Game team recommendation method, device, server and storage medium

Abstract

The present application provides a game team recommendation method, device, server and storage medium, the method comprising: obtaining object portrait data of a target virtual object and team portrait data of a candidate game team, obtaining a first teaming probability of the target virtual object joining the candidate game team and a second teaming probability of the candidate game team agreeing to the target virtual object joining through a preset matching model according to the object portrait data and the team portrait data, determining a two-way matching probability between the target virtual object and the candidate game team according to the first teaming probability and the second teaming probability, and recommending a game team to the target virtual object according to the two-way matching probability. Among them, the two-way matching probability between the virtual object and the game team takes into account both the virtual object's teaming preference for the game team and the game team's teaming demand preference for the virtual object, and recommending a game team to the target virtual object based on the two-way matching probability can reduce the time consumption of teaming and improve the efficiency of teaming.

Description

Technical Field

[0001] The present application relates to the field of game technology, and in particular to a game team recommendation method, device, server, and computer-readable storage medium (referred to as storage medium). Background Art

[0002] With the popularization of smart terminals, electronic games running on smart terminals have developed rapidly. In the process of electronic games, forming a game team by teaming up and participating in the game together in the form of a game team is one of the common ways of playing. In electronic games, the game system recommends game teams that have not been completed to game players, and the game players select the corresponding game teams to apply for teaming. The game team can accept or reject the game players' joining based on the game players' teaming requests. However, the existing game team recommendation methods often recommend game teams to game players according to the teaming time of the game team that has not been completed or the game level of the team members, or randomly recommend game teams to game players. However, the above recommendation methods make the matching degree between game players and the recommended game teams low. This is because the considerations when game players form teams are not limited to the teaming time or game level of the players, which leads to the rejection of the game players' teaming applications by the game teams, increasing the time consumption of teaming and reducing the efficiency of teaming. Summary of the invention

[0003] Based on this, it is necessary to provide a game team recommendation method, device, computer equipment and storage medium to address the above technical problems, so as to improve the matching degree between the recommended game team and the game players and improve the team formation efficiency.

[0004] In a first aspect, the present application provides a game team recommendation method, comprising:

[0005] Obtaining object portrait data of a target virtual object and team portrait data of a candidate game team;

[0006] According to the object portrait data and the team portrait data, a first teaming probability of the target virtual object joining the candidate game team and a second teaming probability that the candidate game team agrees to the target virtual object joining are obtained through a preset matching model;

[0007] Determine a two-way matching probability between the target virtual object and the candidate game team according to the first teaming probability and the second teaming probability;

[0008] Recommend game teams to target virtual objects based on the two-way matching probability.

[0009] In some embodiments of the present application, the matching model includes a first matching model;

[0010] According to the object portrait data and the team portrait data, the first teaming probability of the target virtual object joining the candidate game team is obtained through a preset matching model, including:

[0011] extracting a first object portrait feature of the target virtual object from the object portrait data;

[0012] Extracting first team portrait features of candidate game teams from team portrait data;

[0013] The first object portrait feature and the first team portrait feature are input into the first matching model to obtain the first team formation probability.

[0014] In some embodiments of the present application, before inputting the first object portrait feature and the first team portrait feature into the first matching model to obtain the first team formation probability, the following steps are included:

[0015] Obtaining a first training sample and a first sample team label corresponding to the first training sample, the first training sample comprising a first object portrait feature of a first sample virtual object and a first team portrait feature of a first sample game team;

[0016] Inputting the first object portrait feature of the first sample virtual object and the first team portrait feature of the first sample game team into the machine learning regression model to obtain a first predicted team formation probability;

[0017] A first predicted teaming result is determined according to the first predicted teaming probability, and a model parameter of the machine learning regression model is adjusted according to an error between the first predicted teaming result and the first sample teaming label to obtain a first matching model.

[0018] In some embodiments of the present application, the matching model includes a second matching model;

[0019] According to the object portrait data and the team portrait data, the probability of the candidate game team agreeing to the target virtual object joining the second team is obtained through the preset matching model, including:

[0020] Extracting a second object portrait feature and a historical teaming sequence of the target virtual object from the object portrait data;

[0021] Extracting second team portrait features of the candidate game team from the team portrait data;

[0022] The historical team formation sequence, the second object portrait features, and the second team portrait features are input into the second matching model to obtain the second team formation probability.

[0023] In some embodiments of the present application, the historical team formation sequence includes the team capability parameters and team level of the game team that the target virtual object joins each time in the historical team formation.

[0024] In some embodiments of the present application, before inputting the historical team formation sequence, the second object portrait feature, and the second team portrait feature into the second matching model to obtain the second team formation probability, the following steps are included:

[0025] Obtaining a second training sample and a second sample team label corresponding to the second training sample, the second training sample including a second object portrait feature of a second sample virtual object, a historical team sequence of the second sample virtual object, and a second team portrait feature of a second sample game team;

[0026] Inputting the second object portrait feature of the second sample virtual object, the historical team formation sequence of the second sample virtual object, and the second team portrait feature of the second sample game team into the deep learning model to obtain a second predicted team formation probability;

[0027] A second predicted teaming result is determined according to the second predicted teaming probability, and model parameters of the deep learning model are adjusted according to the second predicted teaming result and the error between the second sample teaming label to obtain a second matching model.

[0028] In some embodiments of the present application, obtaining object portrait data of a target virtual object includes:

[0029] Obtain individual ability data, historical team behavior data, and virtual props data of the target virtual object;

[0030] According to individual ability data, historical team behavior data and virtual props data, object portrait data of the target virtual object is constructed.

[0031] In some embodiments of the present application, obtaining team portrait data of a candidate game team includes:

[0032] Obtain individual capability data and individual team data of virtual objects corresponding to each team member in the candidate game team;

[0033] Generating team capability data of the candidate game team based on the individual capability data of the virtual objects corresponding to the team members;

[0034] Generate team formation data of the candidate game team based on the individual team formation data of the virtual objects corresponding to the team members and the total team formation time of the candidate game team;

[0035] Based on the team capability data and team formation data of the candidate game teams, team portrait data of the candidate game teams are constructed.

[0036] In some embodiments of the present application, before obtaining the object portrait data of the target virtual object and the team portrait data of the candidate game team, the method further includes:

[0037] Get the game team that has not been completed;

[0038] Determine the team level of the game team that has not completed team formation according to the team portrait data of the game team that has not completed team formation;

[0039] Based on the team level of the game team that has not completed the team formation, determine the team formation capability threshold value of the game team that has not completed the team formation;

[0040] The personal ability parameters of the target virtual object are obtained, and candidate game teams are screened from game teams that have not completed team formation according to the team formation ability threshold value and the personal ability parameters of the target virtual object.

[0041] In some embodiments of the present application, determining the two-way matching probability between the target virtual object and the candidate game team according to the first teaming probability and the second teaming probability includes:

[0042] Obtain the weights of the first teaming probability and the second teaming probability;

[0043] The first teaming probability and the second teaming probability are weighted and summed according to the weights to obtain the two-way matching probability between the target virtual object and the game team.

[0044] In some embodiments of the present application, recommending a game team to a target virtual object according to the two-way matching probability includes:

[0045] Sort the candidate game teams according to the two-way matching probability to obtain a recommended team list;

[0046] A game team is recommended to the target virtual object based on the recommended team list.

[0047] In a second aspect, the present application provides a game team recommendation device, comprising:

[0048] A game team acquisition module, used to acquire object portrait data of a target virtual object and team portrait data of a candidate game team;

[0049] A team formation probability acquisition module is used to obtain a first team formation probability of a candidate game team of a target virtual object and a second team formation probability that the candidate game team agrees to the target virtual object to join, based on the object portrait data and the team portrait data, through a preset matching model;

[0050] A two-way matching acquisition module, used to determine the two-way matching probability between the target virtual object and the candidate game team according to the first teaming probability and the second teaming probability;

[0051] The game team recommendation module is used to recommend a game team to a target virtual object based on a two-way matching probability.

[0052] In a third aspect, the present application further provides a server, the server comprising:

[0053] one or more processors;

[0054] Memory; and

[0055] One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the game team recommendation method.

[0056] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, and the computer program is loaded by a processor to execute the steps in the game team recommendation method.

[0057] In a fifth aspect, an embodiment of the present application provides a computer program product or a computer program, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in the first aspect above.

[0058] The above-mentioned game team recommendation method, device, server and storage medium obtain the object portrait data of the target virtual object and the team portrait data of the candidate game team, and obtain the first teaming probability of the target virtual object joining the candidate game team and the second teaming probability of the candidate game team agreeing to the target virtual object joining through a preset matching model according to the object portrait data and the team portrait data, and then determine the two-way matching probability between the target virtual object and the candidate game team according to the first teaming probability and the second teaming probability, and then recommend the game team to the target virtual object according to the two-way matching probability. Among them, the two-way matching probability between the target virtual object and the candidate game team takes into account both the virtual object's teaming preference for the game team and the game team's teaming demand preference for the virtual object. When recommending a game team to the target virtual object, recommending the game team based on the two-way matching probability can improve the accuracy of recommending the game team to the target virtual object, reduce the situation where the teaming request is rejected, reduce the time consumption of teaming, and improve the efficiency of teaming. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0060] Figure 1 Schematic diagram of a game team recommendation method in an embodiment of the present application;

[0061] Figure 2 is a flowchart of a method for recommending a game team in an embodiment of the present application;

[0062] Figure 3 It is a flowchart of the first teaming probability acquisition step in the embodiment of the present application;

[0063] Figure 4 It is a flowchart of the second teaming probability acquisition step in an embodiment of the present application;

[0064] Figure 5 It is a flowchart of the steps of recommending game teams based on two-way matching probability in an embodiment of the present application;

[0065] Figure 6 is a flowchart of the candidate game team screening step in the embodiment of the present application;

[0066] Figure 7 is a flowchart of the first matching model training step in an embodiment of the present application;

[0067] Figure 8 is a flowchart of the second matching model training step in an embodiment of the present application;

[0068] Fig. 9 is a schematic diagram of the structure of a game team recommendation device in an embodiment of the present application;

[0069] Fig.10 It is a schematic diagram of the structure of the server in the embodiment of the present application. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0071] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0072] In the description of the present application, the word "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.

[0073] Among them, the virtual object mentioned in the embodiment of the present application refers to a dynamic object that can be controlled in a virtual scene. Optionally, the dynamic object can be a virtual character, a virtual animal, a cartoon character, etc. The virtual object is a character controlled by the player through an input device, or an artificial intelligence (AI) set in a virtual environment battle through training, or a non-player character (NPC) set in a virtual scene battle. Optionally, the virtual object is a virtual character competing in a virtual scene. In a possible implementation, the user can control the virtual object to move in the virtual scene, for example, control the virtual object to run, jump, crawl, etc., and can also control the virtual object to use the skills, virtual props, etc. provided by the application to fight with other virtual objects. (Game) props refer to props that virtual objects can use in a virtual environment, including but not limited to firearms, cold weapons, grenades, shields, heroic pets, etc., which can be used by virtual objects to speed up their own attributes, assist in combat, or initiate damage to other virtual objects. Virtual props can also be supply props such as bullets, and can also be accessories such as extended magazines, aiming scopes, flash suppressors, and butts that are assembled on designated virtual weapons. The game team mentioned in the embodiments of this application includes multiple virtual objects.

[0074] In the embodiment of the present application, it should be noted that the game team recommendation method provided in the embodiment of the present application can be applied to Figure 1In the game team recommendation system shown in the figure. The game team recommendation system includes a terminal 100 and a server 200. The terminal 100 can be a desktop terminal or a mobile terminal. The terminal 100 can also be a mobile phone, a tablet computer, a laptop computer, etc. The server 200 can be an independent server, or a server network or server cluster composed of servers, including but not limited to a computer, a network host, a single network server, a plurality of network server sets or a cloud server composed of multiple servers. The cloud server is composed of a large number of computers or network servers based on cloud computing.

[0075] The terminal 100 may be a device running a game application, which may be a game application in the form of an application program or a web version of a game application, and the running form of the game application is not limited in any way. The terminal 100 faces the game player, and responds to the operation of the game player by sending a corresponding instruction request to the server 200 corresponding to the game application, and at the same time, receives data information returned by the server 200, and displays the relevant data information to the game player through a display device.

[0076] Specifically, before the server 200 executes the game team recommendation method, the terminal 100 responds to the operation of the game player and sends a team formation request to the server 200. The team formation request carries the identifier of the target virtual object operated by the game player. After receiving the team formation request, the server 200 can obtain the two-way matching probability between the target virtual object and the candidate game team that has not completed the team formation according to the game team recommendation method, and recommend the game team to the target virtual object based on the two-way matching probability; or, before the server 200 executes the game team recommendation method, the server 200 obtains the virtual object that is not currently in a team, determines any virtual object as the target virtual object, and can recommend the game team to the target virtual object according to the game team recommendation method. After the server 200 obtains the recommended game team, it can send the recommended game team to the terminal 100 corresponding to the target virtual object, and the terminal 100 corresponding to the target virtual object can display the recommended game team to the game player on the game interface through the display device.

[0077] It should be noted that the game team recommendation method provided in the embodiment of the present application is applicable to various game scenarios that require team formation. For example, the game team recommendation method can be applied to MMO (Massive Multiplayer Online) games. Figure 1The scenario diagram of the game team recommendation system shown is only an example. The game team recommendation system and scenario described in the embodiment of the present invention are intended to more clearly illustrate the technical solution of the embodiment of the present invention, and do not constitute a limitation on the technical solution provided by the embodiment of the present invention. Those skilled in the art can know that with the evolution of the game team recommendation system and the emergence of new business scenarios, the technical solution provided by the embodiment of the present invention is also applicable to similar technical problems.

[0078] Those skilled in the art will understand that Figure 1 The application environment shown in the figure is only one application scenario of the present application solution and does not constitute a limitation on the application scenario of the present application solution. Other application environments may also include Figure 1 More or less computer equipment as shown in Figure 1 Only one server 200 is shown in the figure. It is understandable that the game team recommendation system may also include one or more other servers, which are not specifically limited here. Figure 1 As shown, the game team recommendation system may also include a memory for storing data, such as object portrait data of virtual objects, game portrait data of game teams, etc.

[0079] See also Figure 2 The present application embodiment provides a game team recommendation method, which is mainly applied to the above Figure 1 Taking the server 200 in FIG. 1 as an example, the method includes steps S210 to S240, which are specifically as follows:

[0080] S210, obtaining object portrait data of the target virtual object and team portrait data of the candidate game teams.

[0081] The target virtual object refers to a virtual object that has not joined any game team and needs to form a team. The object portrait data of the virtual object may include information data describing the individual capabilities, historical operation behaviors, etc. of the virtual object.

[0082] In one embodiment, obtaining object portrait data of the target virtual object may specifically include: obtaining individual ability data, historical team behavior data, and virtual props data of the target virtual object; and constructing object portrait data of the target virtual object based on the individual ability data, historical team behavior data, and virtual props data.

[0083] Specifically, object portrait data includes but is not limited to individual ability data of virtual objects, historical team behavior data and virtual props data, among which individual ability data includes but is not limited to individual ability parameters of virtual objects (such as level value, cultivation value, experience value, etc.), character occupation, consumption information and login information, etc., historical team behavior data includes but is not limited to the number of team formations within a historical time window (such as the past month, week, day, hour, etc.), the team ability parameters of the team joined each time, the game time of each team formation, the experience improvement of each team formation, the level improvement of each team formation and the number of deaths, etc., and virtual props data includes but is not limited to the level value and experience value of virtual props, etc.

[0084] Among them, according to the rules of different games, game teams often have a team size threshold, such as 4, 5 or other specified numbers of virtual objects (i.e., game players) must form a team successfully before the game can start; candidate game teams are game teams whose team size does not reach the team size threshold. The team portrait data of the game team includes information data such as the overall capabilities and operation records of the game team, which can be used to reversely predict the matching degree between the newly added virtual objects and the corresponding game team.

[0085] In one embodiment, obtaining team portrait data of a candidate game team may specifically include: obtaining individual capability data and individual team formation data of virtual objects corresponding to each team member in the candidate game team; generating team capability data of the candidate game team based on the individual capability data of virtual objects corresponding to each team member; generating team formation data of the candidate game team based on the individual team formation data of virtual objects corresponding to each team member and the total team formation time of the candidate game team; constructing team portrait data of the candidate game team based on the team capability data and team formation data of the candidate game team.

[0086] Among them, the team portrait data may include the team capability data, team formation data, etc. of the game team; specifically, the team capability data may include but are not limited to team capability parameters, team influence values, and individual capability data of virtual objects corresponding to each current team member, wherein the team capability parameters may be described by the average value of the individual capability data of the virtual objects corresponding to the team members, and the team influence value may be represented by the maximum value of the individual capability data of the virtual objects corresponding to the team members; the team formation data includes but is not limited to the total formation time of the current team, the individual formation data of the virtual objects corresponding to each team member (for example, the formation time and number of formations of the virtual objects corresponding to each team member), etc.

[0087] It is understandable that the team portrait data changes with the changes in the corresponding virtual objects of the team members in the game team. For example, every time a virtual object is added to the game team, the team portrait data can be updated in real time based on the data information related to the newly added virtual object.

[0088] Specifically, after determining the target virtual object and the candidate game team, the server can extract portrait data corresponding to the target virtual object and the candidate game team from the game log data, and characterize the target virtual object and the candidate game team through the object portrait data and the team portrait data, thereby improving the accuracy of the subsequent target virtual object and the candidate game team formation probability.

[0089] S220, according to the object portrait data and the team portrait data, obtain a first teaming probability that the target virtual object joins the candidate game team and a second teaming probability that the candidate game team agrees to the target virtual object joining through a preset matching model.

[0090] Among them, the first teaming probability is used to represent the probability score that the target virtual object is willing to join the candidate game team, that is, the player's teaming probability; the second teaming probability is used to represent the candidate game team's demand score for the target virtual object, that is, the team's teaming probability.

[0091] The matching model is used to obtain the matching degree between the virtual object and the game team, that is, the team formation probability. Specifically, the matching model here is a matching model that has been trained, and the matching model can be a machine learning model, a deep learning model, or a neural network model.

[0092] Specifically, the matching model may include a first matching model and a second matching model, wherein the first matching model is used to obtain the probability score of the game team that the virtual object is willing to join, that is, the first teaming probability, and the second matching model is used to obtain the game team's demand score for the player corresponding to the virtual object, that is, the second teaming probability that the game team agrees to the virtual object to join. After obtaining the object portrait data of the target virtual object and the team portrait data of the candidate game team, the server inputs the object portrait data and the team portrait data into the first matching model and the second matching model respectively, obtains the first teaming probability through the first matching model, and obtains the second teaming probability through the second matching model.

[0093] In specific applications, the number of candidate game teams is often multiple. The server inputs the object portrait data of the target virtual object and the team portrait data of a candidate game team into the matching model, and the matching model correspondingly outputs the first teaming probability and the second teaming probability of the target virtual object and the candidate game team.

[0094] In one embodiment, see Figure 3, according to the object portrait data and the team portrait data, obtaining the first teaming probability of the target virtual object joining the candidate game team, including:

[0095] S310, extracting a first object portrait feature of a target virtual object from the object portrait data;

[0096] S320, extracting first team portrait features of the candidate game team from the team portrait data;

[0097] S330: Input the first object portrait feature and the first team portrait feature into a first matching model to obtain a first team formation probability.

[0098] Among them, the first object portrait feature is information extracted from the object portrait data of the target virtual object that can reflect the relevant attributes of the target virtual object, including at least the basic object portrait feature of the target virtual object and the team capability feature of the historical game team; the historical game team is the game team that the target virtual object joined in the historical team formation. Among them, the basic object portrait feature can reflect the game level information of the target virtual object, including at least feature information representing individual capability data, feature information representing historical team formation behavior data, and feature information representing virtual props data; the team capability feature of the historical game team can reflect the capability feature of the game team that the target virtual object is willing to join, which can be constructed by the team capability parameter of the game team with the longest team formation time in the historical team formation, or by the average value of the team capability parameters of all game teams in the historical team formation.

[0099] Among them, the first team portrait feature is information extracted from the team portrait data of the candidate game team, which can reflect the attribute level of the candidate game team, including at least the basic team portrait feature and team ability feature of the candidate game team; the basic team portrait feature is used to reflect the game level information of the candidate game team, including at least feature information representing the team ability data and feature information representing the team formation data; the team ability feature is used to represent the ability characteristics of the selected game team, which can be constructed by team ability parameters.

[0100] Among them, the first matching model can be a trained machine learning model, specifically a machine learning regression model based on the GBM (GradientBoosting Machine) algorithm. The first matching model analyzes and processes the first object portrait features and the first team portrait features, and outputs the first teaming probability of the target virtual object and the candidate game team successfully forming a team.

[0101] Specifically, after obtaining the first object portrait feature and the first team portrait feature, the first object portrait feature and the first team portrait feature are input into the first matching model, and the difference feature vector between the team capability characteristics of the historical game team and the team capability characteristics of the candidate game team is obtained through the first matching model, and then based on the difference feature vector, the basic object portrait feature of the target virtual object and the basic team portrait feature of the candidate game team, the probability of the target virtual object and the candidate game team forming a team is predicted.

[0102] It is understandable that when evaluating the first teaming probability of a game player (corresponding to the target virtual object) joining a candidate game team, the team capability characteristics of the target virtual object in the game team that has successfully teamed up in the past have a high reference value, that is, the team capability characteristics of the game team that has successfully teamed up in the past can be used to represent the relevant feature information of the game team that the target virtual object is willing to join. When calculating the degree of match between the target virtual object and the game team (e.g., the first teaming probability), by taking into account the relevant feature information of the game team that has successfully teamed up in the past, the teaming probability of the target virtual object joining the candidate game team can be more accurately predicted, the accuracy of the match can be improved, and the probability of the target virtual object being rejected by the candidate game team can be reduced.

[0103] In one embodiment, see Figure 4 , according to the object portrait data and the team portrait data, obtaining the first teaming probability of the target virtual object joining the candidate game team, including:

[0104] S410, extracting a second object portrait feature and a historical team formation sequence of the target virtual object from the object portrait data;

[0105] S420, extracting a second team portrait feature of the candidate game team from the team portrait data;

[0106] S430, inputting the historical team formation sequence, the second object portrait feature and the second team portrait feature into the second matching model to obtain a second team formation probability.

[0107] The second object portrait feature is information extracted from the object portrait data of the target virtual object that can reflect the relevant attributes of the target virtual object, and at least includes the basic object portrait feature and object ability feature of the target virtual object. The basic object portrait feature can reflect the game level information of the target virtual object, and at least includes feature information representing individual ability data, feature information representing historical team behavior data, and feature information representing virtual props data; the object ability feature can reflect the ability feature of the target virtual object, and can be constructed by the individual ability parameters of the target virtual object.

[0108] Among them, the historical team sequence includes the team information of the game team that the target virtual object joined each time in the historical team, and the team information of the game team includes team capability parameters, team level, etc. The historical team sequence can reflect the team information of the game team that the target virtual object is willing to join. Specifically, when the second matching model predicts the probability of the second team, the target virtual object pays different attention to different candidate game teams. The candidate game teams related to the history (i.e., the game teams joined in the historical team, and the candidate game teams similar to the game teams joined in the historical team) pay more attention, and the candidate game teams not related to the history (i.e., the candidate game teams that are not similar to the game teams joined in the historical team) can even be ignored.

[0109] For example, when the candidate game team includes a virtual object that has teamed up with a historical team, the probability of the target virtual object teaming up with the candidate game team is greater; for another example, if the team level of the candidate game team is the same as the team level of the game team joined in the historical team, the probability of the target virtual object teaming up with the candidate game team is greater; if the team level of the candidate game team is different from the team level of the game team joined in the historical team, the probability of the target virtual object teaming up with the candidate game team is smaller.

[0110] By introducing the historical team formation sequence, the model can be helped to characterize the relevant information of the game team that the target virtual object is interested in, which can effectively improve the accuracy of the second team formation probability.

[0111] Among them, the second team portrait feature is information extracted from the team portrait data of the candidate game team that can reflect the attribute level of the candidate game team, including at least the basic team portrait feature of the candidate game team and the object capability feature of the historical virtual object. The historical virtual object refers to the virtual object that applied to join the candidate game team but was not agreed by the candidate game team. Among them, the basic team portrait feature is used to reflect the game level information of the candidate game team, including at least the feature information representing the team capability data and the feature information representing the team formation data; the object capability feature of the historical virtual object is used to reflect the capability feature of the historical virtual object rejected by the candidate game team during the team formation process, and can be constructed by the average value of the object capability parameters of the historical virtual object.

[0112] Among them, the second matching model can be a trained deep learning model, specifically a Transformer model. The second matching model analyzes and processes the historical team formation sequence, the second object portrait features and the second team portrait features, and outputs the second team formation probability of successful team formation between the target virtual object and the candidate game team.

[0113] Specifically, after obtaining the second object portrait feature and the second team portrait feature, the second object portrait feature and the second team portrait feature are input into the second matching model, and the object difference feature vector between the object ability feature of the target virtual object and the object ability feature of the historical virtual object is obtained through the second matching model, and the attention weights of different candidate game teams are obtained based on the historical team formation sequence of the target virtual object and the basic team portrait features of the candidate game teams, and then based on the object difference feature vector, the basic object portrait features of the target virtual object, the basic team portrait features of the candidate game teams and the attention weights, the team formation probability of the target virtual object and the candidate game team is predicted to obtain the second team formation probability.

[0114] It is understandable that when evaluating whether the candidate game team agrees to the probability of the second teaming of the game player (corresponding to the target virtual object), the object capability characteristics of the historical virtual object rejected by the candidate game team in the historical teaming process have a high reference value, that is, the object capability characteristics of the historical virtual object can be used to characterize the relevant feature information of the candidate game team rejecting the game player. In addition, the attention weights of different candidate game teams are obtained through the historical teaming sequence, so that the second teaming probability of the candidate game team related to the history and the target virtual object is increased, and the second teaming probability of the candidate game team not related to the history and the target virtual object is reduced. When calculating the matching degree (such as the second teaming probability) between the target virtual object and the game team, by taking into account the relevant feature information of the historical virtual object that the candidate game team has rejected or cooperated with in the historical teaming process, and the team information of the target virtual object in the game team that has successfully teamed in the history, the candidate game team is more truly and accurately predicted to agree to the target virtual object to join the teaming probability, improve the accuracy of the matching degree, and reduce the probability of the target virtual object being rejected by the candidate game team.

[0115] For example, virtual object A and virtual object B simultaneously send team application to game team Team C. Among them, virtual object A has better individual ability data performance, such as higher level value, cultivation value or experience value, virtual object A has no team history with game team Team C, while virtual object B has team history with game team C. According to the game team recommendation method of the prior art, the game team is recommended according to the game level of the game members. Compared with virtual object B, virtual object A has a higher matching degree with game team Team C, that is, game team Team C is more likely to agree to virtual object A's team application. However, in actual situations, considering the historical team situation, game team Team C often rejects virtual object A's team application and agrees to virtual object B's team application. In an embodiment of the present application, the attention weights of the candidate game teams are adjusted by the historical teaming sequence, so as to increase the probability of the second teaming between the historically relevant candidate game teams and the target virtual object, and reduce the probability of the second teaming between the historically irrelevant candidate game teams and the target virtual object. This increases the probability of the second teaming that the game team Team C agrees to the teaming application of the virtual object B, and reduces the probability of the second teaming that the game team Team C agrees to the teaming application of the virtual object A. This enables the game team's preference for the teaming demand of the virtual object to be taken into consideration when recommending the game team to the target virtual object, thereby reducing the number of situations in which the teaming request is rejected.

[0116] When recommending a game team to a target virtual object, the first matching model and the second matching model are used to predict the matching degree between the target virtual object and the candidate game team based on multi-dimensional feature factors, such as historical virtual objects that the candidate game team has rejected or cooperated with in the historical team formation process, game teams that the target virtual object has successfully teamed with in the past, and other relevant feature information, thereby improving the accuracy of recommending game teams to the target virtual object, reducing the number of rejections of team formation requests to reduce the time spent on team formation, improving the gaming experience of the game players corresponding to the target virtual object, and improving the retention rate of the game players.

[0117] S230: Determine a two-way matching probability between the target virtual object and the candidate game team according to the first teaming probability and the second teaming probability.

[0118] Among them, the two-way matching probability represents the probability of mutual selection between the target virtual object and the candidate game team.

[0119] Specifically, after obtaining the first teaming probability and the second teaming probability, the server can obtain the sum of the first teaming probability and the second teaming probability as the two-way matching probability between the target virtual object and the candidate game team; it can also obtain the weighted sum between the first teaming probability and the second teaming probability as the two-way matching probability between the target virtual object and the candidate game team.

[0120] More specifically, in one embodiment, step S230 includes: obtaining weights of the first teaming probability and the second teaming probability; and performing weighted summation of the first teaming probability and the second teaming probability according to the weights to obtain a two-way matching probability between the target virtual object and the game team.

[0121] The server configures weights for the first team probability and the second team probability respectively, and uses the weighted sum of the two as the two-way matching probability. The weights of the first team probability and the second team probability can be set according to actual conditions. For example, when the player's first team probability is more concerned, the weight of the first team probability can be set larger than the weight of the second team.

[0122] S240: recommending a game team to the target virtual object according to the two-way matching probability.

[0123] Among them, after determining the two-way matching probability between the target virtual object and the candidate game teams, the target game team recommended to the target virtual object can be determined from the candidate game teams according to the two-way matching probability. For example, the candidate game teams ranked in descending order of the two-way matching probability are taken as the target game teams and recommended to the target virtual object. The candidate game teams can also be ranked according to the two-way matching probability, and when recommending the game teams to the target virtual object, they are recommended according to the ranking results, so that the candidate game teams with higher two-way matching probabilities are ranked first, and the candidate game teams with lower two-way matching probabilities are ranked last.

[0124] In one embodiment, Figure 5 As shown, step S240 includes: step S510, sorting the candidate game teams according to the two-way matching probability to obtain a recommended team list; step S520, recommending the game team to the target virtual object based on the recommended team list.

[0125] Among them, the server sorts the candidate game teams according to the two-way matching probability, obtains a recommended team list, and then pushes the recommended team list to the terminal corresponding to the target virtual object. The terminal corresponding to the target virtual object can display the recommended team list to the game players on the game interface through the display device, so that the terminal corresponding to the target virtual object can recommend and display the candidate game teams according to the two-way matching probability.

[0126] In the above-mentioned game team recommendation method, the object portrait data of the target virtual object and the team portrait data of the candidate game team are obtained; according to the object portrait data and the team portrait data, the first teaming probability of the target virtual object joining the candidate game team and the second teaming probability of the candidate game team agreeing to the target virtual object joining are obtained through a preset matching model; according to the first teaming probability and the second teaming probability, the two-way matching probability between the target virtual object and the candidate game team is determined; and the game team is recommended to the target virtual object according to the two-way matching probability. Among them, the two-way matching probability between the target virtual object and the candidate game team takes into account both the virtual object's teaming preference for the game team and the game team's teaming demand preference for the virtual object. When recommending a game team to the target virtual object, the game team is recommended based on the two-way matching probability, which can improve the accuracy of recommending the game team to the target virtual object, reduce the situation where the teaming request is rejected and reduce the time spent on teaming, improve the efficiency of teaming, so that the game players corresponding to the target virtual object have a better game experience and improve the retention rate of game players.

[0127] In specific game applications, game teams with poor overall capabilities often hope to get game benefits that exceed expectations, and therefore expect the capabilities of the virtual objects to be added to be higher; game teams with medium overall capabilities can usually pass levels smoothly, and hope to improve the efficiency of passing levels, and therefore expect the capabilities of the virtual objects to be added to be similar; game teams with high overall capabilities can usually pass levels easily, and do not need to consider the capabilities of the newly added virtual objects. Game teams with higher capabilities will not lose players because they are matched with players with poor capabilities (i.e., virtual objects), but game teams with lower capabilities will find it difficult to get a good game experience if they are matched with players with poor capabilities (i.e., virtual objects), and the game user churn rate is greatly increased.

[0128] Therefore, in one embodiment, see Figure 6 , before obtaining the object portrait data of the target virtual object and the team portrait data of the candidate game team, it also includes:

[0129] S610, obtaining a game team that has not yet been formed;

[0130] S620, determining the team level of the game team that has not yet completed team formation according to the team portrait data of the game team that has not yet completed team formation;

[0131] S630, determining a team formation capability threshold value of the game team that has not completed team formation based on the team level of the game team that has not completed team formation;

[0132] S640, obtaining personal ability parameters of the target virtual object, and screening candidate game teams from game teams that have not yet completed team formation according to the team formation ability threshold value and the personal ability parameters of the target virtual object.

[0133] Among them, the team level is used to characterize the ability level of the game team. For example, it can be divided into low-end team level, mid-level team level, and high-end team level, and for example, it can be divided into A level, B level, C level, D level and E level, and the ability level from A level to E level gradually decreases. After obtaining the game team that has not completed the team, the server can classify the game team according to the team portrait data of the game team to obtain the team level of the game team. Specifically, the team portrait data of the game team that has not completed the team can be input into a pre-trained classifier, and the game team that has not completed the team is classified using the classifier to obtain its corresponding team level, wherein the classifier can adopt a classifier based on machine learning, such as based on logistic regression (Logistic Regression, LR) or GBM classifier, or it can be a classifier based on deep learning, such as a classifier based on CNN (Convolutional Neural Networks, Convolutional Neural Network) model construction.

[0134] Among them, the team capability threshold value refers to the minimum value that the individual capability parameters of the virtual object joining the corresponding game team need to reach. It is understandable that different team levels correspond to different team capability threshold values. Specifically, the server obtains the team level of the game team that has not yet completed the team, and determines the team capability threshold value corresponding to the game team that has not yet completed the team based on the team level of each unfinished team. After determining the personal capability parameters of the target virtual object, for any unfinished team, it is determined whether the capability parameters of the target virtual object are greater than the team capability threshold value of the game team that has not yet completed the team. When the capability parameters of the target virtual object are greater than the team capability threshold value of the game team that has not yet completed the team, the game team that has not yet completed the team can be determined as a candidate game team.

[0135] Taking the team levels of the game team including the low-end team level, the mid-level team level and the high-end team level as an example, assuming that the team formation ability threshold values ​​of the low-end team level game team include: the level threshold value is 50, the cultivation threshold value is 100 and the experience threshold value is 150, the team formation ability threshold values ​​of the mid-level team level game team include: the level threshold value is 35, the cultivation threshold value is 70 and the experience threshold value is 100, and the team formation ability threshold values ​​of the high-end team level game team include: the level threshold value is 10, the cultivation threshold value is 40 and the experience threshold value is 50; and the personal ability parameters of the target virtual object corresponding to the game player are: level threshold The value is 45, the cultivation threshold is 72, and the experience threshold is 101. Among the game teams that have not yet completed team formation, the game teams with mid-level and high-end team levels can be used as candidate game teams corresponding to the target virtual object, so as to avoid recommending low-ability game teams to the game player, resulting in the game player being frequently rejected when sending team-forming invitations to low-ability game teams, resulting in the target virtual object taking too long to form a team and reducing efficiency, or after the game player joins the low-ability game team, the game team has a low clearance rate or poor income, resulting in the game experience of the game players in the game team being greatly reduced and leading to the loss of game players.

[0136] In order to perform player-oriented matching scoring for the target virtual object and the candidate game team, it is necessary to train the first matching model in advance, and then obtain the trained first matching model. Figure 7 , before inputting the first object portrait feature and the first team portrait feature into the first matching model to obtain the first team formation probability, including:

[0137] S710, obtaining a first training sample and a first sample team label corresponding to the first training sample, the first training sample including a first object portrait feature of a first sample virtual object and a first team portrait feature of a first sample game team;

[0138] S720, inputting a first object portrait feature of a first sample virtual object and a first team portrait feature of a first sample game team into a machine learning regression model to obtain a first predicted team formation probability;

[0139] S730, determining a first predicted teaming result according to the first predicted teaming probability, and adjusting model parameters of the machine learning regression model according to an error between the first predicted teaming result and the first sample teaming label to obtain a first matching model.

[0140] When the server trains the first matching model, it is necessary to first obtain the first training sample and the first sample team label corresponding to the first training sample. Each first training sample corresponds to a sample virtual object and a portrait feature corresponding to a sample game team. The first sample team label can be specifically identified as "0" or "1". When the first sample team label is identified as "0", it indicates that the sample virtual object and the sample game team in the corresponding first training sample have failed to form a team. When the first sample team label is identified as "1", it indicates that the sample virtual object and the sample game team in the corresponding first training sample have successfully formed a team.

[0141] Specifically, the server can randomly collect teaming records of any virtual object and game team from the log data of the game application, and then obtain the first object portrait feature of the corresponding virtual object and the first game team feature of the game team according to the teaming record to generate a first training sample, and generate a corresponding first sample teaming label according to the teaming result in the teaming record.

[0142] Among them, the server pre-builds a machine learning regression model as the trained model. After the construction is completed, the machine learning regression model is trained using the acquired first training sample and its corresponding first sample team label. After the machine learning regression model meets the training end condition, the first matching model for actual application is constructed according to the model parameters of the machine learning regression model that meets the training end condition. Specifically, the server inputs each first training sample into the machine learning regression model, and the machine learning regression model analyzes and processes the first object portrait feature and the first team portrait feature in the first training sample, outputs the first predicted teaming probability corresponding to the first training sample, and determines the corresponding classification result according to the first predicted teaming probability, that is, the first predicted teaming result, and the classification result includes a successful teaming category and a failed teaming category; then, according to the error between the first predicted teaming result and the first sample teaming label, the model parameters in the machine learning regression model are adjusted to optimize the machine learning regression model. When the machine learning regression model meets the training end condition, the first matching model for practical application is obtained, wherein the training end condition can be set or adjusted according to actual needs. For example, the error between the first predicted teaming result and the first sample teaming label can be set to be less than a predetermined smaller value, and then it can be considered that the training end condition is reached. It can also be set that the model parameter adjustment change in the machine learning regression model between the previous and next two iterations is very small, and a threshold can be set. When it is less than this threshold, it can be considered that the training end condition is reached. The maximum number of iterations can also be set. When the iteration exceeds the maximum number of iterations, it can be considered that the training end condition is reached.

[0143] Similarly, in order to score the target virtual object and the candidate game team for team matching, it is necessary to train the second matching model in advance, and then obtain the trained second matching model. Figure 8 , before inputting the historical team formation sequence, the second object portrait feature and the second team portrait feature into the second matching model to obtain the second team formation probability, including:

[0144] S810, obtaining a second training sample and a second sample team label corresponding to the second training sample, wherein the second training sample includes a second object portrait feature of a second sample virtual object, a historical team sequence of the second sample virtual object, and a second team portrait feature of a second sample game team;

[0145] S820, inputting the second object portrait feature of the second sample virtual object, the historical team formation sequence of the second sample virtual object, and the second team portrait feature of the second sample game team into the deep learning model to obtain a second predicted team formation probability;

[0146] S830, determining a second predicted teaming result according to the second predicted teaming probability, and adjusting model parameters of the deep learning model according to the error between the second predicted teaming result and the second sample teaming label to obtain a second matching model.

[0147] Among them, when the server trains the deep learning model, it is necessary to first obtain the second training sample and the second sample team label corresponding to the second training sample. Each second training sample corresponds to two sample virtual objects and a sample game team corresponding to the portrait feature. The data format of the second sample team label is the same as the first sample team label, which will not be repeated here. Specifically, the server can randomly collect team records of any virtual object and game team from the log data of the game application, and then obtain the second object portrait feature of the corresponding virtual object and the historical team sequence and the second game team feature of the game team according to the team record to generate the second training sample, and generate the corresponding second sample team label according to the team result in the team record.

[0148] Among them, the server pre-builds a deep learning model as the trained model. After the construction is completed, the deep learning model is trained using the acquired second training samples and their corresponding second sample team labels. After the deep learning model meets the training end conditions, the second matching model for actual application is constructed according to the model parameters of the deep learning model that meets the training end conditions.

[0149] In the training process, the deep learning model introduces the historical team sequence of the second sample virtual object, and trains the deep learning model based on the attention mechanism, so that when the deep learning model predicts the probability of the second team, it adjusts the attention weights of different game teams for the input virtual object, such as paying more attention to historically related game teams (for example, game teams similar to the game teams added in the historical team of the virtual object), and even ignoring historically irrelevant game teams (for example, candidate game teams that are not similar to the game teams added in the historical team of the virtual object). By introducing the historical team sequence, the model can be helped to characterize the game teams that the target virtual object is interested in, adjust the model parameters in the deep learning model, and improve the accuracy of the deep learning model in predicting the probability of the second team.

[0150] Furthermore, the historical teaming sequence may include the team level of the game team joined by the second sample virtual object each time in the historical teaming. By introducing the historical teaming sequence, the deep learning model pays more attention to the second sample game team with the same team level as the game team joined in the historical teaming, and pays less attention to the second sample game team with a different team level from the game team joined in the historical teaming. This achieves the adjustment of the parameters of the deep learning model that influence the second teaming probability based on the historical teaming sequence of the input virtual object and the team level of the game team, so as to improve the accuracy of the deep learning model in predicting the probability of the second teaming.

[0151] In addition, as mentioned above, a game team with poor overall team ability often expects the ability of the virtual object to be added to be higher, a game team with medium overall team ability expects the ability of the virtual object to be added to be similar, and a game team with high overall team ability does not need to consider the ability of the newly added virtual object. Therefore, in the process of training the deep learning model, the influencing parameter of the deep learning model based on the team level of the input game team on the probability of the second team can also be adjusted through the historical team sequence; specifically, the lower the team level of the input game team, the greater the ability parameter of the corresponding input virtual object, the greater the probability of the second team; when the team level of the input game team is medium, the ability parameter of the corresponding input virtual object is greater than the team ability parameter of the input game team, the greater the probability of the second team, and when the ability parameter of the corresponding input virtual object is less than the team ability parameter of the input game team, the greater the probability of the second team; the higher the team level of the input game team, the smaller the ability parameter of the corresponding input virtual object, the greater the probability of the second team.

[0152] Specifically, the server inputs each second training sample into the deep learning model, and the deep learning model analyzes and processes the second object portrait features, historical team sequence, and second team portrait features in the second training sample, outputs the second predicted team probability corresponding to the second training sample, and determines the corresponding classification result based on the second predicted team probability, that is, the second predicted team result, and the classification result includes a successful team category and a failed team category; then, based on the error between the second predicted team result and the second sample team label, the model parameters in the deep learning model are adjusted to optimize the deep learning model. When the deep learning model meets the training end conditions, the second matching model for practical application is obtained. It can be understood that the training end conditions of the deep learning model are the same as the training end conditions of the above-mentioned machine learning regression model, which will not be repeated here.

[0153] In order to better implement the game team recommendation method provided in the embodiment of the present application, on the basis of the game team recommendation method provided in the embodiment of the present application, a game team recommendation device is also provided in the embodiment of the present application, such as Fig. 9 As shown, the game team recommendation device 900 includes:

[0154] A game team acquisition module 910 is used to acquire object portrait data of a target virtual object and team portrait data of a candidate game team;

[0155] A team formation probability acquisition module 920 is used to obtain a first team formation probability of a candidate game team of a target virtual object and a second team formation probability of a candidate game team agreeing to the target virtual object to join the team according to the object portrait data and the team portrait data through a preset matching model;

[0156] A two-way matching acquisition module 930, for determining a two-way matching probability between a target virtual object and a candidate game team according to the first teaming probability and the second teaming probability;

[0157] The game team recommendation module 940 is used to recommend a game team to the target virtual object according to the two-way matching probability.

[0158] In some embodiments of the present application, the matching model includes a first matching model; the team probability acquisition module 920 is also used to extract the first object portrait feature of the target virtual object from the object portrait data; extract the first team portrait feature of the candidate game team from the team portrait data; input the first object portrait feature and the first team portrait feature into the first matching model to obtain the first team probability.

[0159] In some embodiments of the present application, the team probability acquisition module 920 is also used to obtain a first training sample and a first sample team label corresponding to the first training sample, the first training sample including a first object portrait feature of a first sample virtual object and a first team portrait feature of a first sample game team; the first object portrait feature of the first sample virtual object and the first team portrait feature of the first sample game team are input into a machine learning regression model to obtain a first predicted team probability; a first predicted team result is determined based on the first predicted team probability, and according to the error between the first predicted team result and the first sample team label, the model parameters of the machine learning regression model are adjusted to obtain a first matching model.

[0160] In some embodiments of the present application, the matching model includes a second matching model; the team probability acquisition module 920 is also used to extract the second object portrait features of the target virtual object and the historical team sequence from the object portrait data; extract the second team portrait features of the candidate game team from the team portrait data; input the historical team sequence, the second object portrait features and the second team portrait features into the second matching model to obtain the second team probability.

[0161] In some embodiments of the present application, the historical team formation sequence includes the team capability parameters and team level of the game team that the target virtual object joins each time in the historical team formation.

[0162] In some embodiments of the present application, the team probability acquisition module 920 is also used to obtain a second training sample and a second sample team label corresponding to the second training sample, the second training sample including a second object portrait feature of the second sample virtual object, a historical team sequence of the second sample virtual object, and a second team portrait feature of the second sample game team; the second object portrait feature of the second sample virtual object, the historical team sequence of the second sample virtual object, and the second team portrait feature of the second sample game team are input into the deep learning model to obtain a second predicted team probability; a second predicted team result is determined according to the second predicted team probability, and the model parameters of the deep learning model are adjusted according to the error between the second predicted team result and the second sample team label to obtain a second matching model.

[0163] In some embodiments of the present application, the game team acquisition module 910 is also used to obtain individual ability data, historical team behavior data and virtual props data of the target virtual object; and construct object portrait data of the target virtual object based on the individual ability data, historical team behavior data and virtual props data.

[0164] In some embodiments of the present application, the game team acquisition module 910 is also used to obtain individual ability data and individual team data of virtual objects corresponding to each team member in the candidate game team; generate team ability data of the candidate game team based on the individual ability data of virtual objects corresponding to each team member; generate team formation data of the candidate game team based on the individual team formation data of virtual objects corresponding to each team member and the total team formation time of the candidate game team; construct team portrait data of the candidate game team based on the team ability data and team formation data of the candidate game team.

[0165] In some embodiments of the present application, the game team acquisition module 910 is also used to acquire game teams that have not completed team formation; determine the team level of the game team that has not completed team formation based on the team portrait data of the game team that has not completed team formation; determine the team capability threshold value of the game team that has not completed team formation based on the team level of the game team that has not completed team formation; acquire personal capability parameters of the target virtual object, and screen candidate game teams from the game teams that have not completed team formation according to the team capability threshold value and the personal capability parameters of the target virtual object.

[0166] In some embodiments of the present application, the two-way matching acquisition module 930 is also used to obtain the weights of the first teaming probability and the second teaming probability; the first teaming probability and the second teaming probability are weightedly summed according to the weights to obtain the two-way matching probability between the target virtual object and the game team.

[0167] In some embodiments of the present application, the game team recommendation module 940 is also used to sort the candidate game teams according to the two-way matching probability to obtain a recommended team list; and recommend the game team to the target virtual object based on the recommended team list.

[0168] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0169] In some embodiments of the present application, the game team recommendation device 900 can be implemented in the form of a computer program. The computer program can be used in Fig.10 The computer device shown in the figure is run. The memory of the computer device can store various program modules constituting the game team recommendation device 900, for example, Fig. 9 The game team acquisition module 910, team formation probability acquisition module 920, two-way matching acquisition module 930 and game team recommendation module 940 are shown. The computer program composed of various program modules enables the processor to execute the steps of the game team recommendation method of each embodiment of the present application described in this specification.

[0170] For example, Fig.10The computer device shown can be Fig. 9 The game team acquisition module 910 in the game team recommendation device 900 shown in the figure performs step S210. The computer device can perform step S220 through the team formation probability acquisition module 920. The computer device can perform step S230 through the two-way matching acquisition module 930. The computer device can perform step S240 through the game team recommendation module 940. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external computer device through a network connection. When the computer program is executed by the processor, a game team recommendation method is implemented.

[0171] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0172] In some embodiments of the present application, a server is provided, comprising one or more processors; a memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to execute the steps of the above-mentioned game team recommendation method by the processor. The steps of the game team recommendation method here can be the steps of the game team recommendation method in each of the above-mentioned embodiments.

[0173] In some embodiments of the present application, a computer-readable storage medium is provided, which stores a computer program, and the computer program is loaded by a processor, so that the processor executes the steps of the game team recommendation method described above. The steps of the game team recommendation method here can be the steps of the game team recommendation method in each of the above embodiments.

[0174] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0175] It can be understood that in the above-mentioned specific implementation methods, when the above embodiments of the present application are applied to specific products or technologies, the object portrait data and other related data of the virtual objects operated by the user need to obtain the user's permission or consent, and the collection, use and processing of the relevant data need to comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0176] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0177] The above is a detailed introduction to a game team recommendation method, device, computer equipment and storage medium provided in the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for recommending a game team, characterized in that: include: Obtaining object portrait data of a target virtual object and team portrait data of a candidate game team; According to the object portrait data and the team portrait data, obtaining a first teaming probability that the target virtual object joins the candidate game team through a first matching model, and according to the object portrait data and the team portrait data, obtaining a second teaming probability that the candidate game team agrees to the target virtual object joining through a second matching model; Determining a two-way matching probability between the target virtual object and the candidate game team according to the first team formation probability and the second team formation probability; A game team is recommended to the target virtual object according to the two-way matching probability.

2. The method according to claim 1, characterized in that The step of obtaining a first teaming probability of the target virtual object joining the candidate game team through a first matching model according to the object portrait data and the team portrait data includes: Extracting a first object portrait feature of the target virtual object from the object portrait data; Extracting a first team portrait feature of the candidate gaming team from the team portrait data; The first object portrait feature and the first team portrait feature are input into the first matching model to obtain a first team formation probability.

3. The method according to claim 2, characterized in that Before inputting the first object portrait feature and the first team portrait feature into the first matching model to obtain the first team formation probability, the method includes: Obtaining a first training sample and a first sample team label corresponding to the first training sample, wherein the first training sample includes a first object portrait feature of a first sample virtual object and a first team portrait feature of a first sample game team; Inputting the first object portrait feature of the first sample virtual object and the first team portrait feature of the first sample game team into a machine learning regression model to obtain a first predicted team formation probability; A first predicted teaming result is determined according to the first predicted teaming probability, and a model parameter of the machine learning regression model is adjusted according to an error between the first predicted teaming result and the first sample teaming label to obtain a first matching model.

4. The method according to claim 1, characterized in that: The step of obtaining, based on the object portrait data and the team portrait data, a second team formation probability of the candidate game team agreeing that the target virtual object joins by a second matching model includes: Extracting a second object portrait feature and a historical team sequence of the target virtual object from the object portrait data; Extracting a second team portrait feature of the candidate gaming team from the team portrait data; The historical team formation sequence, the second object portrait feature, and the second team portrait feature are input into the second matching model to obtain a second team formation probability.

5. The method according to claim 4, characterized in that The historical team formation sequence includes team capability parameters and team levels of the game teams that the target virtual object joins each time in the historical team formation.

6. The method according to claim 4, characterized in that Before inputting the historical team formation sequence, the second object portrait feature, and the second team portrait feature into the second matching model to obtain the second team formation probability, the method includes: Obtaining a second training sample and a second sample team label corresponding to the second training sample, wherein the second training sample includes a second object portrait feature of a second sample virtual object, a historical team sequence of the second sample virtual object, and a second team portrait feature of a second sample game team; Inputting the second object portrait feature of the second sample virtual object, the historical team formation sequence of the second sample virtual object, and the second team portrait feature of the second sample game team into the deep learning model to obtain a second predicted team formation probability; A second predicted teaming result is determined according to the second predicted teaming probability, and a model parameter of the deep learning model is adjusted according to an error between the second predicted teaming result and the second sample teaming label to obtain a second matching model.

7. The method according to claim 1, characterized in that The step of obtaining the object portrait data of the target virtual object includes: Obtain individual ability data, historical team behavior data, and virtual props data of the target virtual object; Object portrait data of the target virtual object is constructed according to the individual ability data, the historical team behavior data and the virtual props data.

8. The method according to claim 1, characterized in that The obtaining of team portrait data of the candidate game team includes: Obtain individual capability data and individual team data of virtual objects corresponding to each team member in the candidate game team; Generating team capability data of the candidate game team based on the individual capability data of the virtual objects corresponding to the team members; Generate team formation data of the candidate game team based on the individual team formation data of the virtual objects corresponding to the team members and the total team formation time of the candidate game team; According to the team capability data and team formation data of the candidate game team, the team portrait data of the candidate game team is constructed.

9. The method according to claim 1, characterized in that: Before obtaining the object portrait data of the target virtual object and the team portrait data of the candidate game team, the method further includes: Get the game team that has not been completed; Determining the team level of the incomplete game team according to the team portrait data of the incomplete game team; Determining a team formation capability threshold value of the game team that has not completed team formation based on the team level of the game team that has not completed team formation; The personal ability parameter of the target virtual object is obtained, and a candidate game team is screened from the game teams that have not completed the team formation according to the team formation ability threshold value and the personal ability parameter of the target virtual object.

10. The method according to claim 1, characterized in that The determining, according to the first team formation probability and the second team formation probability, a two-way matching probability between the target virtual object and the candidate game team includes: Obtain the weights of the first teaming probability and the second teaming probability; The first team formation probability and the second team formation probability are weightedly summed according to the weight to obtain a two-way matching probability between the target virtual object and the game team.

11. The method according to claim 1, characterized in that: The recommending a game team to the target virtual object according to the two-way matching probability includes: Sort the candidate game teams according to the two-way matching probability to obtain a recommended team list; A game team is recommended to the target virtual object based on the recommended team list.

12. A game team recommendation device, characterized in that: The device comprises: A game team acquisition module, used to acquire object portrait data of a target virtual object and team portrait data of a candidate game team; A team formation probability acquisition module, configured to acquire a first team formation probability of the candidate game team of the target virtual object through a first matching model according to the object portrait data and the team portrait data, and to acquire a second team formation probability of the candidate game team agreeing to the target virtual object to join through a second matching model according to the object portrait data and the team portrait data; A two-way matching acquisition module, used to determine a two-way matching probability between the target virtual object and the candidate game team according to the first teaming probability and the second teaming probability; A game team recommendation module is used to recommend a game team to the target virtual object according to the two-way matching probability.

13. The device according to claim 12, characterized in that The team formation probability acquisition module is used to: Extracting a first object portrait feature of the target virtual object from the object portrait data; Extracting a first team portrait feature of the candidate gaming team from the team portrait data; The first object portrait feature and the first team portrait feature are input into the first matching model to obtain a first team formation probability.

14. The device according to claim 12, characterized in that The team formation probability acquisition module is used to: Extracting a second object portrait feature and a historical team sequence of the target virtual object from the object portrait data; Extracting a second team portrait feature of the candidate gaming team from the team portrait data; The historical team formation sequence, the second object portrait feature, and the second team portrait feature are input into the second matching model to obtain a second team formation probability.

15. A server, characterized in that: The server comprises: one or more processors; Memory; and One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the game team recommendation method according to any one of claims 1 to 11.

16. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program is loaded by a processor to execute the steps in the game team recommendation method described in any one of claims 1 to 11.

Images