Data processing method and device, equipment, medium and program product

Abstract

The invention discloses a data processing method and device, equipment, a medium and a program product, and relates to the field of machine learning. The method comprises the following steps: receiving an object test instruction; obtaining a first test model corresponding to the first capability information based on the object test instruction; and performing a simulation use test on the object parameter information through the first test model to obtain first object test data, the first object test data being used for integrating with object test data corresponding to other capability information to perform test evaluation on the first object. According to the invention, the performance conditions of accounts with different capabilities using the first object can be reflected, so that the test evaluation of the first object is more objective, and the accuracy of the evaluation result is improved.

Description

Technical Field

[0001] Embodiments of the present application relate to the field of machine learning, and in particular, to a data processing method, apparatus, device, medium, and program product. Background Art

[0002] After designing a target object, it is usually necessary to conduct a performance test on the target object and evaluate the usage of the target object in combination with the test data.

[0003] In the related art, the manual test method is usually adopted. The usage data generated after manually using the target object is combined with the usage data to evaluate the performance of the target object.

[0004] However, the manual test method in the related art results in low test efficiency, and the manual test has certain subjectivity, so the accuracy and objectivity of the performance test results are poor. Summary of the Invention

[0005] Embodiments of the present application provide a data processing method, apparatus, device, medium, and program product. By using a test model corresponding to the ability information of different simulated accounts to perform a simulated usage test on a first object, the test efficiency is improved, and the objectivity and accuracy of the final test evaluation can be improved by the automatic test method. The technical solutions are as follows:

[0006] On the one hand, a data processing method is provided. The method includes:

[0007] Receiving an object test instruction for instructing a simulated usage test on a first object in a virtual scene. The object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate the first ability information of the simulated account using the first object during the test, and the object parameter information is the object parameter to be tested of the first object;

[0008] Obtaining a first test model corresponding to the first ability information based on the object test instruction. The first test model is a machine learning model trained by first sample performance data, and the first sample performance data is data obtained by an account using a second object in a virtual scene to exhibit the first ability information during a historical time period;

[0009] Performing a simulated usage test on the object parameter information through the first test model to obtain first object test data.

[0010] On the other hand, a data processing apparatus is provided. The apparatus includes:

[0011] A receiving module, configured to receive an object test instruction for indicating a simulation usage test on a first object in a virtual scene. The object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate first ability information of a simulated account using the first object during the test, and the object parameter information is object parameters of the first object to be tested.

[0012] An obtaining module, configured to obtain a first test model corresponding to the first ability information based on the object test instruction. The first test model is a machine learning model trained by first sample performance data, and the first sample performance data is data obtained by an account with first ability information performing the first ability information in a virtual scene using a second object during a historical time period.

[0013] A testing module, configured to perform a simulation usage test on the object parameter information through the first test model to obtain first object test data.

[0014] On the other hand, a data processing system is provided. The data processing system includes a terminal and a server, and there is a communication connection between the terminal and the server.

[0015] The server is configured to send an object test instruction to the terminal. The object test instruction is used to indicate a simulation usage test on a first object in a virtual scene. The object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate first ability information of an account that simulates using the first object during the test, and the object parameter information is object parameters of the first object to be tested.

[0016] The terminal is configured to receive the object test instruction; obtain a first test model corresponding to the first ability information based on the object test instruction. The first test model is a machine learning model trained by first sample performance data, and the first sample performance data is data obtained by an account with first ability information performing in a virtual scene using a second object during a historical time period; perform a simulation usage test on the object parameter information through the first test model to obtain first object test data; and send the first object test data to the server.

[0017] The server is further configured to receive the first object test data and object test data corresponding to other ability information; integrate the first object test data and the object test data corresponding to other ability information, and perform a test evaluation on the first object.

[0018] On the other hand, a computer device is provided, which includes a processor and a memory. At least one instruction, at least one program, a code set or an instruction set is stored in the memory. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the data processing method described in any one of the above embodiments.

[0019] On the other hand, a computer-readable storage medium is provided. At least one instruction, at least one program, a code set or an instruction set is stored in the storage medium. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the data processing method described in any one of the above embodiments.

[0020] On the other hand, a computer program product or a computer program is provided. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the data processing method described in any one of the above embodiments.

[0021] The beneficial effects brought by the technical solutions provided in the embodiments of the present application at least include:

[0022] After receiving an object test instruction for simulating the use test of a first object in a virtual scene, a first test model corresponding to the first ability simulation information in the object test instruction is obtained, and the object parameter information in the object test instruction is subjected to a simulation use test through the first test model to obtain first object test data. That is to say, in the way of using a machine learning model, according to different ability information corresponding to the simulated use of the first object by the account, the test model corresponding to the ability information is called to perform a simulation use test on the first object, so as to obtain object test data under different ability information. On the one hand, the method of using a machine learning model to perform an ability test on the first object can improve the object performance test efficiency. On the other hand, by dividing the ability information and using different test models for testing, the performance of the first object used by accounts with different abilities can be reflected, making the test evaluation of the first object more objective and improving the accuracy of the evaluation result. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0024] Figure 1 It is a schematic diagram of an implementation environment provided by an exemplary embodiment of the present application;

[0025] Figure 2 It is a flowchart of a data processing method provided by an exemplary embodiment of the present application;

[0026] Figure 3 It is a flowchart of a data processing method provided by another exemplary embodiment of the present application;

[0027] Figure 4 It is a schematic diagram of a test deviation range provided by another exemplary embodiment of the present application;

[0028] Figure 5 It is a comparison diagram of a deviation shooting path and a reference shooting path provided by another exemplary embodiment of the present application;

[0029] Figure 6 It is a schematic diagram of first sample performance data provided by another exemplary embodiment of the present application;

[0030] Figure 7 It is a schematic diagram of bullet position distribution provided by another exemplary embodiment of the present application;

[0031] Figure 8 It is a broken line distribution diagram of object test data corresponding to multiple objects under different ability information provided by another exemplary embodiment of the present application;

[0032] Figure 9 It is a flowchart of a data processing method provided by an exemplary embodiment of the present application;

[0033] Figure 10 It is a schematic diagram of interface call provided by another exemplary embodiment of the present application;

[0034] Figure 11 It is a schematic diagram of a virtual scene provided by an exemplary embodiment of the present application;

[0035] Figure 12 It is an interaction diagram of a data processing system provided by an exemplary embodiment of the present application;

[0036] Figure 13 It is a data flow diagram of a data processing system provided by an exemplary embodiment of the present application;

[0037] Figure 14 It is a structural block diagram of a data processing device provided by an exemplary embodiment of the present application;

[0038] Figure 15 It is a structural block diagram of a computer device provided by an exemplary embodiment of the present application. Detailed implementation manners

[0039] To make the objectives, technical solutions, and advantages of this application clearer, the following will further describe the embodiments of this application in detail with reference to the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts fall within the scope of protection of this application.

[0040] In this application, terms such as "first" and "second" are used to distinguish identical or similar items with basically the same functions. It should be understood that there is no logical or temporal dependence between "first" and "second", nor are the quantity and execution order limited.

[0041] First, a brief introduction to the nouns involved in the embodiments of this application will be given.

[0042] Artificial Intelligence (AI) is a theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines to enable the machines to have the functions of perception, reasoning, and decision-making. Artificial intelligence technology is an interdisciplinary subject with a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-trained model technology, operation / interaction systems, mechatronics, etc. Among them, the pre-trained model, also known as the large model or the foundation model, can be widely applied to downstream tasks in various directions of artificial intelligence after fine-tuning. The software technologies of artificial intelligence mainly include several directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0043] Machine Learning (ML) is an interdisciplinary subject that involves multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent, and its applications cover all fields of artificial intelligence. Machine learning and deep learning usually include technologies such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

[0044] Figure 1 is a schematic diagram of an implementation environment provided by an exemplary embodiment of the present application. As Figure 1 shown, the implementation environment includes a terminal 110, a server 120, and a communication network 130. Among them, the terminal 110 and the server 120 are connected through the communication network 130. Optionally, the communication network 130 can be a wired network or a wireless network, which is not limited herein.

[0045] In an embodiment of the present application, the server 120 is used to send an object test instruction to the terminal 110. Among them, the object test instruction is used to indicate a simulation usage test on a first object in a virtual scene. The object test instruction includes first ability information for indicating an account used to simulate the use of the first object during the test, and object information parameters of the first object to be tested.

[0046] After receiving the object test instruction, the terminal 110 obtains a first test model corresponding to the first ability information based on the object test instruction. Among them, the first test model is a machine learning model trained through first sample performance data. Among them, the first sample performance data is data obtained by an account with first ability information using second data in a virtual scene during a historical time period.

[0047] The object parameter information is subjected to a simulation usage test through the first test model to obtain first object test data, and the first object test data is fed back to the server 120.

[0048] In addition to receiving the first object test data, the server 120 also receives object test data corresponding to other ability information, and integrates the first object test data and the object test data of other ability information, so as to perform a test evaluation on the first object.

[0049] In some alternative embodiments, the terminal 110 is a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart home appliance, a smart vehicle terminal, a smart speaker, a smart voice interaction device, an aircraft, etc., but is not limited thereto.

[0050] It should be noted that the server 120 can be an independent physical server, can also be a server cluster or a distributed system composed of multiple physical servers, and can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0051] Among them, cloud technology refers to a hosting technology that unifies a series of resources such as hardware, software, and networks within a wide area network or a local area network to achieve data calculation, storage, processing, and sharing. Cloud technology is the general term for network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model. It can form a resource pool, be used as needed, and is flexible and convenient. Cloud computing technology will become an important support. The background services of the technical network system require a large amount of computing and storage resources, such as video websites, picture websites, and more portal websites. With the highly developed application of the Internet industry, in the future, each item may have its own identification mark and needs to be transmitted to the background system for logical processing. Data at different levels will be processed separately, and various industry data requires the support of a powerful system background, which can only be achieved through cloud computing. Optionally, the server 120 can also be implemented as a node in the blockchain system.

[0052] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data for analysis, stored data, displayed data, etc.), and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant laws, regulations, and standards in the relevant region.

[0053] Combined with the above introduction and implementation environment, Figure 2 is a flowchart of a data processing method provided by an embodiment of this application. This method can be applied to a terminal, can also be applied to a server, or can be applied to both the terminal and the server at the same time. In this embodiment of the application, this method is described by taking it as an example that it is applied to a terminal. The method includes:

[0054] Step 210, receive an object test instruction.

[0055] Among them, the object test instruction is used to indicate to perform a simulation usage test on the first object in the virtual scene. The object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate the first ability information of the account that simulates the use of the first object during the test, and the object parameter information is the object parameter to be tested of the first object.

[0056] Schematically, the object test instruction refers to an instruction to simulate an account using the first object in a test environment. For example, if the first object is a virtual firearm, the object test instruction is used to indicate that the simulated account uses the virtual firearm to perform a shooting operation.

[0057] In some embodiments, the first ability simulation information refers to the ability information corresponding to using a first object by a simulated account in a test environment. The ability information is implemented as the operation ability of the simulated account to perform interaction operations using the first object, including at least one of ability categories such as operation accuracy, operation speed, and operation coherence.

[0058] Among them, the operation accuracy refers to the offset between the actual operation result and the reference operation result of the first object when performing an interaction operation using the first object. For example: when the first object is a virtual shooting prop, with position a as the reference, the reference operation result of the first object can be realized as hitting point a after shooting with the virtual shooting prop. However, after the account shoots with the first object, it hits a position to the left of point a. That is, the shooting result is left-offset compared to position a. The larger the offset, the lower the operation accuracy. On the contrary, the smaller the offset, the higher the operation accuracy.

[0059] Among them, the operation speed refers to the number of operation times of performing an interaction operation using the first object within a specified time period. For example: when the first object is a virtual shooting prop, the operation speed refers to the number of times of defeating the virtual objects in the game during the process of using the virtual shooting prop to play against the virtual objects in the virtual scene within a specified time period, or the number of virtual objects defeated.

[0060] Among them, the operation coherence refers to the coherence before continuously performing several interaction operations using the first object. For example: when the first object is a virtual bow and arrow, the operation coherence refers to the time interval between the three interaction operations of drawing the bow, aiming, and shooting with the virtual object. The shorter the time interval, the better the operation coherence. On the contrary, the longer the time interval, the worse the operation coherence.

[0061] Schematically, different ability information is used to indicate the performance data of the first object when performing interaction operations using the first object, including at least one of label types such as operation accuracy, operation offset, and operation result.

[0062] In some embodiments, the object parameter information refers to the performance data during the use process of the first object in a virtual screen. For example: when the first object is realized as a virtual firearm, the object parameter information includes at least one of parameter information such as the hit rate of the virtual firearm, the shortest time consumed to defeat the virtual objects in the game, and the offset between the shooting result and the reference shooting position of the first object.

[0063] Optionally, a single object test instruction is received, that is, the first object is tested using a single type of ability information; or, multiple object test instructions are received at one time, and the first ability simulation information in each object test instruction is different. Therefore, the first object is tested using the ability information of different simulated accounts; or, a single object test instruction is received, and the object test instruction includes multiple different ability simulation information.

[0064] Step 220, obtain a first test model corresponding to the first ability information based on the object test instruction.

[0065] Wherein, the first test model is a machine learning model trained by first sample performance data, and the first sample performance data is data obtained by an account with the first ability information using a second object in a virtual scenario during a historical time period.

[0066] Schematically, the first test model refers to a model used to simulate the corresponding execution process during the process of an account using a first object, so as to obtain the usage result of the simulated account using the first object.

[0067] In some embodiments, the test models corresponding to different ability information are also different, and are specifically manifested as at least one of the following situations:

[0068] First, the model types are different. For example, the first test model corresponding to the first ability information is a linear regression model, and the second test model corresponding to the second ability information is a neural network model;

[0069] Among them, the linear regression model is used to determine the relationship between variables, and predicts new dependent variable values by minimizing the difference between the actual observed values and the model predicted values. For example, when the first object is a virtual firearm, the first ability information is the shooting hit rate, and the parameter affecting the shooting hit rate of the virtual firearm is the shooting offset parameter. The larger the offset, the lower the hit rate. Therefore, the relationship between the offset parameter and the shooting hit rate is analyzed through a linear regression model, so as to predict the landing point of the bullet after the virtual firearm fires a bullet.

[0070] Among them, the neural network model is a model used to identify input signals and output corresponding identification results. For example, when the first object is a virtual firearm, the second ability information is shooting stability, and the parameters affecting the shooting stability of the virtual firearm include the recoil of the virtual firearm and the ability of the account to control the recoil when using the virtual firearm to shoot. By pre-training a stability identification model and inputting the recoil of the virtual firearm and the ability of the account to control the recoil each time the virtual firearm is used to shoot into the stability identification model, the stability prediction result of the virtual firearm is output.

[0071] That is, different types of models are selected according to different ability information to test the first object. On the one hand, it can meet the test requirements of the first object for different object parameter information. On the other hand, by selecting the most matching model for testing according to different ability information, the accuracy of the test results can be improved 。

[0072] Second, the model parameter values are different. For example, the first ability information corresponds to the first test model, and the second ability information corresponds to the second test model. The first test model and the second test model belong to two test models with the same model structure but different model parameter values.

[0073] Optionally, the obtaining method of the first test model includes at least one of the following methods:

[0074] First, obtain multiple candidate test models obtained by pre-training. Each candidate test model corresponds to one type of ability information, and select the first ability information as the first test model from the multiple candidate test models according to the first ability information.

[0075] Second, after obtaining the object test instruction, initialize a sample test model, and according to the first ability information in the object test instruction, select an account with the first ability information to use the first sample performance data of the second object in the historical time period to train the sample test model, so as to obtain the trained first test model.

[0076] It should be noted that the above obtaining methods of the first test model are only illustrative examples, and the embodiments of the present application are not limited thereto.

[0077] Illustratively, the first sample performance data refers to the interaction data when an account with the first ability information uses the second object to perform interaction operations in a virtual scenario in the historical time period. For example, when the second object is a virtual grenade, data such as the hit rate, throwing path, and throwing angle of the virtual grenade during the process of the target account using the virtual grenade to fight can all be used as the first sample performance data.

[0078] Optionally, the first object and the second object are the same object; or, the first object and the second object are different objects.

[0079] Step 230, perform a simulation usage test on the object parameter information through the first test model to obtain the first object test data.

[0080] Among them, the first object test data is used to be integrated with the object test data corresponding to other ability information to test and evaluate the first object.

[0081] Illustratively, after obtaining the first test model corresponding to the first ability information, input the object parameter information into the first test model, and output the first object test data corresponding to the first object.

[0082] Among them, the first object test data is only used to simulate the test data generated when an account with the first ability information uses the first object.

[0083] In some embodiments, the current process is used to test the first object corresponding to the first ability information. If testing the object test data corresponding to other ability information, a distributed process structure can be adopted, and multiple different processes can be run in parallel to ensure that the ability information tested in each process is different. Thus, the first object can be tested under different ability information through multiple different processes, and the object test data corresponding to various different ability information can be obtained. Finally, after integrating the object test data corresponding to different ability information, it is used to evaluate the usage performance of the first object under different ability information.

[0084] Optionally, the test evaluation is used to determine the usage type of the first object. For example, if the object test data corresponding to the first object is excellent under various ability information conditions, it can be considered that the first object belongs to the type of object with simple operation. If the difference in the object test data is obvious under different ability information conditions, it is considered that the first object belongs to the type of object with relatively difficult operation. Or, the test evaluation is used to evaluate the usage performance of the first object. If the usage performance status does not meet the pre-set performance requirements, the design parameters of the first object can be adjusted in a timely manner, where the design parameters refer to the parameters used when creating the usage status corresponding to the first object.

[0085] In summary, the data processing method provided by the embodiments of the present application, after receiving an object test instruction for simulating the use test of the first object in a virtual scene, obtains a first test model corresponding to the first ability simulation information in the object test instruction, and performs a simulated use test on the object parameter information in the object test instruction through the first test model to obtain the first object test data. That is, in the way of using a machine learning model, according to the different ability information corresponding to the simulated use of the first object by the account, the test mode corresponding to the ability information is called to perform a simulated use test on the first object, so as to obtain the object test data under different ability information. On the one hand, the method of using a machine learning model to test the ability of the first object can improve the efficiency of object performance testing. On the other hand, by using different test models for testing through the division of ability information, the performance of the first object used by accounts with different abilities can be reflected, making the test evaluation of the first object more objective and improving the accuracy of the evaluation result.

[0086] In an optional embodiment, the acquisition method of the first test model is described in detail. Schematically, please refer to Figure 3, which shows a flowchart of a data processing method provided by an exemplary embodiment of the present application. That is, steps 221 to 222 are further included in step 220. Schematically, as Figure 3 shown, the method includes the following steps.

[0087] Step 221, obtain multiple candidate test models.

[0088] Among them, the multiple candidate test models respectively correspond to different ability information.

[0089] Schematically, before determining the first test model, multiple trained candidate test models are obtained in advance. Among them, each candidate test model corresponds to corresponding ability information, and the multiple candidate test models are used to simulate different performance situations corresponding to account usage objects with different ability information.

[0090] For example: different ability information includes novice (completely unable to use the first object), relatively proficient in operation (able to use the object relatively proficiently when interacting with the object using the object), proficient in operation (operation ability stronger than relatively proficient in operation), and master level (the strongest ability to use the object). Therefore, the novice corresponds to candidate test model 1, relatively proficient in operation corresponds to candidate test model 2, proficient in operation corresponds to candidate test model 3, and master level corresponds to candidate test model 4. Taking operation accuracy as the evaluation criterion, after obtaining the object parameter information corresponding to object a, input the object parameter information into candidate test model 1, and output object test result 1. Input the object parameter information into candidate test model 4, and output object test result 4. Then, from the perspective of the strength of ability information, the operation accuracy of object test result 4 will be higher than that of object test result 1.

[0091] Next, the acquisition method of multiple candidate test models will be described in detail.

[0092] In some embodiments, obtain the reference object data and object parameter information of the second object. The reference object data refers to the reference object data that meets the deviation requirements during the use of the second object; obtain multiple test deviation ranges, and the multiple test deviation ranges respectively correspond to different ability information; based on the reference object data, obtain the reference deviation data corresponding to the second object in multiple test deviation ranges respectively. The reference deviation data refers to the data whose deviation rate relative to the reference object data is within the test deviation range; obtain a sample test model, which is a machine learning model to be trained for associating object parameter information and ability information; based on the object parameter information of the second object and the reference deviation data corresponding to different ability information respectively, train the sample test model respectively to obtain candidate test models corresponding to different ability information respectively.

[0093] First, taking the second object as an example for training to obtain a candidate test model will be described.

[0094] Schematically, the reference object data corresponding to the second object refers to the reference object data when the second object is used under the condition of meeting the deviation requirements during the design of the second object. For example, when the second object is a virtual firearm, the reference object data means that when the virtual firearm is used for shooting, the position where the bullet hits is exactly the same as the position aimed by the virtual firearm, that is, the hit rate is 100%.

[0095] In this embodiment, taking the second object as a virtual firearm as an example, when the virtual firearm is controlled to perform continuous multiple shootings, the hit positions of the multiple shootings will form a reference shooting path, and this reference design path is used as the reference object data of the second object.

[0096] Schematically, the test deviation range refers to the range corresponding to the deviation between the object generated by using the second object and the reference object data, and the test deviation range corresponds to the ability information. The stronger the account ability indicated by the ability information, the smaller the test deviation range. On the contrary, the weaker the account ability indicated by the ability information, the larger the test deviation range.

[0097] In this embodiment, taking the offset angle and the deviation probability as the deviation parameters, the deviation shooting path obtained by offsetting the reference shooting path according to the offset angle and the deviation probability is used as the reference deviation data generated by the virtual firearm within the test deviation range.

[0098] Schematically, please refer to Figure 4 , which shows a schematic diagram of the test deviation range provided by an exemplary embodiment of the present application. As shown in Figure 4 , the current test deviation range coordinate diagram 400 of the second object is displayed. Taking the second object as a virtual firearm as an example, after presetting the reference shooting path corresponding to the virtual firearm, taking the origin 401 in the test deviation range coordinate diagram 400 as the pressure gun point corresponding to the reference shooting path, then when an offset distance occurs on the x-axis and / or y-axis, it is considered that a new deviation shooting path is generated within the test deviation range according to the reference shooting path. The pressure gun point coordinates corresponding to the deviation shooting path can be represented by (_x_n, _y_n), as shown in Figure 4As shown, the current pressure control points 410, 420, 430, 440, 450, and 460 are displayed, corresponding to the deviation shooting paths obtained through different offsets. Among them, the influencing factors for the change of the pressure control points include: the offset angles p1 and p2 are used to represent two different offset angles, the probability pro1 of offsetting once and the probability pro2 of offsetting twice. That is, the origin 401 is the pressure point reached by the virtual firearm after shooting through the reference shooting path. After offsetting through the offset probability, the offset of the pressure point is realized, thereby obtaining the deviation shooting path. Therefore, the pressure control point 410 can be expressed as (1 + p1) * x, with the probability pro1; the pressure control point 420 can be expressed as (1 + p1 + p2) * x, with the probability pro2; the pressure control point 430 can be expressed as (1 + p1) * y; the pressure control point 440 can be expressed as (1 + p1 + p2) * y; the pressure control point 450 can be expressed as (1 - p1) * x; the pressure control point 460 can be expressed as (1 - p1 - p2) * x.

[0099] In this embodiment, the deviation range between the deviation angle of 0.05 and the deviation angle of 1.0 is used as the entire deviation range. Among them, it is divided based on every 0.2 of the deviation angle, so as to obtain the shooting paths of the second object under different angle deviation ranges. Schematically, please refer to Figure 5 , which shows a comparison diagram of the deviation shooting path and the reference shooting path provided by an exemplary embodiment of the present application. As Figure 5 shown, the current comparison diagrams between the deviation shooting paths and the reference shooting paths under a variety of different test deviation ranges are displayed, including the comparison diagram 501 corresponding to the offset angle of 0, the comparison diagram 502 corresponding to the offset angle of 0.2, the comparison diagram 503 corresponding to the offset angle of 0.4, the comparison diagram 504 corresponding to the offset angle of 0.6, the comparison diagram 505 corresponding to the offset angle of 0.8, and the comparison diagram 506 corresponding to the offset angle of 1.0. Among them, the X-axis and Y-axis of each comparison diagram respectively indicate the length and width corresponding to the shooting range. The solid line represents the reference shooting path, the dotted line represents the deviation shooting path, and the dot represents the pressure control point. It can be seen from Figure 5 that the larger the offset angle, the more serious the offset between the deviation shooting path and the reference shooting path. By setting the test deviation range corresponding to the ability information, the reference deviation data is obtained according to the deviation rate of the reference object data within the test deviation range, and the sample test model is supervised and trained with the reference deviation data as the label, so as to obtain the candidate test models corresponding to different ability information respectively. On the one hand, It can improve the accuracy of label generation and make the generated labels conform to the deviation rate corresponding to the ability information ; on the other hand, the sample test model is trained with the reference deviation data obtained based on the deviation rate, improving The matching degree between the high-candidate test model and the ability information is improved, and the model accuracy of the candidate test model is improved .

[0100] In some embodiments, taking the reference object data as a reference, data traversal is performed within the i-th test deviation range to obtain reference deviation data corresponding to the i-th test deviation range, where i is a positive integer.

[0101] Schematically, in the process of obtaining the reference deviation data, according to the i-th test deviation range, all deviation rates within this range are traversed, so as to obtain multiple reference deviation data within the i-th test deviation range.

[0102] That is to say, taking the test deviation range as a coarse-grained scale, further fine-grained division is performed on the test deviation range, so as to traverse multiple deviation rates within the test deviation range, and thus obtain multiple reference deviation data corresponding to this test deviation range. Through the way of fine-grained traversal under the coarse-grained range, It can obtain as much reference deviation data as possible, thereby improving the training accuracy of the subsequent sample test model.

[0103] In some embodiments, the object parameter information of the second object is input into the sample test model, and the first sample performance data is output; based on the reference deviation data corresponding to different ability information and the first sample performance data, the sample test model is trained to obtain candidate test models corresponding to different ability information respectively, where the reference deviation data corresponding to the k-th ability information and the first sample performance data are used to train the sample test model, and the candidate test model corresponding to the k-th ability information is obtained, where k is a positive integer.

[0104] Schematically, in the process of training the sample test model, first, the object parameter information of the second object is input into the sample test model, so as to output the first sample performance data corresponding to the second object. Using the reference deviation data corresponding to different ability information as data labels, the sample test model is trained through the difference between the first sample performance data and the reference deviation data, and the candidate test model corresponding to the ability information corresponding to the reference deviation data is obtained.

[0105] That is to say, by initializing a sample test model, taking the first sample performance data predicted by it for the second object as the prediction result, and training the sample test model multiple times with the reference deviation data corresponding to different ability information, thus Candidate test models corresponding to different ability information are obtained, and the model training efficiency is improved .

[0106] Schematically, please refer to Figure 6 , which shows a schematic diagram of the first sample performance data provided by an exemplary embodiment of the present application. As Figure 6As shown, the first set of sample performance data 600 output after multiple objects input candidate test models corresponding to different ability information is currently displayed. Among them, the first set of sample performance data is implemented as the shortest time to kill (TTK). Among them, different TTK values represent accounts with different segment capabilities. The abscissa is the object category, and the ordinate is the test deviation range.

[0107] Schematically, please refer to Figure 7 , which shows a schematic diagram of bullet position distribution provided by an exemplary embodiment of the present application, including a bullet landing point distribution diagram 710 with a smaller test deviation range and a bullet landing point distribution diagram 720 with a larger test deviation range. From Figure 7 it can be seen that under different test offset ranges (used to represent different ability information), the bullet landing point diagrams fired by the virtual firearm are significantly different; the larger the offset, the less concentrated the bullets, representing weaker account capabilities.

[0108] In some embodiments, obtain the reference performance data corresponding to different ability information of the third object and the object parameter information corresponding to the third object. The reference performance data is the data obtained by accounts with different ability information performing in the virtual scene using the second object during the historical time period; input the object parameter information corresponding to the third object into the candidate test models corresponding to different ability information respectively, and output the predicted performance data corresponding to different ability information respectively; based on the matching degree between the predicted performance data corresponding to different ability information and the reference performance data, screen the candidate test models corresponding to multiple different ability information respectively to obtain a model screening result, and the model screening result is used to determine the candidate test model whose predicted performance data matches the reference performance data.

[0109] Schematically, after training the candidate test models corresponding to different ability information through the first set of sample performance data of the second object, the model accuracy of the candidate test models can also be checked. By obtaining the object parameter information of the third object and the reference performance data corresponding to different ability information of the third object, and then inputting the object parameter information into the candidate test models corresponding to different ability information respectively, match the predicted performance data output by the candidate test model with the reference performance data corresponding to the ability information, and determine the candidate test model with data fit as the candidate performance data finally used for the simulation use test. Among them, the third object is different from the second object. That is, further verifying the candidate test model can ensure a relatively high accuracy of the candidate data model finally applied to the simulation use test. , to ensure the accuracy of the test results .

[0110] Among them, the reference performance data of the third object can be obtained from the real data executed by the account within the historical time period, or the reference performance data can also be predicted by the candidate test models corresponding to different ability information. If the difference between the reference performance data corresponding to the third object and the first sample performance data of the second object meets the difference condition, it can also be considered that the candidate test model can be used for the final simulation usage test.

[0111] Schematically, please refer to Table 1 below. Table 1 shows the performance data obtained by multiple test objects through candidate test models under different ability information. Among them, different ability information is classified into high level, medium level, and low level, and the performance data is implemented as TTK.

[0112] Table 1

[0113] High level Medium level Low level Object 1 0.66 0.58 0.51 Object 2 0.67 0.58 0.48 Object 3 0.62 0.50 0.39 Object 4 0.70 0.66 0.61

[0114] Step 222: Determine the candidate test model that matches the first ability information from multiple candidate test models based on the object test instruction as the first test model.

[0115] Schematically, the object test instruction contains the first ability information. Taking the first ability information as the standard, select the candidate test model that matches the first ability information from multiple candidate test models for the simulation usage test of the first object. That is, by screening out the candidate test model that matches the first ability information for the simulation usage test of the first object, the object test data finally obtained can be based on the test data corresponding to the current first ability information. Improve the accuracy and objectivity of the test 。

[0116] In some embodiments, obtain the first ability label corresponding to the first ability information based on the object test instruction; use the candidate test model corresponding to the first ability label from multiple candidate ability labels as the first test model.

[0117] In this embodiment, different ability information corresponds to different ability labels. For example, ability information 1 corresponds to label 001, and ability information 2 corresponds to label 002. Screening multiple candidate test models according to the ability labels to determine the first object test model can screen the models by label. Improve the model screening efficiency 。

[0118] Schematically, please refer to Table 2, which shows the test data corresponding to multiple objects tested by different object test models. Among them, different object test models are distinguished as gun god model, gun king model, master model, and expert model.

[0119] Table 2

[0120]

[0121]

[0122] Schematic. Please refer to Figure 8 , which shows the object test data discount distribution diagrams respectively corresponding to multiple objects under different ability information provided by an exemplary embodiment of the present application. As Figure 8 shown, four discount lines are currently displayed, which respectively correspond to the TTK value distribution broken line diagram 801 of different virtual firearms under the Gun God model, the TTK value distribution broken line diagram 802 of different virtual firearms under the Gun King model, the TTK value distribution broken line diagram 803 of different virtual firearms under the Master model, and the TTK value distribution broken line diagram 804 of different virtual firearms under the Expert model. Among them, for the target firearm corresponding to the discount distribution area 810, due to the large difference in TTK values corresponding to different ability information, this target firearm belongs to a virtual firearm with relatively high operation requirements.

[0123] In summary, for the data processing method provided by the embodiment of the present application, after receiving an object test instruction for simulating the use test of a first object in a virtual scene, a first test model corresponding to the first ability simulation information in the object test instruction is obtained, and the object parameter information in the object test instruction is simulated and used for testing through the first test model to obtain first object test data. That is, in the way of using a machine learning model, according to different ability information corresponding to the use of the first object in the simulated account, the test mode corresponding to the ability information is called to simulate the use test of the first object, so as to obtain object test data under different ability information. On the one hand, the way of using a machine learning model to test the ability of the first object can improve the efficiency of object performance testing. On the other hand, by dividing the ability information and using different test models for testing, the performance of using the first object by accounts with different abilities can be reflected, making the test evaluation of the first object more objective and improving the accuracy of the evaluation result.

[0124] In an optional embodiment, the object test instruction is described in detail. Schematic. Please refer to Figure 9 , which shows the flowchart of the data processing method provided by an exemplary embodiment of the present application. That is, 2201 is further included before step 210, and steps 231 and 232 are further included in step 230. As Figure 8 shown, the method includes the following steps.

[0125] Step 2201, start the environment operation data corresponding to the virtual environment based on the object test instruction, and mask the environment rendering data corresponding to the virtual environment.

[0126] Among them, the environment operation data is used to run the virtual environment.

[0127] Schematically, when the terminal receives an object test instruction, it launches the client corresponding to the object test instruction and runs the data corresponding to the client, so as to perform a performance test on the first object in the virtual environment during the process of running the client. Among them, the virtual environment refers to the environment corresponding to the client when it is running.

[0128] In this implementation, in order to reduce data overhead, since automated testing is performed through the first test model, after receiving the object test instruction, only the environment operation data corresponding to the virtual environment in the client is started to support the operation logic of the virtual environment, and subsequent related operations are performed in the virtual environment. For the environment rendering data corresponding to the virtual environment, a shielding operation can be performed, that is, the environment rendering data corresponding to the virtual environment is not rendered. Thus, in the case where there is no need to display the screen, rendering the screen data is avoided, and the normal operation of the virtual environment can still be ensured. This client can also be called a non-rendering client (Norender).

[0129] In this embodiment, when the server issues a test task, in the deployed multi-window (Windows) environment, multiple clients are launched (for batch control and control traversal during testing, the computer version packaged by UE / Unity is used, and the code logic is exactly the same as the experience of the mobile terminal); the client will correspondingly start a Norender client (this client is a client that does not render the screen under UE, and only executes the client logic, which is exactly the same as a normal game. The implementation method is when outputting the rendered screen, Mask the corresponding rendering function, which can greatly reduce the CPU and memory consumption ) and start the corresponding test training.

[0130] Step 231, run the first test model through the environment operation data.

[0131] Schematically, by running the environment operation data, the virtual environment is run, so as to run the first test model in the virtual environment.

[0132] Step 232, input the object parameter information into the first test model and output the first object test data.

[0133] In this embodiment, the object operation interface is called, and the object operation interface includes object operation instructions; the object parameter information and the object operation instructions are input into the first test model, and the first object test data is output.

[0134] In this embodiment, taking the control of weapon firing as an example, based on the project code corresponding to the virtual environment, by binding the corresponding message interface, it is used to send the object operation instruction corresponding to the first object through the message interface; by locally packaging it into a new client interface + PB protocol to communicate with the server, and externally calling interface instructions such as firing and moving, finally, the object operation instruction and the object parameter information are jointly input into the first test model for predicting the performance data, and finally the first object test data is obtained, that is, This method can provide a variety of different operation instructions during the test, improving the comprehensiveness of the test data 。

[0135] Schematically, please refer to Figure 10 , which shows a schematic diagram of interface call provided by an exemplary embodiment of the present application. Currently, the client 1000 is displayed. In the client 1000, it includes the client project code content 1010, the interface list 1020, and the PB protocol module 1030.

[0136] Schematically, please refer to Figure 11 , which shows a schematic diagram of a virtual scene provided by an exemplary embodiment of the present application. As Figure 11 shown, the virtual environment 1100 is currently displayed. In the virtual environment 1100, operations such as firing and moving are performed by controlling the virtual firearm 1120, so as to obtain the object test data corresponding to the virtual firearm.

[0137] In some embodiments, the object test instruction further includes second ability simulation information, and the second ability simulation information is used to indicate the second ability information of using the first object when simulating an account during the test. Based on the object test instruction, a second test model corresponding to the second ability information is obtained. The second test model is a machine learning model trained by second sample performance data, and the second sample performance data is the data obtained when the account uses the third object to perform the second ability information in the virtual scene during the historical time period; the object parameter information is simulated and used for testing through the second test model to obtain the second object test data; the first object test data and the second object test data are integrated to obtain a data integration result; based on the data integration result, the first object is tested and evaluated to obtain an object test result, and the object test result refers to the performance situation of the first object corresponding to different ability information.

[0138] Schematically, taking the example that the object test instruction includes multiple different ability information for illustration. The object test instruction also includes second ability simulation information different from the first ability simulation information, so as to obtain a corresponding second test model according to the second ability information corresponding to the second ability simulation information, and perform a simulation usage test on the object parameter information of the first object through the second test surface model, and finally obtain second object test data. Integrate the first object test data and the second object test data to obtain a data integration result corresponding to the first object, where the data integration result refers to the result of the object parameter information corresponding to the first object under different ability information. For example: the first object is a virtual firearm, and the first ability information refers to the shooting accuracy when a novice player account uses the virtual firearm. Therefore, the first object test data refers to the shooting results when a novice player account uses the virtual firearm for shooting at different distances. The second ability information refers to the shooting accuracy when a technical player account (referring to an account that can proficiently use the virtual firearm) uses the virtual firearm. Therefore, the second object test data refers to the shooting results when a technical player account uses the virtual firearm for shooting at different shooting distances. Therefore, after integrating the first object test data and the second object test data, the shooting accuracy corresponding to the use of the virtual firearm by player accounts with different abilities at different shooting distances is obtained, so as to evaluate the design rationality of the virtual firearm. If the shooting accuracy corresponding to the use of the virtual firearm by player accounts with different abilities at different shooting distances varies greatly, and the shooting accuracy of the technical player account is higher than that of the novice player account, it means that the virtual firearm is more difficult to use.

[0139] In this embodiment, by integrating the object test data corresponding to different ability information, the first object can be comprehensively evaluated through an entire data integration result, improving the accuracy and objectivity of the evaluation result.

[0140] In summary, the data processing method provided in the embodiments of the present application, after receiving an object test instruction for simulating the use test of a first object in a virtual scene, obtains a first test model corresponding to the first ability simulation information in the object test instruction, and performs a simulation use test on the object parameter information in the object test instruction through the first test model to obtain first object test data. That is, in the form of a machine learning model, according to different ability information corresponding to the simulated use of the first object by the account, the test mode corresponding to the ability information is called to perform a simulation use test on the first object, so as to obtain object test data under different ability information. On the one hand, the method of using a machine learning model to perform an ability test on the first object can improve the object performance test efficiency. On the other hand, by dividing the ability information and using different test models for testing, the performance of the first object used by accounts with different abilities can be reflected, making the test evaluation of the first object more objective and improving the accuracy of the evaluation result.

[0141] Schematically, please refer to Figure 12 , which shows an interaction diagram of a data processing system provided by an exemplary embodiment of the present application. As Figure 12 shown, the current content includes the following.

[0142] Step 1210, send an object test instruction to the terminal.

[0143] Among them, the object test instruction is used to indicate a simulation use test of a first object in a virtual scene. The object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate the first ability information of the account that simulates the use of the first object during the test, and the object parameter information is the object parameter to be tested of the first object.

[0144] Schematically, the object test instruction refers to an instruction for simulating the use of a first object by an account in a test environment. For example, when the first object is a virtual firearm, the object test instruction is used to indicate simulating an account using the virtual firearm to perform a shooting operation.

[0145] In some embodiments, the first ability simulation information refers to the ability information corresponding to the account that simulates the use of the first object in a test environment. The ability information is implemented as the operation ability to perform an interactive operation using the first object, including at least one of ability categories such as operation accuracy, operation speed, and operation coherence.

[0146] Among them, the operation accuracy refers to the offset between the actual operation result when using the first object to perform an interactive operation and the benchmark operation result of the first object. For example: when the first object is a virtual shooting prop, with position a as the benchmark, the benchmark operation result of the first object can be realized as that after shooting with the virtual shooting prop, point a will be hit. However, after the account uses the first object to shoot, it hits a position to the left of point a. That is, the shooting result is left-offset compared to position a. The larger the offset, the lower the operation accuracy. On the contrary, the smaller the offset, the higher the operation accuracy.

[0147] Among them, the operation speed refers to the number of operation times when using the first object to perform an interactive operation within a specified time period. For example: when the first object is a virtual shooting prop, the operation speed refers to the number of times of defeating the virtual objects in the game during the process of using the virtual shooting prop to play against the virtual objects in the virtual scene within a specified time period, or the number of virtual objects defeated.

[0148] Among them, the operation coherence refers to the coherence before continuously performing several interactive operations using the first object. For example: when the first object is a virtual bow and arrow, the operation coherence refers to the time interval between the three interactive operations of using the virtual object to draw the bow, aim, and shoot. The shorter the time interval, the better the operation coherence. On the contrary, the longer the time interval, the worse the operation coherence.

[0149] Schematically, different ability information is used to indicate the performance data of the first object when performing interactive operations using it, including at least one of the following types: operation accuracy, operation offset, operation result, etc.

[0150] In some embodiments, the object parameter information refers to the performance data during the use process of the first object in the virtual screen. For example: when the first object is a virtual firearm, the object parameter information includes at least one of the following parameter information: the hit rate of the virtual firearm, the shortest time consumed to defeat the virtual objects in the game, the offset between the shooting result and the benchmark shooting position of the first object.

[0151] Optionally, receive an object test instruction once, that is, test the first object using a single type of ability information; or, receive multiple object test instructions once, and the first ability simulation information in each object test instruction is different. Therefore, test the first object using the ability information of different simulated accounts.

[0152] Step 1220, receive an object test instruction.

[0153] The terminal receives an object test instruction in each process by starting at least one process.

[0154] Step 1230: Obtain a first test model corresponding to the first ability information based on the object test instruction.

[0155] Among them, the first test model is a machine learning model trained through first sample performance data, and the first sample performance data is data obtained from an account with the first ability information using a second object in a virtual scenario during a historical time period.

[0156] Illustratively, the first test model refers to a model used to simulate the corresponding execution process during the process of an account using a first object, so as to obtain the usage result of simulating the account using the first object.

[0157] In some embodiments, the test models corresponding to different ability information are also different, which are specifically manifested as at least one of the following situations:

[0158] First, the model types are different. For example, the first test model corresponding to the first ability information is a linear regression model, and the second test model corresponding to the second ability information is a logistic regression model.

[0159] Second, the numerical values of the model parameters are different. For example, the first ability information corresponds to a first test model, and the second ability information corresponds to a second test model. The first test model and the second test model are two test models with the same model structure but different model parameter values.

[0160] Optionally, the acquisition method of the first test model includes at least one of the following methods:

[0161] First, obtain multiple candidate test models that have been pre-trained. Each candidate test model corresponds to a type of ability information, and select the first ability information as the first test model from the multiple candidate test models according to the first ability information.

[0162] Second, after obtaining the object test instruction, initialize a sample test model, and select the first sample performance data of an account with the first ability information using the second object in a virtual scenario during a historical time period according to the first ability information in the object test instruction to train the sample test model, so as to obtain the trained first test model.

[0163] It should be noted that the above acquisition methods of the first test model are only illustrative examples, and the embodiments of the present application are not limited thereto.

[0164] Illustratively, the first sample performance data refers to the interaction data of an account with the first ability information using a second object in a virtual scenario during a historical time period for interactive operations. For example, when the second object is a virtual grenade, data such as the hit rate, throwing path, and throwing angle of the virtual grenade during the process of the target account using the virtual grenade for combat can all be used as the first sample performance data.

[0165] Optionally, the first object and the second object are the same object; alternatively, the first object and the second object are different objects.

[0166] Step 1240: Perform a simulation usage test on the object parameter information through the first test model to obtain first object test data.

[0167] Illustratively, after obtaining the first test model corresponding to the first capability information, input the object parameter information into the first test model, and output the first object test data corresponding to the first object.

[0168] Among them, the first object test data is only used to simulate the test data generated when an account with the first capability information uses the first object.

[0169] In some embodiments, the current process is used to test and obtain the first object test corresponding to the first capability information. If testing the object test data corresponding to other capability information, a distributed process structure can be adopted to run multiple different processes in parallel, ensuring that the capability information tested in each process is different, so as to test the first object under different capability information through multiple different processes and obtain object test data corresponding to various different capability information.

[0170] Step 1250: Send the first object test data to the server.

[0171] When the terminal obtains the first object test data, it sends it to the server.

[0172] Step 1260: Receive the first object test data and the object test data corresponding to other capability information.

[0173] Step 1270: Integrate the first object test data and the object test data corresponding to other capability information, and perform a test evaluation on the first object.

[0174] After integrating the object test data corresponding to different capability information respectively, it is used to evaluate the usage performance of the first object under different capability information.

[0175] Optionally, the test evaluation is used to determine the usage type of the first object. For example, if the corresponding object test data of the first object performs excellently under various ability information conditions, it can be considered that the first object belongs to the type of object with simple operation. If the difference in object test data is obvious under different ability information conditions, it is considered that the first object belongs to the type of object with difficult operation. Or, the test evaluation is used to evaluate the usage performance status of the first object. If the usage performance status does not meet the pre-set performance requirements, the design parameters of the first object can be adjusted in a timely manner. Herein, the design parameters refer to the parameters used when creating the usage status corresponding to the first object.

[0176] In summary, for the data processing method provided by the embodiment of the present application, after receiving an object test instruction for performing a simulation usage test on a first object in a virtual scene, a first test model corresponding to the first ability simulation information in the object test instruction is obtained, and the object parameter information in the object test instruction is subjected to a simulation usage test through the first test model to obtain first object test data. That is, in the way of using a machine learning model, according to different ability information corresponding to the simulation of using the first object with an account, the test mode corresponding to the ability information is called to perform a simulation usage test on the first object, so as to obtain object test data under different ability information. On the one hand, the method of using a machine learning model to perform an ability test on the first object can improve the efficiency of object performance testing. On the other hand, by dividing the ability information and using different test models for testing, the performance of using the first object by accounts with different abilities can be reflected, making the test evaluation of the first object more objective and improving the accuracy of the evaluation result.

[0177] Schematically, please refer to Figure 13 , which shows the data flow diagram of the data processing system provided by an exemplary embodiment of the present application, as Figure 13As shown, the current test system includes an intelligent server 1310. After the intelligent server 1310 issues a test task, it launches multiple clients, including client 1311, client 1312, and client 1313. Client 1311, client 1312, and client 1313 respectively launch their corresponding non-rendering clients 1314, non-rendering client 1315, and non-rendering client 1316 through a specified communication method (Py-Socket-8000, where 8000 is the port number). Among them, each non-rendering client corresponds to an agent, which is used to launch a local client for initializing virtual objects and entering the game. The system 1300 includes a game server, which is used to pull virtual games for non-rendering clients 1314, non-rendering client 1315, and non-rendering client 1316 after the intelligent server 1310 issues a test task, so as to test the first object in the virtual game and obtain the first test data of the first object. Finally, the intelligent server 1310 sends the first test data to the analysis module 1320 to evaluate the first object by combining the object test data with different ability information.

[0178] Illustratively, taking object performance testing for testing virtual guns as an example, by using the TTK of the virtual gun as an evaluation index, in the model training stage, candidate test models corresponding to different segment pressure capabilities are trained, so as to use the candidate test models corresponding to different segment pressure capabilities to perform recoil control testing on the virtual gun, and obtain the TTK value, segment recoil control angle, and bullet trajectory corresponding to the virtual gun under different segment capabilities, so as to evaluate the usage difficulty corresponding to the virtual gun. Among them, in the training stage, starting from the benchmark TTK value, all deviation data within the deviation range are traversed, so as to adjust the model parameters of the candidate test model. Taking the adjustment of the candidate model parameters with a lower segment as an example, it includes operations such as reducing accuracy (increasing the single-angle offset of recoil control) and reducing stability (increasing the randomness parameter), so as to output candidate test models with different segment capabilities. And use the performance data corresponding to other guns to verify the candidate test models. When the TTK values under different segment capabilities for the same gun are all consistent, the candidate test model is used as the candidate test model for simulation usage testing.

[0179] In summary, for the data processing method provided in the embodiments of the present application, after receiving an object test instruction for simulating the use test of a first object in a virtual scene, a first test model corresponding to the first ability simulation information in the object test instruction is obtained, and the object parameter information in the object test instruction is subjected to a simulated use test through the first test model to obtain first object test data. That is, in the form of a machine learning model, according to the different ability information corresponding to the simulated use of the first object by the account, the test model corresponding to the ability information is called to perform a simulated use test on the first object, so as to obtain object test data under different ability information. On the one hand, the method of using a machine learning model to perform an ability test on the first object can improve the efficiency of object performance testing. On the other hand, by dividing the ability information and using different test models for testing, the performance of the first object used by accounts with different abilities can be reflected, making the test evaluation of the first object more objective and improving the accuracy of the evaluation result.

[0180] Figure 14 FIG. is a structural block diagram of a data processing device provided by an exemplary embodiment of the present application. As Figure 14 shown, the device includes the following parts:

[0181] A receiving module 1410, configured to receive an object test instruction for instructing a simulated use test of a first object in a virtual scene. The object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate the first ability information of the account that simulates the use of the first object during the test, and the object parameter information is the object parameter to be tested of the first object;

[0182] An obtaining module 1420, configured to obtain a first test model corresponding to the first ability information based on the object test instruction. The first test model is a machine learning model trained through first sample performance data, and the first sample performance data is data obtained from the performance of an account with first ability information using a second object in a virtual scene during a historical time period;

[0183] A testing module 1430, configured to perform a simulated use test on the object parameter information through the first test model to obtain first object test data, where the first object test data is used to integrate with object test data corresponding to other ability information to perform a test evaluation on the first object.

[0184] In some embodiments, the obtaining module 1420 is configured to obtain a plurality of candidate test models, where the plurality of candidate test models respectively correspond to different capability information; and determine, based on the object test instruction, a candidate test model that matches the first capability information from the plurality of candidate test models as the first test model.

[0185] In some embodiments, the plurality of candidate test models respectively correspond to different candidate capability labels;

[0186] The obtaining module 1420 is configured to obtain, based on the object test instruction, a first capability label corresponding to the first capability information; and use, as the first test model, the candidate test model corresponding to the first capability label from the plurality of candidate capability labels.

[0187] In some embodiments, the obtaining module 1420 is configured to obtain reference object data and object parameter information corresponding to the second object, where the reference object data refers to reference object data that meets the deviation requirements during the use of the second object; obtain a plurality of test deviation ranges, where the plurality of test deviation ranges respectively correspond to different capability information; obtain, based on the reference object data, reference deviation data respectively corresponding to the second object within the plurality of test deviation ranges, where the reference deviation data refers to data whose deviation rate relative to the reference object data is within the test deviation range; obtain a sample test model, where the sample test model is a machine learning model to be trained for associating object parameter information with capability information; and train the sample test model respectively based on the object parameter information of the second object and the reference deviation data respectively corresponding to different capability information to obtain candidate test models respectively corresponding to different capability information.

[0188] In some embodiments, the obtaining module 1420 is configured to perform data traversal within the i-th test deviation range with the reference object data as a benchmark to obtain reference deviation data corresponding to the i-th test deviation range, where i is a positive integer.

[0189] In some embodiments, the obtaining module 1420 is configured to input the object parameter information of the second object into the sample test model and output first sample performance data; and train the sample test model based on the reference deviation data respectively corresponding to different capability information and the first sample performance data to obtain candidate test models respectively corresponding to different capability information, where after training the sample test model with the reference deviation data corresponding to the k-th capability information and the first sample performance data, a candidate test model corresponding to the k-th capability information is obtained, where k is a positive integer.

[0190] In some embodiments, the obtaining module 1420 is configured to obtain reference performance data corresponding to different ability information of a third object and object parameter information corresponding to the third object, where the reference performance data is data obtained by accounts with different ability information using a second object to perform respectively in a virtual scene during a historical time period; input the object parameter information corresponding to the third object into candidate test models respectively corresponding to the different ability information, and output predicted performance data respectively corresponding to the different ability information; based on the matching degree between the predicted performance data respectively corresponding to the different ability information and the reference performance data, screen the candidate test models respectively corresponding to the multiple different ability information to obtain a model screening result, where the model screening result is used to determine a candidate test model whose predicted performance data matches the reference performance data.

[0191] In some embodiments, the obtaining module 1420 is configured to start environment operation data corresponding to the virtual environment based on the object test instruction, and mask environment rendering data corresponding to the virtual environment, where the environment operation data is used to run the virtual environment;

[0192] The testing module 1430 is configured to run the first test model through the environment operation data; input the object parameter information into the first test model, and output first object test data.

[0193] In some embodiments, the testing module 1430 is configured to call an object operation interface, where the object operation interface includes object operation instructions; input the object parameter information and the object operation instructions into the first test model, and output first object test data.

[0194] In summary, after receiving an object test instruction for performing a simulation usage test on a first object in a virtual scene, the data processing device provided by the embodiments of the present application obtains a first test model corresponding to the first ability simulation information in the object test instruction, and performs a simulation usage test on the object parameter information in the object test instruction through the first test model to obtain first object test data. That is, in a manner of using a machine learning model, according to different ability information obtained by simulating an account using the first object, a test model corresponding to the ability information is called to perform a simulation usage test on the first object, so as to obtain object test data under different ability information. On the one hand, the method of using a machine learning model to perform an ability test on the first object can improve the efficiency of object performance testing. On the other hand, by dividing the ability information and using different test models for testing, the performance of the first object used by accounts with different abilities can be reflected, making the test evaluation of the first object more objective and improving the accuracy of the evaluation result.

[0195] It should be noted that: The data processing device provided in the above embodiments is only illustrated by the division of the above functional modules. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the data processing device provided in the above embodiments and the embodiments of the data processing method belong to the same concept. For the specific implementation process, please refer to the method embodiments and will not be elaborated here.

[0196] Figure 15 FIG. 4 shows a block diagram of a computer device 1500 provided by an exemplary embodiment of the present application. The computer device 1500 may be a portable mobile terminal, such as: a smart phone, a tablet computer, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a notebook computer or a desktop computer. The computer device 1500 may also be referred to by other names such as a user device, a portable terminal, a laptop terminal, a desktop terminal, etc.

[0197] Generally, the computer device 1500 includes a processor 1501 and a memory 1502.

[0198] The processor 1501 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 1501 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1501 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 1501 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 1501 may further include an AI (Artificial Intelligence) processor, and the AI processor is used to process computational operations related to machine learning.

[0199] The memory 1502 may include one or more computer-readable storage media, which may be non-transitory. The memory 1502 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1502 is used to store at least one instruction for being executed by the processor 1501 to implement the training method of the classification model and the data classification method provided in the method embodiments of the present application.

[0200] In some embodiments, the computer device 1500 may also optionally include other components. Those skilled in the art can understand that Figure 15 the structure shown does not constitute a limitation on the computer device 1500, and it may include more or fewer components than shown in the figure, or combine certain components, or adopt a different component arrangement.

[0201] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program. The program may be stored in a computer-readable storage medium, which may be the computer-readable storage medium included in the memory in the above embodiments; or it may exist separately and not be assembled into the terminal. The computer-readable storage medium stores at least one instruction, at least one segment of program, a code set or an instruction set, and the at least one instruction, the at least one segment of program, the code set or the instruction set is loaded and executed by the processor to implement the training method of the classification model and the data classification method in any one of the above embodiments.

[0202] Optionally, the computer-readable storage medium may include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), solid state drives (SSD, Solid State Drives) or optical discs, etc. Among them, the random access memory may include resistive random access memory (ReRAM, Resistance RandomAccess Memory) and dynamic random access memory (DRAM, Dynamic Random Access Memory). The serial numbers of the embodiments of the present application above are only for description and do not represent the advantages and disadvantages of the embodiments.

[0203] Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiments can be completed by hardware or by a program instructing the relevant hardware. The program can be stored in a computer-readable storage medium, and the above-mentioned storage medium can be a read-only memory, a disk, an optical disc, etc.

[0204] The above are only alternative embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A data processing method, characterized in that, The method includes: Receiving an object test instruction for indicating to perform a simulation usage test on a first object in a virtual scenario, where the object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate the first ability information for simulating an account to use the first object during the test, and the object parameter information is the object parameter to be tested for the first object; Obtaining a first test model corresponding to the first ability information based on the object test instruction. The first test model is a machine learning model trained by first sample performance data, and the first sample performance data is data obtained by an account using a second object to exhibit the first ability information in a virtual scenario during a historical time period; Performing a simulation usage test on the object parameter information through the first test model to obtain first object test data.

2. The method according to claim 1, characterized in that The obtaining a first test model corresponding to the first ability information based on the object test instruction includes: Obtaining a plurality of candidate test models, where the plurality of candidate test models respectively correspond to different ability information; Determining, based on the object test instruction, a candidate test model that matches the first ability information from the plurality of candidate test models as the first test model.

3. The method according to claim 2, wherein The plurality of candidate test models respectively correspond to different candidate ability labels; The selecting, based on the object test instruction, a candidate test model that matches the first ability information from the plurality of candidate test models as the first test model includes: Obtaining a first ability label corresponding to the first ability information based on the object test instruction; Regarding the candidate test model corresponding to the first ability label among the plurality of candidate ability labels as the first test model.

4. The method according to claim 2, characterized in that The obtaining a plurality of candidate test models includes: Obtaining reference object data and object parameter information corresponding to the second object, where the reference object data refers to reference object data that meets deviation requirements during the process of using the second object; Obtaining a plurality of test deviation ranges, where the plurality of test deviation ranges respectively correspond to different ability information; Obtaining reference deviation data respectively corresponding to the second object within the plurality of test deviation ranges based on the reference object data. The reference deviation data refers to data whose deviation rate relative to the reference object data is within the test deviation range; Obtaining a sample test model, where the sample test model is a machine learning model to be trained for associating object parameter information and ability information; Training the sample test model respectively based on the object parameter information of the second object and the reference deviation data respectively corresponding to different ability information to obtain candidate test models respectively corresponding to different ability information.

5. The method according to claim 4, characterized in that The obtaining reference deviation data respectively corresponding to the second object within the plurality of test deviation ranges based on the reference object data includes: Taking the reference object data as a benchmark, performing data traversal within the i-th test deviation range to obtain reference deviation data corresponding to the i-th test deviation range, where i is a positive integer.

6. The method according to claim 4, characterized in that, Training the sample test model respectively based on the object parameter information of the second object and the reference deviation data corresponding to different ability information to obtain candidate test models corresponding to different ability information respectively, includes: Inputting the object parameter information of the second object into the sample test model, and outputting the first sample performance data; Training the sample test model based on the reference deviation data corresponding to different ability information respectively and the first sample performance data to obtain candidate test models corresponding to different ability information respectively, wherein, after training the sample test model with the reference deviation data corresponding to the k-th ability information and the first sample performance data, the candidate test model corresponding to the k-th ability information is obtained, and k is a positive integer.

7. The method according to claim 6, wherein The method further includes: Obtaining the reference performance data corresponding to different ability information of the third object and the object parameter information corresponding to the third object, where the reference performance data is the data obtained by accounts with different ability information performing in the virtual scenario using the second object during the historical time period; Inputting the object parameter information corresponding to the third object into the candidate test models corresponding to different ability information respectively, and outputting the predicted performance data corresponding to different ability information respectively; Based on the matching degree between the predicted performance data corresponding to different ability information respectively and the reference performance data, screening the candidate test models corresponding to the multiple different ability information respectively to obtain a model screening result, and the model screening result is used to determine the candidate test model whose predicted performance data matches the reference performance data.

8. The method according to any one of claims 1 to 7, characterized in that, Before obtaining the first test model corresponding to the first ability information based on the object test instruction, it further includes: Starting the environment operation data corresponding to the virtual environment based on the object test instruction, and shielding the environment rendering data corresponding to the virtual environment, where the environment operation data is used to run the virtual environment; The process of obtaining the first object test data by performing a simulated usage test on the object parameter information through the first test model includes: Running the first test model through the environment operation data; Inputting the object parameter information into the first test model, and outputting the first object test data.

9. The method according to claim 8, wherein The process of inputting the parameter information into the first test model and outputting the first object test data includes: Invoking an object operation interface, where the object operation interface includes object operation instructions; Inputting the object parameter information and the object operation instructions into the first test model, and outputting the first object test data.

10. The method according to any one of claims 1 to 7, characterized in that, The object test instruction further includes second ability simulation information, and the second ability simulation information is used to indicate simulating the second ability information of using the first object by the simulated account during the test; After obtaining the first object test data by performing a simulated usage test on the object parameter information through the first test model, it further includes: Obtain a second test model corresponding to the second ability information based on the object test instruction. The second test model is a machine learning model trained through second sample performance data, and the second sample performance data is data obtained by the account demonstrating the second ability information in the virtual scene using a third object during a historical time period; Conduct a simulated usage test on the object parameter information through the second test model to obtain second object test data; Integrate the first object test data and the second object test data to obtain a data integration result; Based on the data integration result, conduct a test evaluation on the first object to obtain an object test result, where the object test result refers to the performance of the first object corresponding to different ability information.

11. A data processing system, characterized in that, The data processing system includes a terminal and a server, and there is a communication connection between the terminal and the server; The server is used to send an object test instruction to the terminal. The object test instruction is used to indicate a simulated usage test of a first object in a virtual scene. The object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate that when testing, the simulated account uses the first ability information of the first object, and the object parameter information is the object parameter to be tested for the first object; The terminal is used to receive the object test instruction; Based on the object test instruction, obtain a first test model corresponding to the first ability information. The first test model is a machine learning model trained through first sample performance data, and the first sample performance data is data obtained by an account with first ability information demonstrating the first ability information in the virtual scene using a second object during a historical time period; conduct a simulated usage test on the object parameter information through the first test model to obtain first object test data; send the first object test data to the server; The server is further used to receive the first object test data and object test data corresponding to other ability information.

12. A data processing device, characterized in that, The device includes: A receiving module, used to receive an object test instruction. The object test instruction is used to indicate a simulated usage test of a first object in a virtual scene. The object test instruction includes first ability simulation information and object parameter information. The first ability simulation information is used to indicate that when testing, the simulated account uses the first ability information of the first object, and the object parameter information is the object parameter to be tested for the first object; An obtaining module, used to obtain a first test model corresponding to the first ability information based on the object test instruction. The first test model is a machine learning model trained through first sample performance data, and the first sample performance data is data obtained by an account with first ability information demonstrating the first ability information in the virtual scene using a second object during a historical time period; A testing module, used to conduct a simulated usage test on the object parameter information through the first test model to obtain first object test data.

13. A computer device, characterized in that, The computer device includes a processor and a memory, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the data processing method according to any one of claims 1 to 10.

14. A computer-readable storage medium, characterized in that, The storage medium stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the data processing method according to any one of claims 1 to 10.

15. A computer program product, characterized in that, The method comprises computer instructions, which, when executed by a processor, implement the data processing method according to any one of claims 1 to 10.

Images