A method and system for testing the performance of an end-to-end game

Abstract

This invention discloses a performance testing method and system for PC games. After collecting player behavior paths and real-time performance data of the PC game client in a preset game scenario, the method analyzes the collected data to obtain a correlation model between the behavior paths and performance data in the preset game scenario. Based on historical performance data and the correlation model, the method dynamically calculates warning thresholds for each performance indicator in the real-time performance data, thereby providing performance warnings based on the warning thresholds. Finally, based on the performance warning results, the method generates a performance test report for the PC game client in the preset game scenario. In this way, by collecting real-time performance data and combining it with player behavior paths, the method can monitor game performance changes in real time. Based on the correlation model between player behavior paths and performance data, the method can analyze performance issues related to player behavior paths in a preset game scenario, improving the player-specificity and accuracy of performance testing.

Description

Technical Field

[0001] This invention relates to the field of performance testing technology, and in particular to a performance testing method and system for PC games. Background Technology

[0002] In the field of PC game client performance testing, the generation of performance test reports is crucial for evaluating the performance stability of game clients, optimizing the gaming experience, and promptly identifying potential problems.

[0003] Existing performance testing methods typically use fixed warning thresholds. However, the performance metrics of client-side games are often affected by various factors (such as game scenario, number of players, network environment, etc.), and fixed warning thresholds cannot adapt to dynamically changing performance requirements. This leads to a high false alarm rate, making it difficult for development teams to accurately identify the real performance problems and increasing the difficulty of troubleshooting and resolving issues.

[0004] Furthermore, existing performance testing methods typically focus only on static data such as performance metrics (e.g., CPU utilization, memory usage), resulting in performance test results that lack player-specificity and accuracy. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a performance testing method and system for PC games, which can improve the player-specificity and accuracy of performance testing.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A performance testing method for client-side games, including the following steps: S1. Collect player behavior paths and real-time performance data of PC game clients in preset game scenarios; S2. Analyze the player behavior path and the real-time performance data and historical performance data under the preset game scenario to obtain the correlation model between the behavior path and performance data under the preset game scenario. S3. Dynamically calculate the warning threshold of each performance indicator in the real-time performance data based on the historical performance data and the correlation model, and issue a performance warning for the real-time performance data based on the warning threshold of each performance indicator in the real-time performance data. S4. Generate a performance test report of the client game under the preset game scenario based on the performance warning results.

[0007] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: A performance testing system for a client-side game includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps of the aforementioned performance testing method for a client-side game.

[0008] The beneficial effects of this invention are as follows: After collecting player behavior paths and real-time performance data of a PC game client in a preset game scenario, the collected data is analyzed to obtain a correlation model between the behavior paths and performance data in the preset game scenario; based on historical performance data and the correlation model, early warning thresholds for each performance indicator in the real-time performance data are dynamically calculated, thereby achieving performance early warning of real-time performance data based on the early warning thresholds, and then generating a performance test report of the PC game client in the preset game scenario based on the performance early warning results. In this way, by collecting real-time performance data and combining it with player behavior paths, changes in game performance can be monitored in real time, shortening the time for problem discovery and resolution; based on the correlation model between player behavior paths and performance data, performance issues related to player behavior paths can be analyzed for preset game scenarios, improving the player-specificity and accuracy of performance testing. Attached Figure Description

[0009] Figure 1 This is a flowchart of a performance testing method for a client-side game according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a performance testing system for a client-side game according to an embodiment of the present invention; Label Explanation: 1. A performance testing system for PC games; 2. Memory; 3. Processor. Detailed Implementation

[0010] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0011] Please refer to Figure 1 This invention provides a performance testing method for PC games, including the following steps: S1. Collect player behavior paths and real-time performance data of PC game clients in preset game scenarios; S2. Analyze the player behavior path and the real-time performance data and historical performance data under the preset game scenario to obtain the correlation model between the behavior path and performance data under the preset game scenario. S3. Dynamically calculate the warning threshold of each performance indicator in the real-time performance data based on the historical performance data and the correlation model, and issue a performance warning for the real-time performance data based on the warning threshold of each performance indicator in the real-time performance data. S4. Generate a performance test report of the client game under the preset game scenario based on the performance warning results.

[0012] As described above, the beneficial effects of this invention are as follows: After collecting player behavior paths and real-time performance data of a PC game client in a preset game scenario, the collected data is analyzed to obtain a correlation model between the behavior paths and performance data in the preset game scenario; based on historical performance data and the correlation model, the warning thresholds for each performance indicator in the real-time performance data are dynamically calculated, thereby achieving performance warnings for real-time performance data based on the warning thresholds, and then generating a performance test report for the PC game client in the preset game scenario based on the performance warning results. In this way, by collecting real-time performance data and combining it with player behavior paths, changes in game performance can be monitored in real time, shortening the time for problem discovery and resolution; based on the correlation model between player behavior paths and performance data, performance issues related to player behavior paths can be analyzed for preset game scenarios, improving the player-specificity and accuracy of performance testing.

[0013] Further, step S2 includes: Identify the changing trends of each performance indicator in the historical performance data and the real-time performance data under the preset game scenario, and identify the correlation between each performance indicator; By combining the player behavior path, the historical performance data, and the changing trends of the real-time performance data under the preset game scenario, as well as the correlation between various performance indicators, the performance characteristics under the preset game scenario are analyzed. A correlation model between behavioral paths and performance data is established based on the performance characteristics described in the preset game scenario.

[0014] As described above, by identifying the changing trends and correlations of performance indicators and combining them with player behavior paths, we can analyze the performance characteristics of preset game scenarios more precisely, improving the depth and accuracy of performance analysis. Furthermore, by establishing a correlation model based on the performance characteristics of preset game scenarios, we can provide personalized performance optimization suggestions for different game scenarios, thereby improving the accuracy of performance testing.

[0015] Further, step S3, which dynamically calculates the warning thresholds for each performance indicator in the real-time performance data based on the historical performance data and the correlation model, includes: The initial thresholds for each performance indicator are calculated by using machine learning algorithms to perform cluster analysis on each performance indicator of the historical performance data and each performance indicator of the real-time performance data. The initial thresholds of each performance indicator are adjusted according to the correlation model to obtain the early warning thresholds of each performance indicator.

[0016] As described above, using machine learning algorithms to perform cluster analysis on historical and real-time performance data, dynamically calculating the initial thresholds of performance indicators, and adjusting them according to the correlation model can ensure the accuracy and adaptability of the early warning thresholds.

[0017] Furthermore, the step of using machine learning algorithms to perform cluster analysis on various performance indicators of the historical performance data and various performance indicators of the real-time performance data also includes: Obtain game time sequence information of the PC game client in a preset game scenario; Based on the game time series information, the performance indicators of historical performance data and real-time performance data are divided, and cluster analysis is performed on the division results using machine learning algorithms.

[0018] As described above, by combining game time series information to divide and cluster performance data, it is possible to more accurately identify performance characteristics in different time periods and improve the accuracy of early warning thresholds.

[0019] Furthermore, the initial thresholds of each performance indicator are adjusted according to the correlation model to obtain the early warning thresholds for each performance indicator, including: The influence weight of each performance indicator in the real-time performance data is calculated based on the association model. The initial threshold of each performance indicator is adjusted according to the influence weight of each performance indicator to obtain the warning threshold of each performance indicator.

[0020] As described above, by calculating the influence weights of each performance indicator through the correlation model, the initial threshold can be adjusted more specifically to ensure the rationality and accuracy of the warning threshold; by dynamically adjusting the influence weights, the system can adapt to changes in performance data in real time, ensuring the stability and reliability of the warning system in different scenarios.

[0021] Please refer to Figure 2 Another embodiment of the present invention provides a performance testing system for a client-side game, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps of the aforementioned performance testing method for a client-side game.

[0022] The performance testing method and system for PC games described above are applicable to PC game client performance testing. The following detailed implementation methods illustrate these methods: Please refer to Figure 1 Embodiment 1 of the present invention is as follows: A performance testing method for client-side games involves performing the following steps using a real-time streaming data processing framework: S0. Initialize and configure the performance testing system for client-side games.

[0023] S01. Deploy a distributed computing and storage environment: Install and configure a distributed computing framework (such as Apache Spark) and a distributed storage system (such as Ceph or Hadoop HDFS) on the server cluster. Configure Spark cluster parameters, including the number of nodes and resource allocation strategies; set parameters such as block size and number of replicas for the storage system to ensure high availability and scalability to meet the processing and storage needs of large-scale performance data.

[0024] S02. Start the real-time streaming data processing framework: Deploy and start the selected real-time streaming data processing framework (such as Apache Flink), configure the data input source (such as obtaining data from the client performance data acquisition interface of the game client), the data processing flow (including data cleaning, transformation and real-time analysis logic), and the data output destination (such as storing the processing results in a distributed storage system or sending them to the early warning module).

[0025] S03. Initialize the interactive report generation platform: Build the backend server for the web-based interactive report generation platform. Develop API interfaces using a suitable web framework (such as Spring Boot) to receive user operation requests (such as data queries, report generation parameter settings, etc.) and return processing results (such as personalized reports and chart data). The frontend uses HTML, CSS, and JavaScript to build the user interface, integrating visualization libraries (such as Echarts, D3.js) to implement interactive functions (such as drag, zoom, pan, data drill-down, etc.) and event handling.

[0026] S04. Load the Natural Language Generation Model: Train and load a natural language generation model (such as a deep learning-based sequence-to-sequence model). This model has been trained on a large amount of performance data and corresponding text description samples, and can generate natural language reports based on the performance data features. At the same time, configure the model parameters (such as vocabulary, number of hidden layers, etc.) to optimize model performance and ensure that the generated report is fluent, accurate, and readable.

[0027] S1. Collect player behavior paths and real-time performance data in preset game scenarios from the client.

[0028] In this embodiment, the client-side game collects performance data and player behavior paths in real time through a performance data acquisition module. Performance data includes CPU utilization, memory usage, GPU performance parameters, network latency, etc., and is converted according to a predefined data format. The streaming data processing framework's data source (such as a Kafka message queue) receives and caches the converted collected data from the client-side game, ensuring data order and reliability. The streaming data processing framework reads data from the data source and first performs data cleaning to remove invalid data (such as incorrectly formatted data, duplicate data, etc.) and outliers (such as data outside the reasonable range).

[0029] In this embodiment, it is also necessary to collect the player behavior path of the PC game client, such as the player's behavior path from entering the combat area, selecting skill release, interacting with other players or monsters, to leaving the area after the battle ends.

[0030] The collected player behavior paths and real-time performance data are all within a preset game scenario, so as to facilitate subsequent analysis of performance data when players perform different operations in a fixed scenario.

[0031] S2. Analyze the player behavior path and the real-time performance data and historical performance data under the preset game scenario to obtain the correlation model between the behavior path and performance data under the preset game scenario.

[0032] S21. Identify the changing trends of each performance indicator in the historical performance data and the real-time performance data under the preset game scenario, and identify the correlation between each performance indicator.

[0033] Specifically, time series analysis (such as ARIMA model, exponential smoothing, etc.) is used to identify the changing trends of performance indicators (such as rising, falling, periodic fluctuations, etc.), and anomaly detection algorithms (such as isolated forest, LOF local anomaly factor, etc.) are used to identify outliers in performance indicators and mark potential performance problems.

[0034] The correlation between various performance metrics is calculated using methods such as Pearson correlation coefficient and Spearman correlation coefficient to identify which metrics have strong correlations (such as the correlation between CPU utilization and memory usage). The causal relationship between performance metrics is analyzed using methods such as Granger causality test (such as whether network latency causes a decrease in GPU rendering frame rate).

[0035] Typically, a player's behavior path in a game will affect game performance to varying degrees. For example, in a multiplayer online battle scenario, the complete behavior path of a player—from entering the battle area, selecting a skill to release, interacting with other players or monsters, to leaving the area after the battle—involves different performance metrics at different stages. The calculation logic can be set up so that when a player is detected entering the battle scene, the system automatically begins recording changes in key performance metrics such as CPU usage, GPU rendering frame rate, network latency, and memory consumption throughout this behavior path, going beyond simple sliding window statistical calculations.

[0036] Specifically, this analysis examines the interrelationships between these performance metrics when players choose different skill combinations. For instance, does the release of certain high-effect skills cause a sharp drop in GPU rendering frame rate while simultaneously leading to a rapid increase in CPU usage, and does network latency also have an impact in such cases?

[0037] This computational logic helps to more accurately pinpoint the key factors affecting performance in specific game scenarios, rather than simply analyzing overall performance metric statistics. Compared to traditional computational logic, it focuses more on the details of performance changes during the player's actual gameplay experience, providing development teams with more targeted optimization directions, thereby improving overall game performance and the player's gaming experience.

[0038] S22. Analyze the performance characteristics of the preset game scenario by combining the player behavior path, the historical performance data and the real-time performance data change trends, and the correlation between various performance indicators.

[0039] In this embodiment, the player behavior path is divided into multiple stages (such as entering the combat area, selecting skill release, interacting with other players or monsters, leaving the area after the battle, etc.), and the performance data of each stage is recorded. For each behavior path stage, the changing trend and abnormal situation of the performance indicators are analyzed. The impact of the player behavior path on the performance indicators is quantified by regression analysis or decision tree methods.

[0040] For different game scenarios (such as combat, exploration, and multiplayer cooperative scenarios), scene performance features are extracted; based on different player behavior paths, player performance features are extracted, and the impact of these features on performance is analyzed; combined with the hardware configuration of the player's device (such as GPU model, CPU model, memory size, etc.), the impact of device performance on game performance is analyzed, and device performance features are extracted. Multi-dimensional device performance features for preset game scenarios are generated by combining scene performance features, player performance features, and device performance features.

[0041] Specifically, in terms of data feature extraction, in addition to the usual focus on the changing trends of performance indicators, outliers, and correlations between different performance indicators, we also paid special attention to the performance characteristics under different in-game scenarios. For example, we extracted and analyzed performance data for different game maps or levels separately to understand the impact of specific scenarios on PC game performance.

[0042] In data mining, a more in-depth player behavior correlation analysis was employed, combining factors such as players' game preferences, gaming habits, and interaction patterns with other players to uncover potential correlations between these behaviors and performance metrics. For example, the performance of players who prefer teamwork in team-based scenarios was analyzed, or the performance data changes of players who prefer exploring the game world in complex maps were studied.

[0043] In terms of statistical analysis, multi-dimensional time series analysis was introduced. In addition to the conventional analysis of performance metric changes in chronological order, further detailed time series statistical analysis was conducted based on dimensions such as player gaming time periods (e.g., weekday evenings, weekends), and game duration. This allows for a more precise understanding of performance variation patterns across different time periods and game durations, enabling better optimization of game performance and a superior gaming experience. For example, identifying performance bottlenecks during certain time periods or game durations allows for targeted optimization.

[0044] Meanwhile, this embodiment places greater emphasis on close integration with game design and optimization during the data feature extraction, mining, and statistical analysis process. Through in-depth analysis of this data, specific and targeted suggestions are provided for subsequent game updates and content optimization, rather than simply presenting data results. For example, based on the analysis results of performance data, optimization schemes are proposed for specific gameplay or maps to improve the overall smoothness and stability of the game.

[0045] In some embodiments, the time points of player actions within the behavior path are aligned with performance data to analyze performance trends before and after the actions, and machine learning algorithms (such as K-means clustering) are used to identify the impact of different action patterns on performance. Furthermore, time series analysis (such as Fourier transform or ARIMA models) can be used to identify periodic fluctuations in performance data, and time series anomaly detection algorithms (such as isolated forest or LSTM neural networks) can be used to detect abnormal fluctuations in performance data in real time.

[0046] S23. Establish a correlation model between behavior paths and performance data based on the performance characteristics described in the preset game scenario.

[0047] By collecting and analyzing behavioral path data of a large number of players in the same combat scenario, a performance index correlation model based on behavioral path was established.

[0048] S3. Dynamically calculate the warning threshold of each performance indicator in the real-time performance data based on the historical performance data and the correlation model, and issue a performance warning for the real-time performance data based on the warning threshold of each performance indicator in the real-time performance data.

[0049] S31. Using machine learning algorithms, cluster analysis is performed on each performance index of the historical performance data and each performance index of the real-time performance data to calculate the initial threshold of each performance index.

[0050] Specifically, the process involves obtaining game time series information of the client in a preset game scenario; dividing the performance indicators of historical performance data and real-time performance data according to the game time series information; performing cluster analysis on the division results using machine learning algorithms; and dynamically calculating the threshold of performance indicators using machine learning algorithms (such as clustering algorithms, time series analysis algorithms, etc.).

[0051] In this embodiment, based on game time series information (such as different stages and time periods of the game), historical performance data and real-time performance data are divided into multiple time windows. Features of key performance indicators, such as the average, maximum, minimum, and standard deviation of CPU utilization, memory usage, GPU rendering frame rate, network latency, etc., are extracted from each time window.

[0052] Choose an appropriate clustering algorithm (such as K-means, DBSCAN, hierarchical clustering, etc.) to perform cluster analysis on historical and real-time performance data, thereby identifying typical patterns in the performance data. The clustering algorithm divides the performance data into multiple clusters, each representing a typical performance pattern. For example, some clusters may represent a high CPU utilization, low frame rate performance pattern, while others may represent a low CPU utilization, high frame rate performance pattern. For each cluster, calculate its centroid, which is the average value of all data points in the cluster. The cluster center represents the typical performance characteristic of that cluster. Based on the performance metric value of the cluster center, set an initial threshold for each performance metric. For example, the performance metric value of the cluster center can be used as the initial threshold, or the initial threshold can be set based on the cluster boundary values ​​(such as the maximum or minimum value of the performance metric in the cluster).

[0053] S32. Adjust the initial threshold of each performance indicator according to the correlation model to obtain the early warning threshold of each performance indicator.

[0054] In this embodiment, the correlation model is established by analyzing the relationship between player behavior paths and performance data, enabling the quantification of the impact of different player behaviors on performance metrics. For example, some player behaviors (such as unleashing high-effect skills) may lead to a decrease in GPU rendering frame rate, while other behaviors (such as a large number of players entering a certain area simultaneously) may lead to an increase in network latency. The correlation model can quantify the impact weight of different player behaviors on various performance metrics, thereby dynamically adjusting the warning threshold.

[0055] Specifically, the influence weight of each performance indicator in the real-time performance data is calculated according to the association model, and the initial threshold of each performance indicator is adjusted according to the influence weight of each performance indicator to obtain the warning threshold of each performance indicator.

[0056] By using a correlation model, the sensitivity of each performance metric to player behavior paths is analyzed. Feature importance analysis (such as feature importance based on decision trees, SHAP values, etc.) can be used to quantify the impact of each performance metric on player behavior. Based on the results of the feature importance analysis, an influence weight is assigned to each performance metric. The influence weights of all performance metrics are then normalized to ensure that the sum of the weights is 1.

[0057] The initial threshold is adjusted based on the influence weight. Assuming the initial threshold for a performance indicator is T, and its influence weight is w, the adjusted warning threshold T' can be calculated using the following formula: T'=T×(1+α×w) In the formula, α is an adjustment coefficient used to control the degree of influence of the weight on the threshold. α can be adjusted according to actual needs.

[0058] In this embodiment, the performance metrics calculated in real time are compared with dynamic thresholds. If a metric exceeds the threshold range, an intelligent early warning mechanism is triggered. The early warning module sends early warning information to the development team by calling a message push service (such as an email sending service, SMS gateway, or instant messaging platform API), including the name of the abnormal performance metric, its current value, the time of occurrence, and possible causes (based on a predefined rule base).

[0059] S4. Generate a performance test report of the client game under the preset game scenario based on the performance warning results.

[0060] In this embodiment, the performance warning results include various performance indicators and their warning status.

[0061] In one embodiment, operations can be performed on the front-end interface of the interactive report generation platform. Users can select performance metrics of interest by dragging and dropping, specify a time range by sliding a time range selector, and select analysis dimensions (such as by game scene, player region, device type, etc.) by clicking a button. The front-end interface converts the operation request into a JSON-formatted API request and sends it to the back-end server.

[0062] After receiving an API request, the backend server generates a corresponding Spark SQL query based on the request parameters and submits it to the distributed computing framework (Spark). Spark reads performance data from the distributed storage system, filters, aggregates, and performs calculations according to the query, obtaining a performance data result set that meets the user's needs. For example, if a user chooses to view the CPU usage trend for a specific game scenario over the past week, Spark will retrieve the CPU usage data for the corresponding time period from the storage system and calculate statistical information such as the average, maximum, and minimum values.

[0063] The backend server converts the data results calculated by Spark into a data format suitable for the frontend visualization library (such as JSON) and passes it to the frontend. The frontend, based on the data format and the visualization type selected by the user (such as line chart, bar chart, pie chart, etc.), uses the visualization library to generate dynamic charts. Simultaneously, the backend server calls the natural language generation module to convert the data results into a text report describing the data in natural language. Finally, the frontend combines the generated charts and text report and presents them to the user, who can interact with the interface (such as zooming in and out of the chart, panning the timeline, and clicking on data points to drill down for detailed information) to deeply analyze performance trends.

[0064] Furthermore, performance features can be converted into the input format required by the natural language generation model (such as vector representation or sequence form) and input into the trained natural language generation model. The model generates natural language descriptive text based on the input feature information, combined with predefined language templates and grammatical rules. For example, if CPU usage continuously increases over a certain period, the model might generate a text description such as, "During the [time period], CPU usage showed an upward trend, increasing from [initial value] to [end value]. This may indicate that the game client's computational tasks increased during this period, requiring further investigation."

[0065] The generated natural language report undergoes optimization, including grammar checking, semantic coherence adjustment, and vocabulary optimization, to ensure the report is fluent, accurate, and easy to understand. Finally, the optimized report is returned to the backend server, which integrates it into an interactive report for user viewing or stores it directly in the file system for later review and analysis.

[0066] The performance testing method in this embodiment also includes performance monitoring and maintenance steps, specifically: Deploy monitoring tools (such as Prometheus and Grafana) to monitor the performance of the entire system (including distributed computing frameworks, storage systems, streaming data processing frameworks, and report generation platforms) in real time. Monitoring metrics include CPU utilization, memory usage, network traffic, and data processing latency. Based on the monitoring data, promptly identify performance bottlenecks (such as excessive load on a computing node or decreased read / write speed of the storage system) and take corresponding optimization measures (such as adjusting resource allocation, optimizing configuration parameters, and increasing the number of nodes) to ensure that the system always operates at high efficiency.

[0067] Regularly back up performance data in the distributed storage system to off-site storage devices or cloud storage services to prevent data loss. Simultaneously, establish a data recovery mechanism to quickly restore data from backups in the event of corruption or loss, ensuring system reliability and data integrity.

[0068] As technology advances and business needs change, the system should be upgraded and updated promptly. This includes updating the versions of the distributed computing framework, storage system, and streaming data processing framework to obtain new features and performance optimizations; improving the natural language generation model to enhance the quality and accuracy of report generation; and optimizing the user experience of the interactive report generation platform by adding new features and interaction methods. During the upgrade process, it is crucial to ensure system compatibility and stability, avoiding any impact on ongoing performance testing and report generation.

[0069] Please refer to Figure 2 Embodiment two of the present invention is as follows: A performance testing system 1 for a client-side game includes a memory 2, a processor 3, and a computer program stored on the memory 2 and executable on the processor 3. When the processor 3 executes the computer program, it implements the various steps of a performance testing method for a client-side game according to Embodiment 1.

[0070] In summary, the performance testing method and system for PC games provided by this invention collects player behavior paths and real-time performance data of the PC game client under a preset game scenario. It then analyzes the collected data to obtain a correlation model between the behavior paths and performance data under the preset game scenario. Based on historical performance data and the correlation model, it dynamically calculates warning thresholds for each performance indicator in the real-time performance data, thereby implementing performance warnings based on the warning thresholds. Finally, it generates a performance test report for the PC game client under the preset game scenario based on the performance warning results. In this way, by collecting real-time performance data and combining it with player behavior paths, it is possible to monitor game performance changes in real time, shortening the time for problem discovery and resolution. Based on the correlation model between player behavior paths and performance data, it is possible to analyze performance issues related to player behavior paths for preset game scenarios, improving the player-specificity and accuracy of performance testing.

[0071] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A performance testing method for client-side games, characterized in that, Including the following steps: S1. Collect player behavior paths and real-time performance data of PC game clients in preset game scenarios; S2. Analyze the player behavior path and the real-time performance data and historical performance data under the preset game scenario to obtain the correlation model between the behavior path and performance data under the preset game scenario. S3. Dynamically calculate the warning threshold of each performance indicator in the real-time performance data based on the historical performance data and the correlation model, and issue a performance warning for the real-time performance data based on the warning threshold of each performance indicator in the real-time performance data. S4. Generate a performance test report of the client game under the preset game scenario based on the performance warning results.

2. The performance testing method for a client-side game according to claim 1, characterized in that, Step S2 includes: Identify the changing trends of each performance indicator in the historical performance data and the real-time performance data under the preset game scenario, and identify the correlation between each performance indicator; By combining the player behavior path, the historical performance data, and the changing trends of the real-time performance data under the preset game scenario, as well as the correlation between various performance indicators, the performance characteristics under the preset game scenario are analyzed. A correlation model between behavioral paths and performance data is established based on the performance characteristics described in the preset game scenario.

3. The performance testing method for a client-side game according to claim 1, characterized in that, Step S3, which dynamically calculates the warning thresholds for each performance indicator in the real-time performance data based on the historical performance data and the correlation model, includes: The initial thresholds for each performance indicator are calculated by using machine learning algorithms to perform cluster analysis on each performance indicator of the historical performance data and each performance indicator of the real-time performance data. The initial thresholds of each performance indicator are adjusted according to the correlation model to obtain the early warning thresholds of each performance indicator.

4. The performance testing method for a client-side game according to claim 3, characterized in that, The method of using machine learning algorithms to perform cluster analysis on various performance indicators of the historical performance data and various performance indicators of the real-time performance data also includes: Obtain game time sequence information of the PC game client in a preset game scenario; Based on the game time series information, the performance indicators of historical performance data and real-time performance data are divided, and cluster analysis is performed on the division results using machine learning algorithms.

5. The performance testing method for a client-side game according to claim 3, characterized in that, The initial thresholds for each performance indicator are adjusted based on the correlation model to obtain the warning thresholds for each performance indicator, including: The influence weight of each performance indicator in the real-time performance data is calculated based on the association model. The initial threshold of each performance indicator is adjusted according to the influence weight of each performance indicator to obtain the warning threshold of each performance indicator.

6. A performance testing system for a client-side game, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: S1. Collect player behavior paths and real-time performance data of PC game clients in preset game scenarios; S2. Analyze the player behavior path and the real-time performance data and historical performance data under the preset game scenario to obtain the correlation model between the behavior path and performance data under the preset game scenario. S3. Dynamically calculate the warning threshold of each performance indicator in the real-time performance data based on the historical performance data and the correlation model, and issue a performance warning for the real-time performance data based on the warning threshold of each performance indicator in the real-time performance data. S4. Generate a performance test report of the client game under the preset game scenario based on the performance warning results.

7. The performance testing system for client-side games according to claim 6, characterized in that, Step S2 includes: Identify the changing trends of each performance indicator in the historical performance data and the real-time performance data under the preset game scenario, and identify the correlation between each performance indicator; By combining the player behavior path, the historical performance data, and the changing trends of the real-time performance data under the preset game scenario, as well as the correlation between various performance indicators, the performance characteristics under the preset game scenario are analyzed. A correlation model between behavioral paths and performance data is established based on the performance characteristics described in the preset game scenario.

8. The performance testing system for client-side games according to claim 6, characterized in that, Step S3, which dynamically calculates the warning thresholds for each performance indicator in the real-time performance data based on the historical performance data and the correlation model, includes: The initial thresholds for each performance indicator are calculated by using machine learning algorithms to perform cluster analysis on each performance indicator of the historical performance data and each performance indicator of the real-time performance data. The initial thresholds of each performance indicator are adjusted according to the correlation model to obtain the early warning thresholds of each performance indicator.

9. The performance testing system for client-side games according to claim 8, characterized in that, The method of using machine learning algorithms to perform cluster analysis on various performance indicators of the historical performance data and various performance indicators of the real-time performance data also includes: Obtain game time sequence information of the PC game client in a preset game scenario; Based on the game time series information, the performance indicators of historical performance data and real-time performance data are divided, and cluster analysis is performed on the division results using machine learning algorithms.

10. A performance testing system for client-side games according to claim 8, characterized in that, The initial thresholds for each performance indicator are adjusted based on the correlation model to obtain the warning thresholds for each performance indicator, including: The influence weight of each performance indicator in the real-time performance data is calculated based on the association model. The initial threshold of each performance indicator is adjusted according to the influence weight of each performance indicator to obtain the warning threshold of each performance indicator.

Images