performance optimization directions are provided, offering clear and feasible reference for managers to formulate anomaly handling strategies and performance optimization measures; Anomaly Information Management Unit: Abnormal data recording: All detected performance anomalies are recorded in detail and completely, including core information such as abnormal indicators, anomaly type, abnormal behavior, occurrence time, detection algorithm, and related data, forming a standardized abnormal data archive to ensure the traceability of abnormal information; Archiving of handling results: Synchronously record the handling process and final results of management personnel in handling performance anomalies, including handling measures, execution time, handling effect, recovery status of performance indicators, etc., and link the handling results with the corresponding anomaly information for archiving, forming a complete data closed loop from anomaly detection to handling; Anomaly Pattern Analysis: Regularly analyze the patterns of all anomaly data and handling results accumulated in the system to uncover the frequency of occurrence, high-incidence phases, related factors, and effective handling measures of different types of anomalies, and generate an anomaly pattern analysis report to provide data support for subsequent optimization of anomaly detection algorithms, adjustment of early warning rules, and improvement of the test performance management system.

[0030] An AI-based test performance analysis and prediction method, based on the core execution logic of an AI-based test performance analysis and prediction system, relies on the collaborative linkage of various core modules of the system to integrate intelligent data collection, deep analysis using AI large-scale models, multi-dimensional quantitative evaluation, time-series trend prediction, visual interactive exploration, and intelligent report generation into a standardized and automated full-process execution system. This overcomes the inefficiencies, subjective biases, and delays inherent in traditional manual test performance analysis, achieving intelligent processing across the entire chain from raw data in the test management system to performance analysis and prediction results. This provides a scientific, accurate, and forward-looking decision-making basis for software test performance management. The method closely adheres to the functional positioning and data flow logic of each module in the system, ensuring seamless data exchange between each execution stage. It combines flexibility and adaptability, allowing for adjustments to execution parameters and analysis strategies based on business needs at different project stages, adapting to the performance management needs of software test projects of various sizes and types. The method includes the following steps: Step S1: Intelligent Data Acquisition and Preprocessing: The test management system database structure is automatically identified through dynamic mapping, defect data is extracted, and cleaned and formatted. Specific operation steps are as follows: The system automatically adapts and connects to the database, establishing a connection with the test management system's database through the system's preset database connection library. It supports two configuration methods: direct connection and connection string. There is no need to manually adjust the underlying database settings, achieving non-intrusive integration with heterogeneous databases. At the same time, the system automatically queries the database system tables, identifies the data table structure and field mapping relationships of the test management system, and completes intelligent database adaptation. The system accurately extracts target data. Based on the identification results of the database structure, it dynamically constructs a data extraction model and uses parameterized queries to accurately extract core data related to test performance, such as defect data, user information, and product information, from the test management system. This ensures the completeness and relevance of the extracted data and eliminates redundant data that is irrelevant to performance analysis. Raw data cleaning and formatting involves a full-process cleaning and standardization preprocessing of the extracted raw data, removing duplicate, missing, and outlier values, standardizing data format and units, classifying and aggregating data from different dimensions, and finally converting the processed raw data into structured data that can be directly called by various modules of the system, ensuring the quality of the data source for subsequent analysis and evaluation steps. Step S2: AI Large Model Interaction and Performance Analysis: Construct structured prompt words, call the large language model API to perform multi-dimensional analysis of performance data, and parse the returned structured analysis results. Specific operations are as follows: The structured prompt word construction system pre-builds standardized prompt word templates that include system role settings, analysis dimension definitions, and output format constraints. It clarifies the exclusive role of the large language model in the test performance analysis scenario, defines fixed analysis dimensions such as response time, defect quality, test coverage, and team collaboration, and standardizes the structure and format of the output content. Then, according to the differences in project type, stage, and analysis object, the prompt word template is dynamically adjusted. The pre-processed aggregated performance data is filled into the template according to the rules to complete the final construction of the prompt words and perform compliance verification. For large model API interaction calls, the system will encapsulate the verified prompt words into a standardized HTTP request, configure the corresponding interface authentication information request parameters timeout control rules and retry mechanism, initiate a call request to the large language model API, and monitor the entire link status of the interface call in real time. In case of call timeout failure, the system will automatically retries and switch to a backup interface to ensure the stability of the call process. The system receives JSON-formatted response content from the large language model API, first verifying the legality and completeness of the content, checking whether it covers all preset analysis dimensions and conforms to format constraints, then performing in-depth analysis on the verified response content to extract three core contents: multi-dimensional analysis results, key performance insights, and improvement suggestions. Finally, this content is deserialized into structured data and converted into a format that downstream modules such as the system's multi-dimensional evaluation module can directly read, achieving seamless flow of analysis results within the system. Step S3: Multi-dimensional assessment and weight calculation: Calculate scores for each dimension based on the preset assessment system, and dynamically adjust the weights according to the project stage to generate a comprehensive performance score. The specific operation is as follows: The system collects data for the evaluation dimensions. It receives the structured analysis results output by the AI ​​analysis engine module in real time, and simultaneously retrieves the standardized raw performance data processed by the intelligent data acquisition module. It completes the full collection of data corresponding to the four core evaluation dimensions of response timeliness, defect quality, test coverage, and team collaboration. At the same time, it verifies the completeness and accuracy of the collected data, removes invalid data, and fills in missing data. For each dimension score calculation, the system follows the preset standardized scoring rules and judgment criteria, traverses all the collected valid performance data, calculates each of the four core evaluation dimensions, obtains the independent score results for each dimension, records the original data support and calculation process corresponding to each dimension score, then verifies the rationality of the calculated scores, traces and corrects abnormal scores that exceed the pre-designed partitions, and determines the final score for each dimension. The system dynamically adjusts the weights of project phases, connects to relevant project management data ports in real time, accurately obtains the project's current lifecycle phase information, identifies which node the project is in, such as requirements review, test case design, integration testing, system testing, deployment, and regression, and then automatically adjusts the weight ratio of each evaluation dimension in the comprehensive performance score according to the preset matching rules of project phase and dimension weights. At the same time, the adjusted weights are normalized and verified to ensure that the weight allocation conforms to the evaluation logic. The system generates a comprehensive performance score by using a weighted average method based on the final scores of each core dimension and the dynamically adjusted dimension weights. This results in the overall performance quantification of the corresponding test project or test team. At the same time, the system converts the final scores of each dimension and the comprehensive performance score into a standardized data format that is compatible with visualization components such as radar charts, providing data support for subsequent visualization. Step S4: Time Series Analysis and Trend Prediction: Train a time series prediction model based on historical performance data to generate future performance trend predictions and information on abnormal inflection points. The specific steps are as follows: The historical performance data is processed in a time series manner. The system retrieves historical and current quantitative performance data from the multi-dimensional evaluation module, covering the comprehensive performance score of each dimension. At the same time, it integrates the raw performance data from the intelligent data acquisition module to complete the collection and cleaning of all historical performance data. Then, the Pandas library is used to aggregate the cleaned historical performance data according to the preset time granularity of daily or weekly, converting it into a standardized time series format. The correspondence between the time dimension of the data and the performance indicator dimension is sorted out, and the processed time series dataset is divided into training set and test set according to the preset ratio. For predictive model building and training, the system automatically matches the appropriate Prophet or LSTM algorithm based on the characteristics of the time series dataset and the time range requirements of the prediction task. It also supports manual selection of algorithms and provides standardized model parameter configuration items. Users can adjust the parameters according to data characteristics and prediction goals. The system uses the training set to systematically train the predictive model with the selected algorithm and configured parameters, captures the time change patterns and internal correlations in historical performance data, and then uses the test set to verify the effect of the trained model. For models whose verification accuracy does not meet the standard, the parameters are readjusted and the model is trained a second time. The system uses a trained and validated prediction model to predict test performance trends within a preset time frame. It generates complete future performance trend data, including scores from various dimensions and a comprehensive performance score. Simultaneously, it combines historical data fluctuation characteristics and model training errors to calculate the confidence interval for each prediction result, quantifying the error range and reliability of the prediction results. This allows managers to clearly understand the reference value of the prediction results. Potential abnormal inflection point identification: The system uses a mutation point detection algorithm to comprehensively analyze the generated future performance trend data, monitor the magnitude and rate of change of performance data in real time, identify potential abnormal inflection points in the trend, and accurately locate the time node of the inflection point and the corresponding performance indicator, providing a precise basis for the pre-emptive prevention and control of performance risks. Step S5: Visualization and Interactive Exploration: Dynamically display analysis results and predicted trends in the web interface, and provide interactive functions such as data drill-down and filtering. Specific operations are as follows: The system analyzes data format adaptation. It receives quantitative scoring data from the multi-dimensional evaluation module and trend prediction data from the performance prediction module. It performs front-end format adaptation processing on this data, converting it into a data format that can be directly called by visualization components such as charts and tables. At the same time, it classifies and stores the data according to data type to ensure the smoothness and accuracy of data display. The system offers multi-format visualizations. In the web interface, it displays performance trend prediction charts and confidence intervals through chart components, scores for each dimension through radar charts, and performance details and personnel details through table components. All visualizations are dynamically updated in real time based on backend data updates to ensure that the displayed content is consistent with the latest analysis and prediction results. Interactive data exploration features a rich set of interactive components on the web interface, allowing users to adjust the time range and filter evaluation dimensions. Users can choose the analysis period and the evaluation dimensions they are interested in. The system re-filters and aggregates performance data based on user actions and updates the displayed content accordingly. The system also provides an expandable drill-down view component, allowing users to view more granular performance data layer by layer through clicks, until they can view specific defect details, achieving deep penetration from summary data to detailed data. Step S6: Intelligent Report Generation and Management: Automatically integrates and analyzes data and improvement suggestions to generate structured performance reports and supports export in multiple formats. The specific steps are as follows: The system integrates multi-source analysis data, establishes standardized data interfaces with each core module, collects the comprehensive performance score and scores of each dimension from the multi-dimensional evaluation module, the trend prediction results and abnormal inflection point information from the performance prediction module, and the multi-dimensional analysis conclusions and optimization suggestions from the AI ​​analysis engine module. At the same time, it synchronizes the core performance details data filtered by the user in the visualization interaction module, performs consistency verification and redundancy cleaning on the collected multi-source data, and then restructures it according to the business logic of the performance analysis report. Structured reports are automatically generated. The system retrieves preset standardized report templates and also supports users to select custom templates. Through a data binding mechanism, the structured and reorganized performance data is automatically filled into the corresponding positions in the template, realizing the quantitative display of core performance indicators, the textual presentation of multi-dimensional analysis results, the embedding of trend prediction information in charts, and the point-by-point summary of optimization suggestions. At the same time, the system verifies the completeness and logical rationality of the generated preliminary report, automatically corrects the problems found, and optimizes the text description of the report and adjusts the chart style based on natural language processing technology. Multi-format report export: The system has built a standardized format conversion engine that supports converting the generated structured performance analysis reports into a variety of commonly used document formats. During the conversion process, the text content, chart information and layout structure of the report are fully preserved. At the same time, users can personalize the export parameters, select specific chapters of the report to be exported, adjust page settings, and also support batch export of multiple reports with different templates. The system performs integrity verification on the exported files after conversion to ensure that the files can be opened and edited normally. The system manages reports systematically, storing all generated performance analysis reports and categorizing them by project name, generation time, report type, and other dimensions. It also links reports with corresponding raw performance data, builds an efficient report retrieval engine, supports multi-condition combined search and keyword search, and can assign viewing, downloading, and modification permissions to reports based on users' system roles. It also records user operation logs for reports, achieving full lifecycle management of reports.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AI-based test performance analysis and prediction system, characterized in that: include: The intelligent data acquisition module is configured to automatically identify and connect to the database of the test management system, extract defect data and related information through a dynamic mapping table structure, and perform cleaning and formatting preprocessing on the raw data. The AI ​​analysis engine module communicates with the intelligent data acquisition module and is configured to perform multi-dimensional performance analysis on preprocessed defect data by calling the large language model API and combining it with preset structured prompt word templates, generating structured analysis results containing key insights and optimization suggestions. The multi-dimensional evaluation module is configured to receive the analysis results from the AI ​​analysis engine module. Based on an evaluation system that includes response timeliness, defect quality, test coverage, and team collaboration, it calculates scores for each dimension and dynamically adjusts the weights according to the project stage to generate a comprehensive performance score. The performance prediction module is configured to acquire historical performance data, build and train a prediction model based on time series analysis or machine learning algorithms, and generate future performance trend prediction results and information on potential abnormal inflection points. The visualization and interaction module is configured to provide a web-based interactive interface, enabling database connection configuration, analysis parameter settings, and dynamic display of comprehensive performance scores, scores for each dimension, and performance trend prediction results in the form of charts and tables. It also supports interactive exploration through data drill-down and dimension filtering. The intelligent report generation module is configured to automatically integrate comprehensive performance scores, scores for each dimension, performance trend prediction results, and optimization suggestions to generate a structured performance analysis report, and supports export in multiple formats.

2. The AI-based test performance analysis and prediction system according to claim 1, characterized in that: The intelligent data acquisition module is further configured to: Establish a connection through the database connection library, and support two configuration methods: direct connection and connection string; By querying the database system tables, the data table structure and field mapping relationships of the test management system are automatically identified; Based on the identification results, a data extraction model is dynamically constructed, and parameterized queries are executed to obtain defect data, user information, and product information.

3. The AI-based test performance analysis and prediction system according to claim 1, characterized in that: The AI ​​analysis engine module is further configured to: Construct structured prompts that include system role settings, analysis dimension definitions, and output format constraints; The aggregated performance data is populated into the structured prompt words, and the large language model API is invoked via an HTTP request; Parse the JSON response returned by the API, extract multi-dimensional analysis results, key performance insights and improvement suggestions, and deserialize them into structured data for use by subsequent modules.

4. The AI-based test performance analysis and prediction system according to claim 1, characterized in that: The multi-dimensional evaluation module is further configured to: Iterate through the performance data and calculate the score for each preset dimension. Obtain information on the current stage of the project and dynamically adjust the weight of each dimension in the overall performance score based on preset rules; The weighted average method is used to calculate the overall performance score, and score data for each dimension are generated for radar chart visualization.

5. The AI-based test performance analysis and prediction system according to claim 1, characterized in that: The performance prediction module is further configured to: Historical performance data is aggregated according to a preset time granularity and converted into a time series format; Train a time series forecasting model using the Prophet algorithm or LSTM algorithm to generate forecast data that includes future performance trends; Calculate the confidence interval of the prediction results and use a mutation point detection algorithm to identify potential abnormal inflection points in the performance trend.

6. The AI-based test performance analysis and prediction system according to claim 1, characterized in that: The visualization interaction module is further configured to: Provides a dynamic configuration interface for database connection parameters, API keys, and prediction model parameters; The chart component displays performance trend forecasts and confidence intervals, while the table component displays performance details and personnel information. Interactive components enable time range adjustment and evaluation dimension filtering, while expandable components provide drill-down access to detailed defect information.

7. The AI-based test performance analysis and prediction system according to claim 1, characterized in that: The system also includes an anomaly detection and early warning module, configured for: Based on statistical process control or machine learning anomaly detection algorithms, monitor fluctuations in performance indicators in real time; When an indicator is detected to exceed a preset threshold or to experience abnormal fluctuations, an early warning notification is automatically triggered, and root cause analysis suggestions based on data correlation analysis are provided.

8. An AI-based method for test performance analysis and prediction, applied to the system according to any one of claims 1 to 7, characterized in that, Includes the following steps: Step S1: Intelligent data acquisition and preprocessing: Automatically identify the database structure of the test management system through dynamic mapping, extract defect data, and clean and format it; Step S2: AI Large Model Interaction and Performance Analysis: Construct structured prompt words, call the large language model API to perform multi-dimensional analysis of performance data, and parse the returned structured analysis results; Step S3: Multi-dimensional assessment and weight calculation: Calculate the scores of each dimension based on the preset assessment system, and dynamically adjust the weights according to the project stage to generate a comprehensive performance score; Step S4: Time Series Analysis and Trend Prediction: Train a time series prediction model based on historical performance data to generate future performance trend predictions and information on abnormal inflection points; Step S5: Visualization and Interactive Exploration: Dynamically display analysis results and predicted trends in the web interface, and provide interactive functions such as data drill-down and filtering; Step S6: Intelligent Report Generation and Management: Automatically integrates analytical data and improvement suggestions to generate structured performance reports and supports export in multiple formats.

9. The AI-based test performance analysis and prediction system and method according to claim 8, characterized in that: Step S2 specifically includes: Construct a structured prompt word template that includes system roles, analysis dimensions, and output format constraints; The preprocessed performance data is populated into the prompt word template and sent to the large language model via API request; Receive and parse the JSON response returned by the API, and extract structured multi-dimensional analysis conclusions and optimization suggestions in natural language description.

10. The AI-based test performance analysis and prediction system and method according to claim 8, characterized in that: Step S4 specifically includes: Use the Pandas library to aggregate historical performance data by day or week and convert it into a time series format; Use the Prophet or LSTM algorithm library to train the model on the time series data; Based on the trained model, generate performance trend predictions for a future preset time range and calculate the confidence interval of the prediction results; The mutation point detection algorithm is used to identify potential performance inflection points within the prediction period.

Images