A game object processing method and device, computer equipment and a storage medium

Abstract

Embodiments of the present application provide a game object processing method and device, computer equipment and a storage medium, which can be applied to various scenes such as cloud technology, artificial intelligence, intelligent transportation and auxiliary driving. Embodiments of the present application can obtain a set of battle state information of a target object and a game object in a game; perform feature extraction on the battle state information included in the set of battle state information to obtain battle state feature information corresponding to the battle state information; perform operation level prediction on the target object according to the battle state feature information to obtain operation level information corresponding to each game; and perform fusion processing on the operation level information to obtain target operation level information of the target object. Embodiments of the present application can accurately determine the operation level of the target object.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a method, apparatus, computer device, and storage medium for processing game objects, wherein the storage medium is a computer-readable storage medium. Background Technology

[0002] Currently, the most common method for valuing players' skill level in multiplayer online battle arena (MOBA) games is to use a score to represent their skill level.

[0003] Specifically, the player skill assessment method directly determines a player's skill level based on the score they earn in a game. This method is rather simplistic and cannot accurately reflect a player's actual skill level. Summary of the Invention

[0004] This application provides a method, apparatus, computer device, and storage medium for processing game objects, which can accurately determine the operational level of a target object.

[0005] A method for handling game objects, comprising:

[0006] Obtain a set of battle status information for target objects and opponents in the game. The set of battle status information includes battle status information for at least one game.

[0007] Feature extraction is performed on the battle status information included in the battle status information set to obtain the battle status feature information corresponding to the battle status information;

[0008] Based on the characteristics of the battle status, the skill level of the target object is predicted to obtain the skill level information corresponding to each game.

[0009] The operational level information of the target object is fused to obtain the target operational level information for the target object.

[0010] Accordingly, embodiments of this application provide a game object processing apparatus, including:

[0011] The acquisition unit can be used to acquire a set of battle status information of target objects and game objects in the game. The set of battle status information includes battle status information corresponding to at least one game.

[0012] The extraction unit can be used to extract features from the battle status information included in the battle status information set to obtain the battle status feature information corresponding to the battle status information.

[0013] The prediction unit can be used to predict the skill level of a target object based on the characteristics of the battle state, and obtain the skill level information corresponding to each game.

[0014] The fusion unit can be used to fuse the operation level information of the target object to obtain the target operation level information for the target object.

[0015] In some embodiments, the prediction unit may be used to fuse the battle state feature information corresponding to each game to obtain the fused state feature information corresponding to each game; based on the fused state feature information corresponding to each game, the trained operation level prediction model is used to predict the operation level of the target object to obtain the operation level information corresponding to each game.

[0016] In some embodiments, the battle status information includes the battle status information corresponding to each time point in the time series of each game; the extraction unit can be specifically used to extract features from the battle status information corresponding to each time point in the time series of each game to obtain the battle status feature information corresponding to each time point in the time series of each game.

[0017] Correspondingly, the prediction unit can be used to fuse the battle state feature information corresponding to each time point in the time series based on the time series of each game, so as to obtain the fused state feature information corresponding to each game.

[0018] In some embodiments, the game includes multiple rounds; the fusion unit can specifically be used to, for the multiple rounds of the game, take any one of the rounds as the current round, and perform the following operations: for the current round of the game, if the current round of the game is the first round of the game, obtain the initial comprehensive operation level information of the target object, and calculate the comprehensive operation level information of the current round of the game based on the initial comprehensive operation level information and the operation level information of the target object in the current round of the game; if the current round of the game is not the first round of the game, obtain the comprehensive operation level information of the target object in the previous round of the game, and calculate the comprehensive operation level information of the current round of the game based on the comprehensive operation level information of the previous round of the game and the operation level information of the target object in the current round of the game; until each round of the multiple rounds of the game has been calculated, the comprehensive operation level information of each round of the game is obtained; and determine the target operation level information for the target object based on the comprehensive operation level information of the target object.

[0019] In some embodiments, the battle state feature information is information obtained by feature extraction of battle state information using a trained operation level prediction model; the game object processing device further includes a training unit, which can be used to acquire a battle data sample set, the battle data sample set including battle data samples labeled with operation level tags in at least one game, the battle data samples being data samples labeled with operation level tags for the object to be predicted; based on the battle data samples, the operation level prediction model to be trained is used to predict the operation level of the object to be predicted, obtaining the predicted operation level information of the object to be predicted for each game; based on the operation level tags and predicted operation level information corresponding to the battle data samples, the model parameters of the operation level prediction model to be trained are converged to obtain the trained operation level prediction model.

[0020] In some embodiments, the training unit can be specifically used to acquire a first game faction and a second game faction. The first game faction includes a target object to be predicted and at least one first participating object, and the second game faction includes at least one second participating object. The candidate operation level information of the target object, the first participating object, and the second participating object are all randomly assigned operation level information. The target object to be predicted, the first participating object, and the second participating object are subjected to a battle to obtain battle data samples. Based on the candidate operation level information of the target object, the operation level label corresponding to the battle data sample is determined. A battle data sample set is generated based on the battle data samples and the corresponding operation level labels.

[0021] In some embodiments, the training unit may be specifically used to extract features from the battle data samples using the operation level prediction model to be trained, and obtain sample feature information; based on the sample feature information, to predict the operation level of the object to be predicted, and obtain the predicted operation level information of the object to be predicted for each game.

[0022] In some embodiments, the training unit may be used to fuse sample feature information for each game to obtain fused sample feature information; and based on the fused sample feature information, to predict the operation level of the object to be predicted to obtain the predicted operation level information for each game of the object to be predicted.

[0023] In some embodiments, the training unit can be specifically used to process the object to be predicted, the first participating object, and the second participating object into a battle to obtain candidate battle data; and to extract samples from the candidate battle data according to the target time series to obtain the battle data samples corresponding to the object to be predicted in each game.

[0024] In some embodiments, the game object processing device further includes a matching unit, which can be specifically used to determine the target level information of the target object based on the target operation level information of the target object; obtain at least one candidate game object based on the target level information of the target object; and extract the target game object that matches the target object from the candidate game objects.

[0025] Furthermore, this application also provides a computer device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute any of the game object processing methods provided in this application.

[0026] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a computer program adapted for loading by a processor to execute any of the game object processing methods provided in embodiments of this application.

[0027] Furthermore, this application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the game object processing methods provided in this application.

[0028] This application embodiment can obtain a set of battle status information of a target object and a player in a game. The set of battle status information includes battle status information corresponding to at least one game. Features are extracted from the battle status information in the set to obtain corresponding battle status feature information. Based on the battle status feature information, the skill level of the target object is predicted to obtain skill level information corresponding to each game. The skill level information of the target object is then fused to obtain target skill level information for the target object. Because this application embodiment can determine the skill level information for each game based on the battle status information corresponding to each game between the target object and the player, the target skill level information of the target object can be accurately calculated, thereby accurately determining the skill level of the target object. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a schematic diagram of a scenario illustrating the game object processing method provided in an embodiment of this application;

[0031] Figure 2This is a flowchart illustrating the game object processing method provided in an embodiment of this application;

[0032] Figure 3 This is a schematic diagram of the training process for the operation level prediction model to be trained, provided in an embodiment of this application.

[0033] Figure 4 These are two flowchart illustrations of the game object processing method provided in the embodiments of this application;

[0034] Figure 5 This is a structural diagram of the operational level prediction model to be trained provided in the embodiments of this application;

[0035] Figure 6 This is a schematic diagram illustrating the application of the post-training operational level prediction model provided in the embodiments of this application;

[0036] Figure 7 This is a schematic diagram illustrating the convergence of target operation level information for a target object, provided in an embodiment of this application.

[0037] Figure 8 This is a schematic diagram of the structure of the game object processing device provided in the embodiments of this application;

[0038] Figure 9 This is a schematic diagram of the structure of the computer device provided in the embodiments of this application. Detailed Implementation

[0039] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0040] This application provides a game object processing method, apparatus, computer device, and computer-readable storage medium. The game object processing apparatus can be integrated into a computer device, which may be a server or a terminal, etc.

[0041] The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery network (CDN) acceleration services, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein. Terminals include, but are not limited to, mobile phones, computers, smart voice interaction devices, smart home appliances, vehicle terminals, and aircraft. The embodiments of this application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, and assisted driving.

[0042] This application's embodiments may relate to Artificial Intelligence (AI). AI is the theory, methods, technology, and application systems that utilize digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce new intelligent machines capable of reacting in a manner similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities.

[0043] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0044] For example, see Figure 1 Taking the integration of a game object processing device into a computer device as an example, the computer device can acquire a set of battle status information of target objects and game objects in the game; extract features from the battle status information included in the battle status information set to obtain the battle status feature information corresponding to the battle status information; predict the operation level of the target object based on the battle status feature information to obtain the operation level information corresponding to each game; and fuse the operation level information of the target object to obtain the target operation level information for the target object.

[0045] The battle status information set includes battle status information corresponding to at least one game, and each game can correspond to several battle status information sets.

[0046] The game can be any game. For example, a game can be a human-computer game, in which a virtual object controlled by a player in the game is played against a virtual object controlled by artificial intelligence in the game; or a game can be a human-human game, in which a virtual object controlled by a player in the game is played against another human.

[0047] The target object can refer to the virtual object controlled by the target player in the game; the target object may include at least one virtual object.

[0048] The game object can refer to a virtual object controlled by artificial intelligence in the game, or a virtual object controlled by a non-target player in the game; the game object can include at least one virtual object.

[0049] Among them, the skill level information can be a representation of the game skill level of the target player in a game, which is predicted to be the skill level of the target player in dealing with the target object; the target skill level information can be a representation of the actual game skill level of the target player in dealing with the target object in at least one game.

[0050] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the preferred order of the embodiments.

[0051] This embodiment will be described from the perspective of a game object processing device, which can be integrated into a computer device, such as a server or a terminal. The terminal can include tablet computers, laptop computers, personal computers (PCs), wearable devices, virtual reality devices, or other smart devices that can acquire data.

[0052] like Figure 2 As shown, the specific flow of this game object processing method is as follows: steps S101-S104:

[0053] S101. Obtain a set of battle status information for the target object and the opponent in the game.

[0054] The battle status information set includes battle status information for at least one game. Each game can correspond to several battle status information sets.

[0055] In the embodiments of this application, the target object may refer to a virtual object controlled by a target player in the game; the target object may include at least one virtual object. The game object may refer to a virtual object controlled by an artificial intelligence neural network model in the game, or it may refer to a virtual object controlled by a non-target player in the game; the game object may include at least one virtual object.

[0056] For example, in this embodiment of the application, the playing factions may include at least two factions, such as a first playing faction and a second playing faction. The first playing faction includes at least one target object, and the second playing faction includes at least one playing object. The first playing faction and the second playing faction engage in combat to obtain game state information corresponding to the game session in which the first playing faction and the second playing faction engage in combat.

[0057] In this embodiment, since not all candidate battle state information generated during the game between the target object and the opponent is required by this embodiment, the candidate battle state information can refer to any information generated by the target object and the opponent during the game. Therefore, for each game, this embodiment can extract candidate battle state information according to each time point in the time sequence. Thus, the computer device can extract the battle state information corresponding to each time point in the time sequence of each game. In this embodiment, the time sequence of each game includes multiple time points.

[0058] S102. Extract features from the battle status information included in the battle status information set to obtain the battle status feature information corresponding to the battle status information.

[0059] Among them, the battle status feature information can be information extracted by using a feature extraction model, and the feature extraction model can be a neural network model in existing related technologies.

[0060] The combat status feature information can also be information extracted using a post-trained operation level prediction model. The post-trained operation level prediction model is the model obtained after training the operation level prediction model to be trained.

[0061] The battle status feature information may include battle status feature information corresponding to at least one status dimension, and the status dimension may include lineup dimension, position dimension and battle dimension.

[0062] The lineup dimension is used to represent the lineup of an object within a match's faction. Objects include the target object and the match object. Specifically, the lineup dimension's match state characteristic information includes the lineup state characteristic information corresponding to the target object and the lineup state characteristic information corresponding to the match object. The lineup state characteristic information corresponding to the target object can be the lineup information within the target object's match faction, and the lineup state characteristic information corresponding to the match object can be the lineup information within the match object's match faction.

[0063] The position dimension is used to represent the position of an object within the game's factions. Objects include target objects and other players. Specifically, the position dimension's battle state feature information includes the position dimension's battle state feature information corresponding to the target object and the position dimension's battle state feature information corresponding to other players. The position dimension's battle state feature information corresponding to the target object can be the target object's position within the game's factions; the position dimension's battle state feature information corresponding to other players can be the other player's position within their respective factions.

[0064] The combat dimension can refer to a dimension other than the position and lineup dimensions. The combat dimension is used to characterize the combat status of objects in a game. Specifically, the combat status feature information of the combat dimension can be the combat information of the target object and the other objects in the game.

[0065] In this embodiment of the application, the battle status information includes the battle status information corresponding to each time point in the time sequence of each game. Based on this, this embodiment of the application can extract features from the battle status information corresponding to each time point in the time sequence of each game to obtain the battle status feature information corresponding to each time point in the time sequence of each game.

[0066] Since the post-trained operational level prediction model is the model obtained after training the operational level prediction model to be trained, the embodiments of this application can train the operational level prediction model to be trained, such as... Figure 3 As shown, the training process is as follows: steps A1-A3:

[0067] A1. Obtain a sample set of battle data.

[0068] The battle data sample set includes battle data samples labeled with skill level tags from at least one game. The battle data samples are data samples labeled with skill level tags for the object to be predicted.

[0069] In this embodiment of the application, the computer device may acquire the battle data sample set in the following ways:

[0070] For example, a computer device can acquire a first game faction and a second game faction, wherein the first game faction includes an object to be predicted and at least one first participating object, and the second game faction includes at least one second participating object; the object to be predicted, the first participating object, and the second participating object are subjected to a battle to obtain a battle data sample; based on the candidate operation level information of the object to be predicted, the operation level label corresponding to the battle data sample is determined; and based on the battle data sample and the operation level label corresponding to the battle data sample, a battle data sample set is generated.

[0071] In this embodiment, to simulate the situation where a player's matched opponents and game difficulty are inaccurate during the beginner stage, the candidate operation level information of the object to be predicted, the first participating object, and the second participating object are all randomly assigned operation level information. In this embodiment, the candidate operation level information can be represented by the game's operation difficulty level. In this embodiment, the object to be predicted, the first participating object, and the second participating object can all be controlled by a neural network model. Based on this, the computer device can randomly assign candidate operation level information to the object to be predicted, the first participating object, and the second participating object.

[0072] In this embodiment of the application, the method of processing the object to be predicted, the first participating object, and the second participating object into a battle to obtain battle data samples can be as follows: the computer device can process the object to be predicted, the first participating object, and the second participating object into a battle to obtain candidate battle data; according to the target time series, the candidate battle data is sampled to obtain the battle data samples corresponding to the object to be predicted in each game.

[0073] In this embodiment, the target time series can refer to the target time series of a game. The target time series can include multiple time points. In this embodiment, the target time series can include twelve time points in a game as an example. In this embodiment, for each game, the computer device can extract samples of candidate battle data according to the twelve time points in the target time series to obtain the battle data samples corresponding to the object to be predicted in each game.

[0074] A2. Based on the battle data samples, the operation level prediction model to be trained is used to predict the operation level of the target, and the predicted operation level information of the target for each game is obtained.

[0075] In this embodiment of the application, the operational skill prediction model to be trained is used to predict the operational skill level of the target object based on the battle data samples. The method to obtain the predicted operational skill level information of the target object for each game can be as follows:

[0076] For example, computer equipment can use a training operation level prediction model to extract features from battle data samples to obtain sample feature information; based on the sample feature information, the operation level of the object to be predicted is predicted to obtain the predicted operation level information for each game of the object to be predicted.

[0077] In this embodiment of the application, the method for predicting the operational level of the object to be predicted based on the sample feature information and obtaining the predicted operational level information for each game of the object to be predicted can be as follows: for each game, the sample feature information is fused to obtain fused sample feature information; based on the fused sample feature information, the operational level of the object to be predicted is predicted to obtain the predicted operational level information for each game of the object to be predicted.

[0078] In this embodiment of the application, for each game, the sample feature information is fused to obtain the fused sample feature information. The method is as follows: the sample feature information corresponding to each time point in the target time series is concatenated in ascending order of the target time series to construct candidate time series feature information, which is the fused sample feature information.

[0079] A3. Based on the operation level labels and predicted operation level information corresponding to the battle data samples, the model parameters of the operation level prediction model to be trained are converged to obtain the trained operation level prediction model.

[0080] Among them, the operation level label can be understood as the actual operation level information of the object to be predicted.

[0081] It is understandable that the skill level labels and predicted skill level information corresponding to the battle data samples are both for the same game of the same target.

[0082] Based on the above, the embodiments of this application can calculate the loss value between the operation level label and the predicted operation level information using a loss function; and converge the model parameters of the operation level prediction model to be trained according to the loss value to obtain the trained operation level prediction model.

[0083] The loss function can be the cross-entropy loss function, the exponential loss function, or the mean squared error loss function, etc.

[0084] The loss function can be either the cross-entropy loss function or the mean squared error loss function.

[0085] S103. Based on the characteristics of the battle status, predict the skill level of the target object to obtain the skill level information corresponding to each game.

[0086] This application embodiment predicts the skill level of a target object based on battle status characteristic information, and there are multiple ways to obtain the skill level information corresponding to each game:

[0087] For example, computer equipment uses existing game skill prediction models based on the characteristics of the battle status to predict the skill level of the target object, thereby obtaining the skill level information corresponding to each game.

[0088] For example, computer equipment can fuse the battle state feature information corresponding to each game to obtain the fused state feature information corresponding to each game; based on the fused state feature information corresponding to each game, a trained operation level prediction model is used to predict the operation level of the target object to obtain the operation level information corresponding to each game.

[0089] In this embodiment, the battle state feature information may include the battle state feature information corresponding to each time point in the time sequence of each game. Based on this, the method of fusing the battle state feature information corresponding to each game to obtain the fused state feature information corresponding to each game can be as follows: according to the time sequence of each game, the battle state feature information corresponding to each time point in the time sequence is fused to obtain the fused state feature information corresponding to each game.

[0090] In this embodiment of the application, the method for fusing the battle state feature information corresponding to each time point in the time series can be to concatenate the battle state feature information corresponding to each time point in the time series in ascending order of the time series to construct temporal feature information, which is the fused state feature information.

[0091] S104. Perform fusion processing on the operation level information of the target object to obtain the target operation level information for the target object.

[0092] In this embodiment of the application, the computer device can determine the target operation level information of the target object based on the operation level information corresponding to each game.

[0093] The embodiments of this application perform fusion processing on the operational level information of the target object to obtain the target operational level information for the target object in the following manner:

[0094] In this embodiment, the game includes multiple rounds. For each round, any one of the rounds is taken as the current round, and the following operations are performed: For example, for the current round, if it is the first round, the initial comprehensive operation level information of the target object is obtained, and the comprehensive operation level information of the current round is calculated based on the initial comprehensive operation level information and the operation level information of the target object in the current round; if the current round is not the first round, the comprehensive operation level information of the target object in the previous round is obtained, and the comprehensive operation level information of the current round is calculated based on the comprehensive operation level information of the previous round and the operation level information of the target object in the current round; until each round in the multiple rounds has been calculated, the comprehensive operation level information of each round is obtained; based on the comprehensive operation level information of the target object, the target operation level information for the target object is determined.

[0095] The initial overall operational level information can be the overall operational level information initialized by the computer equipment.

[0096] The previous game refers to the game before the current game.

[0097] Since the operational level information in this application embodiment can be represented numerically, and the initial comprehensive operational level information can also be represented numerically, based on this, the method by which this application embodiment calculates the comprehensive operational level information of the current game based on the initial comprehensive operational level information and the operational level information of the target object in the current game can be: obtaining a preset function; and mapping the initial comprehensive operational level information and the operational level information of the target object in the current game according to the preset function to obtain the comprehensive operational level information of the current game.

[0098] Similarly, in this embodiment, the overall skill level information of the previous game and the skill level information of the target object in the current game can be mapped according to the preset function to obtain the overall skill level information of the current game.

[0099] Among them, the overall operational level information can be represented in numerical form, and the target operational level information can be represented in numerical form.

[0100] This application embodiment uses multiple methods to determine the target operational level information for the target object based on the comprehensive operational level information of the target object:

[0101] For example, a computer device calculates the average value of the overall skill level information for each game and uses this average value as the target skill level information for the target object.

[0102] For example, a game may consist of multiple rounds, and the overall skill level information may include the overall skill level information of the previous round and the overall skill level information of the current round.

[0103] For the current game, if it is the first game, calculate the difference between the current game's overall skill level information and the initial overall skill level information; if the current game is not the first game, calculate the difference between the previous game's overall skill level information and the current game's overall skill level information. This difference can be represented numerically. Treating each game as the current game, repeat the steps of calculating the difference between the current game's overall skill level information and the initial overall skill level information if it is the first game, and calculating the difference between the previous game's overall skill level information and the current game's overall skill level information if it is not the first game, until the difference is less than or equal to a preset threshold, thus obtaining the target skill level information for the target object. The preset threshold can be 0.

[0104] After step S104 in this embodiment, a corresponding target game object can be matched for the target object. For example, the computer device can determine the target skill level information of the target object based on the target skill level information of the target object; obtain at least one candidate game object based on the target skill level information of the target object; and extract the target game object that matches the target object from the candidate game objects.

[0105] In this embodiment, a mapping relationship set can be obtained, which includes the mapping relationship between preset operation level information and preset level information; based on the mapping relationship set, the target level information corresponding to the target operation level information of the target object can be determined.

[0106] In this embodiment of the application, the candidate game objects and the target skill level information have a corresponding relationship. Based on this, this embodiment of the application can obtain at least one candidate game object.

[0107] There are several ways in which a target game object matches another target object. For example, when the skill level information of the target game object is the same as the target skill level information of the target object, this can be called a match between the target game object and the target object. When the skill level information of the target game object is the same as the target skill level information of the target object, and the object type of the target game object and the object type of the target object meet the preset rules, this can also be called a match between the target game object and the target object.

[0108] This application embodiment can obtain a set of battle status information of a target object and a player in a game. The set of battle status information includes battle status information corresponding to at least one game. Features are extracted from the battle status information in the set to obtain corresponding battle status feature information. Based on the battle status feature information, the skill level of the target object is predicted to obtain skill level information corresponding to each game. The skill level information of the target object is then fused to obtain target skill level information for the target object. Because this application embodiment can determine the skill level information for each game based on the battle status information corresponding to each game between the target object and the player, the target skill level information of the target object can be accurately calculated, thereby accurately determining the skill level of the target object.

[0109] Based on the method described in the above embodiments, the following examples will provide further detailed explanations.

[0110] In this embodiment, the game object processing device is specifically integrated into a computer device, which serves as a server.

[0111] like Figure 4 As shown, a method for handling game objects includes the following steps S201-S207:

[0112] S201. Computer equipment acquires a set of battle data samples.

[0113] In this embodiment of the application, the computer device may acquire the battle data sample set in the following ways:

[0114] For example, a computer device can acquire a first game faction and a second game faction, wherein the first game faction includes an object to be predicted and at least one first participating object, and the second game faction includes at least one second participating object; the object to be predicted, the first participating object, and the second participating object are subjected to a battle to obtain a battle data sample; based on the candidate operation level information of the object to be predicted, the operation level label corresponding to the battle data sample is determined; and based on the battle data sample and the operation level label corresponding to the battle data sample, a battle data sample set is generated.

[0115] Specifically, in the game, the computer can use colors to distinguish between the first and second game factions. For example, the first game faction is red, and the second game faction is blue. Each faction has five virtual objects. The virtual objects in the first game faction include the object to be predicted and four first-participating objects, while the virtual objects in the second game faction include five second-participating objects. Each virtual object is controlled by a corresponding neural network model.

[0116] Then, for the first game faction, the computer device acquires a first preset operation level information set, which includes several first preset operation level information pieces. First preset operation level information pieces are randomly extracted from this set as candidate operation level information for the objects to be predicted. The computer device then acquires a second preset operation level information set, which includes several second preset operation level information pieces. Second preset operation level information pieces are randomly extracted from this set as candidate operation level information for each of the first participating objects.

[0117] For the second game faction, the computer device acquires a third preset operation level information set. This third preset operation level information set includes several third preset operation level information sets. Third preset operation level information is randomly extracted from this set as candidate operation level information for each second participating entity. The first, second, and third preset operation level information sets can be set according to specific requirements.

[0118] Based on the above, under the control of the neural network model, the first game faction and the second game faction engage in battle to obtain battle data samples. That is, the method by which the object to be predicted, the first participating object, and the second participating object are processed to obtain battle data samples in this embodiment of the application can be as follows: the computer device can process the object to be predicted, the first participating object, and the second participating object into a battle to obtain candidate battle data; according to the target time sequence of each game, samples are extracted from the candidate battle data to obtain the battle data samples corresponding to the object to be predicted in each game.

[0119] In this embodiment of the application, since candidate operation level information can be represented by game operation difficulty level, based on this, different preset game operation difficulty level ranges can be set in advance as distinguishing conditions to distinguish between the single-player battle situation of the simulated object to be predicted and the team battle situation of the simulated object to be predicted.

[0120] For example, in the case of a single-player battle for the object to be predicted, the candidate skill level information of the object to be predicted is within the range of a first preset game difficulty level, while the candidate skill level information of the first participating object and the second participating object are not within the range of the first preset game difficulty level. In this case, the computer device can extract samples only from the candidate battle data corresponding to the object to be predicted, thus simulating the single-player battle situation of the object to be predicted.

[0121] Regarding the team battle scenario of the target being predicted, the candidate skill levels of the target and the first participating player can be within the range of a second preset game difficulty level. In this case, the computer device can extract samples of candidate battle data corresponding to all virtual objects in the first game faction to which the target belongs, thus simulating the team battle scenario of the target.

[0122] In this embodiment, the first and second game factions engaged in at least 420 game battles. For team battles involving the target object, this embodiment can generate at least 5 million battle data samples. For single-player battles involving the target object, the target object can also participate in multiple battles, generating a large number of battle data samples.

[0123] In this embodiment, the target time series may include twelve time points from each game. In this embodiment, if the time points in the game duration between the first and second game factions do not meet the target time series, no game data samples need to be acquired.

[0124] S202. The computer equipment uses the operation level prediction model to be trained to predict the operation level of the target object based on the battle data sample, and obtains the predicted operation level information of the target object for each game.

[0125] For example, computer equipment can use the feature extraction layer of the operation level prediction model to extract features from the battle data samples to obtain sample feature information; based on the sample feature information, the operation level of the object to be predicted is predicted to obtain the predicted operation level information for each game of the object to be predicted.

[0126] The computer equipment can fuse sample feature information for each game to obtain fused sample feature information; based on the fused sample feature information, it can predict the operation level of the object to be predicted to obtain the predicted operation level information for each game of the object to be predicted.

[0127] The sample feature information collected in this application embodiment may include sample feature information of multiple candidate dimensions, including: virtual object position, MVP (most valuable player) score, virtual object candidate operation level information, number of factions, probability of the predicted object participating in a faction, damage of the predicted object to the second participating object, virtual object virtual kills, virtual object virtual kill percentage, virtual damage taken by the virtual object, virtual building damage value, virtual assists, virtual last hits, virtual deaths, virtual death percentage, virtual damage taken / death ratio, virtual damage kill ratio, virtual object level, virtual gold, game duration, virtual kills of the faction, virtual deaths of virtual objects of the faction, faction of the predicted object, color of the faction of the predicted object, game result, virtual kills of the game faction, virtual deaths of the game faction, average virtual kills per minute of the predicted object's faction, average virtual deaths per minute of the predicted object's faction, average virtual kills per minute, average virtual assists per minute, average virtual deaths per minute of the predicted object's faction, and average virtual assists per minute. The following metrics are listed: virtual gold earned, average virtual gold difference per minute between factions, average virtual kill difference per minute between factions, average virtual last hits per minute, average virtual damage dealt to the target per minute, average virtual damage received by the target per minute, average number of team fights participated in by the target per minute, average MVP score per minute, average (MVP score - faction average MVP score) per minute, average (MVP score - both factions average MVP score) per minute, average (MVP score - opposing faction average MVP score) per minute, average MVP score of the target's faction, average MVP score of the match, MVP score - faction average MVP score, MVP score - both factions average MVP score, MVP score - match average MVP score, average kills per minute for the match, average virtual gold per minute for the match, the gap between the candidate skill level information of the target's faction and the candidate skill level information of the match, and the match faction.

[0128] The sample feature information in this application embodiment may include discrete feature information and continuous feature information.

[0129] Based on the above, the discrete feature information can include sample feature information of the virtual object identity candidate dimension, sample feature information of the virtual object position candidate dimension, sample feature information of the game result candidate dimension, sample feature information of the virtual object's candidate operation level information candidate dimension, and sample feature information of the game faction candidate dimension. Except for the sample feature information corresponding to the candidate dimensions in the discrete feature information, the sample feature information corresponding to other candidate dimensions can be regarded as continuous feature information.

[0130] For continuous feature information, computer devices perform discretization processing. For example, they can use equal frequency or equal intervals to divide the continuous feature information into bins to discretize it. Specifically, for each continuous feature information, the computer device divides it into several different bins, for example, any number of different bins within the range of 15-20. In this way, the continuous feature information is transformed into discrete feature information.

[0131] Regarding the aforementioned discrete feature information, since this information can be represented numerically, the computer device filters out discrete feature information exceeding a target threshold. This filters out anomalous discrete feature information, resulting in filtered feature information. The filtered feature information is then encoded, using both one-hot and multi-hot encoding methods.

[0132] In this embodiment, the preset discrete feature information can be used to replace the discrete feature information that is greater than the target threshold to obtain the replaced discrete feature information. Based on this, the replaced discrete feature information and the normal discrete feature information can be used as the filtered feature information.

[0133] One-hot encoding can refer to a binary vector encoding form of discrete feature information, which maps discrete values ​​to integer values. Each integer value is represented as a binary vector, and except for the integer index, it is a zero value, which is marked as 1.

[0134] Multi-hot encoding can refer to a binary vector encoding form of discrete feature information, which maps discrete values ​​to integer values. Each integer value is represented as a binary vector, with all integers except for a few integers whose indices are 1.

[0135] Specifically, in this application embodiment, the filtered feature information can be one-hot encoded to obtain one-hot encoded feature information; the one-hot encoded feature information can be multi-hot encoded to obtain multi-hot encoded feature information; and the multi-hot encoded feature information can be used as sample feature information, thus eliminating positional bias.

[0136] For example, the candidate virtual objects include a first candidate virtual object and a second candidate virtual object. The one-hot encoded feature information of the position of the first candidate virtual object can be represented as 010, and the one-hot encoded feature information of the position of the second candidate virtual object can be represented as 001. Through multi-hot encoding, the multi-hot encoded feature information can be represented as 010001 or 001010, thus eliminating positional deviation.

[0137] In this application embodiment, for each game, the sample feature information is fused to obtain the fused sample feature information in the following way: the computer device splices the sample feature information corresponding to each time point in the target time series in ascending order of the target time series to construct candidate time series feature information, and the candidate time series feature information is the fused sample feature information.

[0138] Furthermore, such as Figure 5 As shown, the operation level prediction model to be trained can be an Omninet neural network model, which includes a feature extraction layer, a Transformer layer, an Omnidirectional Attention layer, and an output layer.

[0139] When training the operational level prediction model, the entire dataset (approximately 5 million records) was divided into a training set (80%) and a test set (20%), and the model training batch size was set to 2000.

[0140] The feature extraction layer can be used to map the battle data samples to a dimensional space to obtain the sample feature information corresponding to the battle data samples. In this embodiment, the battle data samples can be displayed in the form of game battle images. For example, for each game, the battle data samples include the battle data samples corresponding to each time point in the target time series. The time points in the target time series include the first time point, the second time point, ..., the Nth time point. N is a positive integer, and in this embodiment, N can be set to twelve.

[0141] In this embodiment, the battle data samples include battle data samples corresponding to each time point in the target time series within each game. The feature extraction layer of the operation level prediction model to be trained extracts features from the battle data samples corresponding to each time point in the target time series to obtain sample feature information corresponding to each time point in the target time series. The sample feature information may include sample feature information corresponding to at least one state dimension. The state dimension may include lineup dimension, position dimension, and battle dimension. The lineup dimension, position dimension, and battle dimension are explained above and will not be repeated here.

[0142] In this embodiment, the sample feature information corresponding to each state dimension may include sample feature information of at least one candidate dimension. For example, the sample feature information of the lineup dimension may include sample feature information of the number of factions, the candidate dimension of the faction to be predicted, the candidate dimension of the color of the faction to be predicted, and the candidate dimension of the match faction. The sample feature information of the position dimension may include sample feature information of the virtual object position. In this embodiment, the sample feature information of other candidate dimensions may be included in the sample feature information of the battle dimension.

[0143] In this embodiment, the sample feature information corresponding to each time point in the target time series is concatenated in ascending order according to the target time series to construct candidate time series feature information. The candidate time series feature information is the fused sample feature information. It can be understood that the fused sample feature information is the sample feature information of the lineup dimension, the sample feature information of the position dimension, and the sample feature information of the battle dimension, which is the feature information obtained by concatenating the target time series in ascending order.

[0144] Since the fused sample feature information consists of candidate temporal feature information collected multiple times in a game, this application adopts the Transformer, which is adept at processing temporal feature information, as the basic architecture of the operational level information prediction model to be trained. The Transformer layer has at least 6 neural network layers. Compared with traditional sequence deep learning models, such as RNN and LSTM, the Transformer layer uses self-attention instead of self-loop dependency. The structure of the Transformer layer is easy to parallelize and has stronger interpretability.

[0145] In this embodiment, the Transformer layer has L-1 neural network layers as an example, where L is a positive integer and L is greater than 1. In this embodiment, the fused sample feature information X is input into the Transformer layer for processing. The processing of the fused sample feature information by the first neural network layer of the Transformer layer can be expressed as formula (1):

[0146]

[0147] Where X1 represents the feature information of the first neural network layer in the Transformer layer. This can represent the battle data sample corresponding to the Nth time point, which has been processed by the first layer of the neural network in the feature extraction layer and the Transformer layer, and the sample feature information collected by the first layer of the neural network.

[0148] The second neural network layer of the Transformer layer processes X1 to obtain the feature information of the second neural network layer in the Transformer layer, as shown in formula (2):

[0149]

[0150] Where X2 represents the feature information collected by the second neural network layer in the Transformer layer. This can represent the battle data sample corresponding to the Nth time point, which has been processed by the first layer of the feature extraction layer and the second layer of the Transformer layer, and the sample feature information collected by the second layer of the Transformer layer.

[0151] By analogy, the feature information of the (L-1)th layer of the neural network in the Transformer layer can be obtained, as shown in formula (3):

[0152]

[0153] Among them, X L-1 This represents the feature information obtained from the (L-1)th layer of the neural network in the Transformer layer. This can represent the battle data sample corresponding to the Nth time point, which has been processed by the L-1 layer neural network in the feature extraction layer and the Transformer layer, and the sample feature information collected by the L-1 layer neural network.

[0154] The Omnidirectional Attention layer, also known as the all-around attention neural network layer, employs self-attention. It allows each sample's feature information to "interact" with all sample features throughout the network, giving it a receptive field across the entire width and depth of the neural network. This dense form of residual connections facilitates gradient propagation and enhances the feature representation capability of the operational-level prediction model after training, resulting in more accurate predictions and higher precision than traditional Transformer models.

[0155] The Omnidirectional Attention layer receives the output feature information of the Transformer layer. The output feature information Xformer(X) of the Transformer layer can be expressed as formula (4):

[0156] Xformer(X)=X1,X2,...,X L-1 Formula (4)

[0157] The Omnidirectional Attention layer processes the output features of the Transformer layer to obtain the processed feature information O, which can be expressed as formula (5):

[0158] O=Attend(IndexSort(X1,X2,...,X L-1 )) Formula (5)

[0159] Here, Attend represents the calculation rule for self-attention, and IndexSort refers to sorting by index.

[0160] In this embodiment, the processed feature information O is pooled to obtain pooled feature information O'.

[0161] Finally, the target output feature information OmniNet(X) of the Omnidirectional Attention layer can be expressed as Equation (6):

[0162] OmniNet(X)=Xformer(X)+O' Formula (6)

[0163] The Omninet neural network model in this application includes a Transformer Encoder structure, namely a feature extraction layer, a Transformer layer, and an Omnidirectional Attention layer. Then, the output layer is directly connected to the Encoder structure for prediction.

[0164] Based on this, in this embodiment, the target output feature information of the Omnidirectional Attention layer is predicted through the output layer to obtain the predicted operation level information. In this embodiment, the output layer can classify the target output feature information to obtain the classification probability for each preset operation level information; the predicted operation level information with the highest classification probability value is the aforementioned operation level information.

[0165] OmniNet's Attention mechanism can dynamically learn the importance of features from different time slices within a local area. In this embodiment, the Transformer layer can use only 6 neural network layers, making the OmniNet neural network model a lightweight model with <2M parameters. Such a lightweight model is easily parallelized, has fast inference speed, and can meet online performance requirements. Using Transformer layers as the basic structure of the OmniNet neural network model better supports model training and facilitates rapid model version updates.

[0166] S203. The computer equipment converges the model parameters of the operation level prediction model to be trained based on the operation level labels and predicted operation level information corresponding to the combat data samples, and obtains the trained operation level prediction model.

[0167] In this embodiment, the loss value between the operation level label and the predicted operation level information can be calculated using a loss function; the model parameters of the operation level prediction model to be trained are converged based on the loss value to obtain the trained operation level prediction model.

[0168] The loss function can be the cross-entropy loss function. The operation level prediction model to be trained uses the Adam optimizer. The learning rate and L2 regularization of the operation level prediction model to be trained are 1e-4, and the training converges in about 200 epochs. The operation level prediction model to be trained is trained using a 4 GPU server, which can enable the operation level prediction model to converge and achieve good results in a relatively short time.

[0169] Based on the above, this application further elaborates on the embodiments. The computer device in this application embodiment can be a server. The server obtains battle data samples and trains the untrained operation level prediction model on the server based on the battle data samples to obtain a trained operation level prediction model. This application embodiment can deploy the trained operation level prediction model online. The trained operation level prediction model can be triggered during the player's novice stage.

[0170] S204. The computer device acquires a set of battle status information of the target object and the opponent in the game.

[0171] The battle status information set includes several battle status information corresponding to at least one game.

[0172] In this embodiment of the application, since the candidate battle state information generated in the game between the target object and the game object is not all required for the implementation of this application, the candidate battle state information can refer to any information generated by the target object and the game object during the game. Based on this, this embodiment of the application extracts the candidate battle state information according to the time point in the time series. In this way, the computer device can extract the battle state information corresponding to the time point in the time series.

[0173] S205. The computer equipment uses a post-training operational level prediction model to extract features from the battle status information set, thereby obtaining the battle status feature information corresponding to the battle status information.

[0174] In this embodiment of the application, the battle status information includes the battle status information corresponding to each time point in the time series. Based on this, this embodiment of the application performs feature extraction on the battle status information corresponding to each time point in the time series to obtain the battle status feature information corresponding to each time point in the time series.

[0175] S206. The computer equipment predicts the operational level of the target object based on the characteristics of the battle status, and obtains the operational level information corresponding to each game.

[0176] For example, computer equipment can fuse the battle state feature information corresponding to each game to obtain the fused state feature information corresponding to each game; based on the fused state feature information corresponding to each game, a trained operation level prediction model is used to predict the operation level of the target object to obtain the operation level information corresponding to each game.

[0177] In this embodiment of the application, the battle state feature information corresponding to each time point in the time sequence of each game can be fused to obtain the fused state feature information corresponding to each game.

[0178] In this embodiment of the application, the method for fusing the battle state feature information corresponding to each time point in the time series can be to concatenate the battle state feature information corresponding to each time point in the time series in ascending order of the time series to construct temporal feature information, which is the fused state feature information.

[0179] The dimensions of the fused state feature information are [number of samples, number of features, time points]. For time series with fewer than twelve time points, zero values ​​are filled in, and all are input into the operational level prediction model to be trained.

[0180] like Figure 6 As shown in this embodiment, the trained operation level prediction model is deployed on server 11. The target player operates on the target object through game client 12 to engage in combat with the opponent, obtaining at least one game-related combat status information. Then, the combat status information is stored in database 13. The processing platform 14 can extract the combat status information corresponding to the target object from the database and send the combat status information corresponding to the target object to server 11. Based on the trained operation level prediction model on server 11, it performs operation level prediction to obtain operation level information and returns the operation level information to game client 12.

[0181] S207. The computer equipment performs fusion processing on the operational level information of the target object to obtain target operational level information for the target object.

[0182] This application embodiment fuses the operation level information of the target object to obtain the target operation level information for the target object in the following way:

[0183] For multiple game sessions, any one of the sessions is taken as the current session, and the following operations are performed: For the current session, if it is the first session, the initial overall skill level information of the target object is obtained, and the overall skill level information of the current session is calculated based on the initial overall skill level information and the target object's skill level information of the current session; if the current session is not the first session, the overall skill level information of the target object's previous session is obtained, and the overall skill level information of the current session is calculated based on the previous session's overall skill level information and the target object's skill level information of the current session; this process continues until the overall skill level information of each session in the multiple game sessions has been calculated, and the overall skill level information of each session is obtained; based on the overall skill level information of the target object, the target skill level information for the target object is determined.

[0184] The method by which computer equipment determines the target operational level information for the target object based on the comprehensive operational level information of the target object can be found in formula (7):

[0185] S t =aC t +(1-a)S t-1 Formula (7)

[0186] Among them, S t This indicates the overall skill level in the current game; 'a' represents the smoothing coefficient of 0. <a<1;S t-1 This indicates the player's overall skill level in the previous game or their initial overall skill level; C t This indicates the player's skill level in the current game. Specifically, when t=1, S... t-1 This provides initial integrated operational level information; when t≠1, S t-1 This provides information on the overall skill level of the player in the previous game.

[0187] This application embodiment can calculate the difference between the overall skill level information of the current game and the initial overall skill level information, and calculate the difference between the overall skill level information of the previous game and the overall skill level information of the current game, in order to converge the target skill level information for the target object. The convergence speed of the target skill level information for the target object in this application embodiment is as follows: Figure 7 As shown, from Figure 7As can be seen, by the third game, the difference between the overall skill level information from the second and third games had widened to form a normal distribution and remained relatively stable. With the increase in the number of games, the difference became smaller. This indicates that, with a limited number of games, quickly identifying the target player's skill level is fast and stable. Both the difference information and the overall skill level information can be represented numerically.

[0188] The existing True Skill algorithm iteratively updates the player's ELO rating based on wins and losses in matches to determine their skill level in controlling virtual objects, a very slow process. Because the existing True Skill algorithm has low accuracy in assessing a player's ELO rating during the beginner phase, using ELO rating for matchmaking reduces the overall quality of gameplay.

[0189] Furthermore, in this embodiment, all players and players with 10 or fewer matches are selected to simulate novice matches. Then, the game object processing method of this embodiment evaluates the player's skill level compared to the existing True Skill algorithm. Table 1 compares the game object processing method of this embodiment with the existing True Skill algorithm.

[0190] High-scoring team's winning games High-scoring teams' winning percentage This application 2844 60.3％ Existing True Skill algorithm 2496 49.4％

[0191] Table 1

[0192] As shown in Table 1, in the beginner stage, the target object's win rate in the team using the game object processing method of this application is 60.3%, while the win rate of the target object in the team using the TrueS kill algorithm is only 49.4%, and the win rate of the TrueSkill algorithm is lower than that of this application. It can be seen that in the beginner stage, the target operation level information obtained by the game object processing method of this application is more accurate than the ELO score used by the existing TrueSkill algorithm.

[0193] In addition to the above, this application embodiment also evaluates the difference between the game object processing method of this application embodiment and the existing True Skill algorithm by using the difference between the virtual kill count and the virtual kill count of the target object. If the strength of the target object and the target opponent are similar, the difference value is small.

[0194] Compared to the True Skill algorithm, the average KD of the target objects in this embodiment is reduced by 0.5 (12.1%), and the proportion of target objects with KD>=8 is reduced by 1.6% (13.5%), indicating that using the target objects as the matching target game objects in this embodiment can significantly improve the quality of the game.

[0195] Where K refers to the virtual kill count of the target object; D refers to the virtual kill count of the target object.

[0196] In addition, after step S207 in this embodiment, a corresponding target game object can be matched for the target object. For example, the computer device can determine the target skill level information of the target object based on the target operation skill information of the target object; obtain at least one candidate game object based on the target skill level information of the target object; and extract the target game object that matches the target object from the candidate game objects.

[0197] Based on the above, the game object processing method provided in this application has many advantages, for example, (1) the target operation level information converges faster in this application, and the target operation level information tends to stabilize after 3 games, that is, the player's game operation level has stabilized; (2) the accuracy of this application in evaluating the player's game operation ability in the novice stage is significantly better than the existing True Skill algorithm. For the accuracy of evaluating the relative strength of the target object in the game camp, the accuracy of this application is 6.9% higher than the existing True Skill algorithm; for the accuracy of evaluating the relative strength of the target object in a single game, the accuracy of this application is 11.1% higher than the existing True Skill algorithm; (3) since this application uses the battle state feature information corresponding to each time point in the time series in each game, the accuracy of this application is improved by nearly 20% compared with a single-layer neural network; (4) the application of this application in real-person games significantly improves the player experience and effectively reduces the proportion of extreme difference values ​​in the game.

[0198] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0199] This application embodiment can obtain a set of battle status information of a target object and a player in a game. The set of battle status information includes battle status information corresponding to at least one game. Features are extracted from the battle status information in the set to obtain corresponding battle status feature information. Based on the battle status feature information, the skill level of the target object is predicted to obtain skill level information corresponding to each game. The skill level information of the target object is then fused to obtain target skill level information for the target object. Because this application embodiment can determine the skill level information for each game based on the battle status information corresponding to each game between the target object and the player, the target skill level information of the target object can be accurately calculated, thereby accurately determining the skill level of the target object.

[0200] To better implement the above methods, this application also provides a game object processing device, which can be integrated into a computer device, such as a server or terminal. The terminal may include a tablet computer, a laptop computer, and / or a personal computer.

[0201] For example, such as Figure 8 As shown, the game object processing device may include an acquisition unit 301, an extraction unit 302, a prediction unit 303, a fusion unit 304, a matching unit 305, and a training unit 306, as follows:

[0202] (1) Obtain unit 301;

[0203] The acquisition unit 301 can be used to acquire a set of battle status information of target objects and game objects in the game. The set of battle status information includes battle status information corresponding to at least one game.

[0204] (2) Extraction unit 302;

[0205] The extraction unit 302 can be used to extract features from the battle status information included in the battle status information set to obtain the battle status feature information corresponding to the battle status information.

[0206] In some embodiments, the battle status information includes the battle status information corresponding to each time point in the time sequence of each game; the extraction unit 302 can be used to extract features from the battle status information corresponding to each time point in the time sequence of each game to obtain the battle status feature information corresponding to each time point in the time sequence of each game.

[0207] (3) Prediction unit 303;

[0208] The prediction unit 303 can be used to predict the operational level of the target object based on the characteristics of the battle state, so as to obtain the operational level information corresponding to each game.

[0209] In some embodiments, the prediction unit 303 may be specifically used to fuse the battle state feature information corresponding to each game to obtain the fused state feature information corresponding to each game; and based on the fused state feature information corresponding to each game, to use a trained operation level prediction model to predict the operation level of the target object to obtain the operation level information corresponding to each game.

[0210] In some embodiments, the prediction unit 303 can be specifically used to perform fusion processing on the battle state feature information corresponding to each time point in the time series based on the time series of each game, so as to obtain the fused state feature information corresponding to each game.

[0211] (4) Fusion unit 304;

[0212] The fusion unit 304 can be used to fuse the operation level information of the target object to obtain the target operation level information for the target object.

[0213] In some embodiments, the game includes multiple rounds; the fusion unit 304 can be specifically used to, for multiple rounds of the game, take any one of the rounds as the current round and perform the following operations: for the current round of the game, if the current round of the game is the first round of the game, obtain the initial comprehensive operation level information of the target object, and calculate the comprehensive operation level information of the current round of the game based on the initial comprehensive operation level information and the operation level information of the target object in the current round of the game; if the current round of the game is not the first round of the game, obtain the comprehensive operation level information of the target object in the previous round of the game, and calculate the comprehensive operation level information of the current round of the game based on the comprehensive operation level information of the previous round of the game and the operation level information of the target object in the current round of the game; until each round of the multiple rounds of the game has been calculated, the comprehensive operation level information of each round of the game is obtained; and determine the target operation level information for the target object based on the comprehensive operation level information of the target object.

[0214] (5) Matching unit 305;

[0215] The matching unit 305 can be used to determine the target skill level information of the target object based on the target operation skill information of the target object; obtain at least one candidate game object based on the target skill level information of the target object; and extract the target game object that matches the target object from the candidate game objects.

[0216] (6) Training Unit 306;

[0217] Training unit 306 can be used to acquire a set of battle data samples, which includes battle data samples labeled with operation level tags from at least one game. The battle data samples are data samples labeled with operation level tags for the object to be predicted. Based on the battle data samples, the operation level prediction model to be trained is used to predict the operation level of the object to be predicted, and the predicted operation level information of the object to be predicted for each game is obtained. Based on the operation level tags and predicted operation level information corresponding to the battle data samples, the model parameters of the operation level prediction model to be trained are converged to obtain the trained operation level prediction model.

[0218] In some embodiments, the training unit 306 can be used to acquire a first game faction and a second game faction. The first game faction includes an object to be predicted and at least one first participating object, and the second game faction includes at least one second participating object. The candidate operation level information of the object to be predicted, the candidate operation level information of the first participating object, and the candidate operation level information of the second participating object are all randomly assigned operation level information. The object to be predicted, the first participating object, and the second participating object are subjected to a battle to obtain a battle data sample. The operation level label corresponding to the battle data sample is determined according to the candidate operation level information of the object to be predicted. A battle data sample set is generated according to the battle data sample and the operation level label corresponding to the battle data sample.

[0219] In some embodiments, the training unit 306 may be specifically used to extract features from the battle data samples using the operation level prediction model to be trained, and obtain sample feature information; based on the sample feature information, to predict the operation level of the object to be predicted, and obtain the predicted operation level information of the object to be predicted for each game.

[0220] In some embodiments, the training unit 306 may be specifically used to fuse sample feature information for each game to obtain fused sample feature information; and to predict the operation level of the object to be predicted based on the fused sample feature information to obtain the predicted operation level information of the object to be predicted.

[0221] In some embodiments, the training unit 306 can be specifically used to process the object to be predicted, the first participating object, and the second participating object into a battle to obtain candidate battle data; and to extract samples from the candidate battle data according to the target time series to obtain the battle data samples corresponding to the object to be predicted in each game.

[0222] As can be seen from the above, the acquisition unit 301 of this application embodiment can be used to acquire a set of battle state information of the target object and the opponent in the game. The set of battle state information includes several battle state information corresponding to at least one game. The extraction unit 302 can be used to extract features from the battle state information included in the set of battle state information to obtain battle state feature information corresponding to the battle state information. The prediction unit 303 can be used to predict the operation level of the target object based on the battle state feature information to obtain the operation level information corresponding to each game. The fusion unit 304 can be used to fuse the operation level information of the target object to obtain the target operation level information for the target object. Since this application embodiment can determine the operation level information of each game based on the battle state information corresponding to each game between the target object and the opponent, the target operation level information of the target object can be accurately calculated, and thus the operation level of the target object can be accurately determined.

[0223] This application also provides a computer device, such as... Figure 9 As shown, it illustrates a structural schematic diagram of the computer device involved in the embodiments of this application, specifically:

[0224] The computer device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 9 The computer device structure shown does not constitute a limitation on the computer device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0225] The processor 401 is the control center of the computer device, connecting various parts of the computer device through various interfaces and lines. It performs various functions and processes data by running or executing software programs and / or modules stored in the memory 402, and by calling data stored in the memory 402. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and computer programs, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0226] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, computer programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0227] The computer device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0228] The computer device may also include an input unit 404, which can be used to receive input digital or character information communication, and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0229] Although not shown, the computer device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the computer device loads the executable files corresponding to the processes of one or more computer programs into the memory 402 according to the following instructions, and the processor 401 runs the computer programs stored in the memory 402 to realize various functions, as follows:

[0230] The system acquires a set of battle status information for the target object and the player in the game, including battle status information for at least one game. It then extracts features from the battle status information in the set to obtain corresponding battle status feature information. Based on the battle status feature information, it predicts the skill level of the target object to obtain skill level information for each game. Finally, it fuses the skill level information of the target object to obtain target skill level information for that target object.

[0231] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0232] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by a computer program, or by a computer program controlling related hardware. The computer program can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0233] Therefore, embodiments of this application provide a computer-readable storage medium storing a computer program that can be loaded by a processor to execute any of the game object processing methods provided in embodiments of this application.

[0234] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0235] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0236] Since the instructions stored in the computer-readable storage medium can execute the steps in any of the game object processing methods provided in the embodiments of this application, the beneficial effects that any of the game object processing methods provided in the embodiments of this application can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0237] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations of the above embodiments.

[0238] The above provides a detailed description of a game object processing method, apparatus, computer device, and computer-readable storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for processing game objects, characterized in that, include: Acquire a set of battle status information for target objects and match objects in the game. The set of battle status information includes battle status information corresponding to at least one game. The target object is a virtual object controlled by the target player in the game. The match object is an artificial intelligence or a virtual object controlled by a non-target player in the game. The battle status information includes battle status information corresponding to each time point in the time sequence of each game. Feature extraction is performed on the battle state information corresponding to each time point in the time sequence of each game to obtain the battle state feature information corresponding to each time point in the time sequence of each game. Based on the time sequence of each game, the battle state feature information corresponding to each time point in the time sequence is fused to obtain the fused state feature information for each game. Based on the fused state feature information corresponding to each game, the operation level prediction model after training is used to predict the operation level of the target object, thereby obtaining the operation level information corresponding to each game. The operation level information of the target object is fused to obtain the target operation level information for the target object.

2. The game object processing method according to claim 1, characterized in that, The game includes multiple rounds; the process of fusing the skill level information of the target object to obtain target skill level information for the target object includes: for the multiple game rounds, taking any one round as the current round, and performing the following operations: For the current game, if the current game is the first game, then the initial comprehensive operation level information of the target object is obtained, and the comprehensive operation level information of the current game is calculated based on the initial comprehensive operation level information and the operation level information of the target object in the current game. If the current game is not the first game, then obtain the overall skill level information of the target object in the previous game, and calculate the overall skill level information of the current game based on the overall skill level information of the previous game and the skill level information of the target object in the current game. The calculation continues until each game in the multiple games has been completed, and the overall operational skill level information for each game is obtained. Based on the comprehensive operational level information of the target object, the target operational level information for the target object is determined.

3. The game object processing method according to claim 1, characterized in that, The battle state feature information is obtained by extracting features from the battle state information using a trained operational level prediction model; before extracting features from the battle state information included in the battle state information set, the method further includes: Obtain a set of battle data samples, which includes battle data samples labeled with operation level tags from at least one game. The battle data samples are data samples labeled with operation level tags for the object to be predicted. Based on the battle data samples, the operation level prediction model to be trained is used to predict the operation level of the object to be predicted, so as to obtain the predicted operation level information of the object to be predicted for each game. Based on the operation level label and the predicted operation level information corresponding to the battle data sample, the model parameters of the operation level prediction model to be trained are converged to obtain the trained operation level prediction model.

4. The game object processing method according to claim 3, characterized in that, The acquisition of the battle data sample set includes: Obtain a first game faction and a second game faction. The first game faction includes the object to be predicted and at least one first participating object. The second game faction includes at least one second participating object. The candidate operation level information of the object to be predicted, the candidate operation level information of the first participating object, and the candidate operation level information of the second participating object are all randomly assigned operation level information. The object to be predicted, the first participating object, and the second participating object are subjected to a battle to obtain a battle data sample. Based on the candidate operation level information of the object to be predicted, determine the operation level label corresponding to the battle data sample; A set of battle data samples is generated based on the battle data samples and the corresponding operation level labels.

5. The game object processing method according to any one of claims 1 to 4, characterized in that, After fusing the operation level information of the target object to obtain the target operation level information for the target object, the method further includes: Based on the target operation level information of the target object, determine the target level grade information of the target object; Based on the target level information of the target object, at least one candidate match object is obtained; Extract the target game object that matches the target object from the candidate game objects.

6. A game object processing device, characterized in that, include: The acquisition unit is used to acquire a set of battle status information of target objects and game objects in the game. The set of battle status information includes battle status information corresponding to at least one game. The target object is a virtual object controlled by the target player in the game. The game object is a virtual object controlled by artificial intelligence or a non-target player in the game. The battle status information includes battle status information corresponding to each time point in the time sequence of each game. The extraction unit is used to extract features from the battle state information corresponding to each time point in the time sequence of each game, and obtain the battle state feature information corresponding to each time point in the time sequence of each game. The prediction unit is used to fuse the battle state feature information corresponding to each time point in the time series of each game according to the time series of each game, to obtain the fused state feature information corresponding to each game, and to use the trained operation level prediction model to predict the operation level of the target object according to the fused state feature information corresponding to each game, so as to obtain the operation level information corresponding to each game. The fusion unit is used to perform fusion processing on the operation level information of the target object to obtain target operation level information for the target object.

7. The game object processing device according to claim 6, characterized in that, The game comprises multiple rounds; the fusion unit is used to select any one of the multiple rounds as the current round and perform the following operations: For the current game, if the current game is the first game, then the initial comprehensive operation level information of the target object is obtained, and the comprehensive operation level information of the current game is calculated based on the initial comprehensive operation level information and the operation level information of the target object in the current game. If the current game is not the first game, then obtain the overall skill level information of the target object in the previous game, and calculate the overall skill level information of the current game based on the overall skill level information of the previous game and the skill level information of the target object in the current game. The calculation continues until each game in the multiple games has been completed, and the overall operational skill level information for each game is obtained. Based on the comprehensive operational level information of the target object, the target operational level information for the target object is determined.

8. The game object processing device according to claim 6, characterized in that, The battle state feature information is obtained by extracting features from the battle state information using a trained operation level prediction model; the game object processing device further includes a training unit, which is used to acquire a battle data sample set, which includes battle data samples labeled with operation level tags from at least one game, and the battle data samples are data samples labeled with operation level tags for the object to be predicted. Based on the battle data samples, the operation level prediction model to be trained is used to predict the operation level of the object to be predicted, so as to obtain the predicted operation level information of the object to be predicted for each game. Based on the operation level label and the predicted operation level information corresponding to the battle data sample, the model parameters of the operation level prediction model to be trained are converged to obtain the trained operation level prediction model.

9. A computer device, characterized in that, It includes a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to perform the game object processing method according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted for loading by a processor to perform the game object processing method according to any one of claims 1 to 5.

11. A computer program product, characterized in that, The computer program product stores a computer program adapted for loading by a processor to execute the game object processing method according to any one of claims 1 to 5.

Images