Business processing method and device, computer equipment, storage medium and program product

Abstract

The embodiment of the invention provides a business processing method and device, computer equipment, a storage medium and a program product. The business processing method comprises the following steps: acquiring business characteristics of a first game object, wherein the business characteristics comprise channel characteristics of the first game object in a business channel and game characteristics of the first game object in a game; generating relation characteristics of the first game object based on the relation network of the first game object in the game; performing feature fusion learning on the channel features, the game features and the relation features, and predicting business evaluation information of the first game object in the business channel; and according to the business evaluation information, business processing is performed on the first game object in the business channel, and the business processing is related to the game. By adopting the embodiment of the invention, the business evaluation information of the game object can be accurately determined, and waste of resources related to business processing is avoided.

Description

Technical Field

[0001] The present application relates to the field of computer technology, in particular to the field of game technology, and specifically to a business processing method, a business processing apparatus, a computer device, a computer-readable storage medium, and a computer program product. Background Art

[0002] With the continuous development of computer technology, games have become a popular electronic entertainment item in daily life. Currently, in order to increase the popularity of games, in addition to taking relevant measures within the game (for example, these measures may include in-game rewards, etc.), games are usually also released through business channels outside of the game. In addition, business channels can also perform business processing on game objects outside the game. Business processing may include providing some game-related business services to maintain the game objects.

[0003] Within a business channel, business processing can be performed based on the business evaluation information of a game object within the business channel. Effective business processing for a game object is closely related to accurately determining this business evaluation information. Currently, business evaluation information for game objects is typically provided by operators based on their experience. This information is subjective and can vary significantly between different operators. This means that it is currently difficult to accurately determine business evaluation information, resulting in ineffective business processing for game objects and a waste of resources associated with business processing. Summary of the Invention

[0004] The embodiments of the present application provide a business processing method, apparatus, computer equipment, storage medium, and program product, which can more accurately determine the business evaluation information of a game object and avoid waste of resources involved in business processing.

[0005] On the one hand, an embodiment of the present application provides a service processing method, which includes:

[0006] Obtaining business characteristics of the first game object, the business characteristics including channel characteristics of the first game object within the business channel and game characteristics of the first game object within the game; the channel characteristics are used to characterize the first game object's access to the game within the business channel; the game characteristics are used to characterize the first game object's access to the game; the business channel is used to provide game-related business services to the first game object outside the game;

[0007] generating a relationship feature of the first game object based on a relationship network of the first game object within the game;

[0008] Perform feature fusion learning on channel features, game features, and relationship features to predict business evaluation information of the first game object in the business channel;

[0009] Based on the business evaluation information, business processing is performed on the first game object within the business channel, where the business processing is related to the game.

[0010] Accordingly, an embodiment of the present application provides a service processing device, which includes:

[0011] an acquisition unit, configured to acquire business characteristics of the first game object, the business characteristics including channel characteristics of the first game object within the business channel and game characteristics of the first game object within the game; the channel characteristics are used to characterize the first game object's access to the game within the business channel; the game characteristics are used to characterize the first game object's access to the game; the business channel is used to provide game-related business services to the first game object outside the game;

[0012] a processing unit, configured to generate a relationship feature of the first game object based on a relationship network of the first game object within the game;

[0013] The processing unit is further configured to perform feature fusion learning on the channel features, the game features, and the relationship features to predict business evaluation information of the first game object in the business channel;

[0014] The processing unit is further configured to perform business processing on the first game object within the business channel according to the business evaluation information, where the business processing is related to the game.

[0015] In one implementation, the relationship network includes one or more relationship sub-networks, each relationship sub-network corresponding to a different relationship type; the processing unit is configured to generate a relationship feature of the first game object based on the relationship network of the first game object in the game, specifically performing the following steps:

[0016] Based on each relationship sub-network, generating a relationship feature of the first game object under the corresponding relationship type;

[0017] The relationship features under each relationship type are determined as the relationship features of the first game object.

[0018] In one implementation, a relationship network includes N relationship sub-networks, any one of the N relationship sub-networks is represented as the i-th relationship sub-network, the i-th relationship sub-network corresponds to the i-th relationship type, i and N are both positive integers, and i is less than or equal to N; the i-th relationship sub-network consists of nodes, edges between any two nodes, and edge weights;

[0019] Among them, the i-th relationship sub-network includes at least a first node and a second node; the first node represents the first game object, and the second node represents the second game object of the game; the edge between the first node and the second node indicates that the first game object and the second game object have a relationship under the i-th relationship type; the weight corresponding to the edge between the first node and the second node represents the relationship quantification value between the first game object and the second game object; the relationship quantification value is used to indicate the intimacy of the relationship between the first game object and the second game object under the i-th relationship type.

[0020] In one implementation, the processing unit is configured to generate, based on each relationship sub-network, a relationship feature of the first game object under the corresponding relationship type, specifically to perform the following steps:

[0021] Take each node in the i-th relationship sub-network as the starting point, traverse the i-th relationship sub-network, and obtain the node sequence corresponding to each node;

[0022] Determine the relationship text corresponding to each node sequence; the relationship text corresponding to each node sequence is composed of the player identifier of the game object represented by the node in the node sequence, and the weight corresponding to the edge between each two nodes in the node sequence;

[0023] Call the language processing network to perform semantic understanding on each relationship text, and obtain the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type;

[0024] The relationship feature of each game object in the i-th relationship sub-network under the i-th relationship type is determined as the relationship feature of the first game object under the i-th relationship type.

[0025] In one implementation, the processing unit is configured to call a language processing network to perform semantic understanding on each relational text, and to obtain the relational features of each game object in the i-th relational subnetwork under the i-th relational type, and specifically to perform the following steps:

[0026] Based on each relational text, the language processing network is trained;

[0027] Call the trained language processing network to perform semantic understanding on the player ID of each game object in the i-th relationship sub-network, and obtain the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type.

[0028] In one implementation, the number of relation texts is M, any one of the M relation texts is represented as the j-th relation text, j and M are both positive integers, and j is less than or equal to M; the j-th relation text includes multiple words; and the training process of the language processing network based on the j-th relation text includes:

[0029] Take each word in the j-th relation text as the central word in turn, and determine the marked background word corresponding to the central word;

[0030] Calling the language processing network to perform semantic understanding on the central word to obtain a semantic vector of the central word; and calling the language processing network to perform semantic understanding on each other word in the j-th relation text to obtain a semantic vector of each other word, where each other word is a word in the j-th relation text other than the central word;

[0031] According to the semantic vector of the central word and the semantic vector of each other word, the probability of each other word being the background word corresponding to the central word is determined;

[0032] Based on the difference between the labeled background word and each other word, the loss information of the language processing network is determined according to the probability of each other word being the center word corresponding to the background word;

[0033] The language processing network is trained based on the loss information of the language processing network.

[0034] In one implementation, the channel characteristics include at least one of the following: a channel activity characteristic, a channel interaction characteristic, and a channel consumption characteristic; the channel activity characteristic is used to characterize the activity of the first game object in the business channel and related to the game; the channel interaction characteristic is used to characterize the interaction of the first game object in the business channel and related to the game; the channel consumption characteristic is used to characterize the resource consumption of the first game object in the business channel and related to the game;

[0035] The game features include at least one of the following: game activity features, game interaction features, and game consumption features; the game activity features are used to characterize the activity of the first game object in the game; the game interaction features are used to characterize the interaction of the first game object in the game; and the game consumption features are used to characterize the resource consumption of the first game object in the game.

[0036] In one implementation, the processing unit is configured to perform feature fusion learning on the channel feature, the game feature, and the relationship feature, and to predict the business evaluation information of the first game object in the business channel, specifically to perform the following steps:

[0037] Perform semantic understanding on channel features to obtain the channel semantic vector corresponding to the channel features;

[0038] Perform semantic understanding on game features to obtain the game semantic vector corresponding to the game features;

[0039] Fuse the channel semantic vector, game semantic vector, and relationship features to obtain fused features;

[0040] Feature learning is performed on the fused features, and the feature learning results of the fused features are used as business evaluation information of the first game object.

[0041] In one implementation, feature fusion learning is performed by calling a trained business evaluation model. The training process of the business evaluation model includes:

[0042] Obtaining sample business features of the sample game object, the sample business features including sample channel features of the sample game object in the business channel, game features of the sample game object in the game, and sample relationship features of the sample game object;

[0043] Perform feature fusion learning on sample channel features, sample game features, and sample relationship features to predict sample business evaluation information of sample game objects in the business channel;

[0044] Predicting a business processing method for a sample game object based on sample business evaluation information;

[0045] According to the difference between the predicted business processing mode and the actual business processing mode corresponding to the sample game object, the loss of the sample business evaluation information is calculated to obtain the loss information of the business evaluation model;

[0046] The business evaluation model is trained based on its loss information.

[0047] In one implementation, the processing unit is configured to perform the following steps when performing business processing on the first game object in the business channel according to the business evaluation information:

[0048] Determine a business processing strategy corresponding to the first game object in the business channel; the business channel provides one or more business processing strategies, each business processing strategy corresponds to a respective business processing standard, and the business processing strategy corresponding to the first game object is any one or more business processing strategies provided by the business channel;

[0049] Checking the business evaluation information according to a business processing standard corresponding to the business processing strategy of the first game object;

[0050] If the business evaluation information meets the business processing standard corresponding to the business processing strategy of the first game object, the business processing corresponding to the business processing strategy of the first game object is performed on the first game object in the business channel.

[0051] Accordingly, an embodiment of the present application provides a computer device, comprising:

[0052] a processor suitable for implementing a computer program;

[0053] Computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing the above-mentioned business processing method.

[0054] Accordingly, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is read and executed by a processor of a computer device, the computer device executes the above-mentioned business processing method.

[0055] Accordingly, an embodiment of the present application provides a computer program product, which includes a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the above-mentioned service processing method.

[0056] In an embodiment of the present application, feature fusion learning can be performed on the channel characteristics of the game object in the business channel, the game characteristics of the game object in the game, and the relationship characteristics of the game object in the game to predict the business evaluation information of the game object in the business channel; that is, the channel characteristics of the game object in the business channel, the game characteristics of the game object in the game, and the relationship characteristics of the game object in the game can be comprehensively considered, and the multi-angle characteristics of the game object in the business channel and in the game can be comprehensively considered to more accurately determine the business evaluation information of the game object. Therefore, based on the more accurate business evaluation information, effective business processing can be performed on the business object to avoid waste of resources involved in business processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0058] Figure 1 This is a schematic diagram of the architecture of a business processing system provided in an embodiment of the present application;

[0059] Figure 2 This is a schematic diagram of an application scenario of a business processing method provided in an embodiment of the present application;

[0060] Figure 3 This is a schematic diagram of an application scenario of another business processing method provided in an embodiment of the present application;

[0061] Figure 4This is a schematic diagram of an application scenario of another business processing method provided in an embodiment of the present application;

[0062] Figure 5 This is a flowchart of a business processing method provided by an embodiment of the present application;

[0063] Figure 6 is a schematic diagram of a relationship network provided in an embodiment of the present application;

[0064] Figure 7 is a schematic diagram of a relationship sub-network provided in an embodiment of the present application;

[0065] Figure 8 This is a schematic diagram of the model structure of a business evaluation model provided in an embodiment of the present application;

[0066] Figure 9 is a schematic diagram of a neural network perception layer provided in an embodiment of the present application;

[0067] Figure 10 This is a schematic diagram of the execution logic of a business evaluation model provided in an embodiment of the present application;

[0068] Figure 11 This is a flowchart of another business processing method provided by an embodiment of the present application;

[0069] Figure 12 This is a schematic diagram of the generation principle of a relationship feature provided in an embodiment of the present application;

[0070] Figure 13 This is a schematic diagram of a node sequence traversal result provided by an embodiment of the present application;

[0071] Figure 14 This is a schematic diagram of the structure of a language processing network provided in an embodiment of the present application;

[0072] Figure 15 This is a schematic diagram of the structure of a service processing device provided in an embodiment of the present application;

[0073] Figure 16 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0075] In order to more clearly understand the technical solutions provided by the embodiments of the present application, the technical terms involved in the embodiments of the present application are first introduced here.

[0076] (1) Games:

[0077] The game mentioned in the embodiments of this application refers to an electronic game product. An electronic game refers to an interactive game that runs on an electronic device platform. Electronic games can include stand-alone games and online games. Among them, a stand-alone game refers to an electronic game that can be run independently using a computer or another game platform. A stand-alone game does not require a game server to run normally. Online games can also be called online games. Generally, they refer to sustainable individual multi-player online games that use the Internet as a communication medium, a game server (Game Server) and a player terminal as game processing devices, and a game client (Game Client) running in the player terminal as an information interaction window, aiming to achieve entertainment, leisure, communication and virtual achievements. The game client refers to a program that corresponds to the game server and provides local game services for game objects participating in the online game (game objects refer to objects that play games); the game server refers to a service device that provides data calculation, verification, storage and forwarding functions for the game client. For online games, the embodiments of this application do not limit the game type of online games; for example, online games can be cloud games, or online games can be ordinary online games.

[0078] Among them, cloud gaming (Cloud Gaming), which can also be called gaming on demand (Gaming on Demand), refers to games that are deployed in the cloud and run on game servers (the game servers here specifically refer to cloud servers); that is, in the cloud gaming scenario, all cloud games are not run in the cloud gaming client of the player's terminal (the cloud gaming client refers to the game client used by the game object to play the cloud game), but in the game server. The game server compresses and encodes the game screen and game audio in the cloud game into a media stream, and then transmits the media stream to the cloud gaming client in the player's terminal through the network for image display and audio playback; specifically, the game screen can be encoded into a video stream, and the game audio can be encoded into an audio stream. The video stream and audio stream are transmitted to the cloud gaming client in the form of media streams, and the cloud gaming client presents the game screen and plays the game audio.

[0079] As can be seen, cloud gaming clients don't need powerful graphics processing and data computing capabilities; they only need basic media streaming capabilities and the ability to receive and transmit operational commands from game objects to the game server. When game objects perform various operations in the cloud gaming client's game interface, the cloud gaming client reports the resulting operational data to the game server. The game server then refreshes the cloud game's screen based on the operational data in the corresponding cloud game, and then returns the refreshed screen to the cloud gaming client for display, enabling cloud gaming.

[0080] The cloud gaming described in the above content is a cloud gaming that transmits the game screen in the form of media streaming. In addition, the technical implementation of cloud gaming can also include the instruction stream method. The instruction stream means that the game server transmits the rendering instructions of the cloud game to the cloud gaming client. The cloud gaming client uses the graphics card of the player's terminal to render and then displays the game screen.

[0081] Regular online games are games that run directly on a game client installed on the player's terminal. The difference between regular online games and cloud games is that regular online games run on a game client installed on the player's terminal, while cloud games run on a game server. The cloud game client is responsible for displaying the game screen, playing the game audio, and receiving input from game objects.

[0082] (2) Business channels:

[0083] Business channels refer to business processing of game objects outside the game to maintain the business products of the game objects. Business processing can be, for example, the provision of business services related to the game. Business channels can support the provision of game-related business services for one or more games, that is, business channels can provide game-related business services for game objects of one or more games. The embodiments of the present application do not limit the product form of business channels. For example, business channels can be applications, applets, software, web pages or public platforms, etc. Moreover, the outside of the game and the inside of the game mentioned in the embodiments of the present application are two relative concepts. Inside the game refers to the business scope covered by the game product itself, for example, within the business scope covered by the game client; and outside the game refers to outside the business scope covered by the game product itself, for example, outside the business scope covered by the game client.

[0084] Business services related to games refer to services whose service objects in business channels are game objects and whose service contents are game-related contents. The embodiments of the present application do not limit the types of business services. For example, business services related to games may include any one or more of the following: interactive services related to games, resource services related to games, publicity services related to games, reward and feedback services related to games, and churn recovery services related to games. Among them, interactive services related to games refer to services whose service objects in business channels are game objects and whose service contents are game interactions, for example, providing game team formation functions to game objects in business channels; resource services related to games refer to services whose service objects in business channels are game objects and whose service contents are game resources, for example, providing game resource recharge, game resource usage and other functions to game objects in business channels; publicity services related to games refer to services whose service objects in business channels are game objects and whose service contents are game promotional contents (for example, game promotional contents may be game advertisements), for example, delivering game advertisements to game objects in business channels; reward and feedback services related to games A game-related churn recovery service refers to a service in which the service object in a business channel is a game object and the service content is game rewards (for example, game rewards can be game resources, game props, etc.), for example, issuing game feedback packages to game objects in the business channel; a game-related churn recovery service refers to a service in which the service object in a business channel is a churned game object (the churned game object here refers to a game object that has not visited the game within a certain time period (for example, a certain time period can be three days, a week or a month, etc.)), and the service content is game recall information, for example, pushing game recall information to the churned game object in the business channel, notifying the churned game object that it can obtain game return rewards after returning to the game.

[0085] Optionally, a range of game objects maintained by the business channel can be set. Specifically, the game objects maintained by the business channel can be all game objects in the game. Alternatively, the game objects maintained by the business channel can be a portion of the game objects in the game, which can include any one or more of the following: game objects with a game level above a set level threshold, game objects with a game experience value above a set experience threshold, game objects with a resource consumption value above a set resource threshold, etc.

[0086] (3) Business evaluation information:

[0087] The business evaluation information of the game object in the business channel is used to indicate the impact of the game object on the business channel. The greater the impact of the game object on the business channel, the greater the value of its business evaluation information; the smaller the impact of the game object on the business channel, the smaller the value of its business evaluation information.

[0088] Based on the above introduction to technical terms such as games, business channels, and business evaluation information, an embodiment of the present application provides a business processing method. This business processing method, for any game that provides business services supported by a business channel, can comprehensively consider the channel characteristics of the game object in the business channel that are related to the game (for example, the channel characteristics may include channel activity characteristics, channel interaction characteristics, and channel consumption characteristics), the game characteristics of the game object in the game (for example, the game characteristics may include game activity characteristics, game interaction characteristics, and game consumption characteristics), and the relationship characteristics of the game object in the game to determine the business evaluation information of the game object in the business channel; that is, the business processing method can determine the business evaluation information of the game object in the business channel from a more comprehensive perspective, so that the business evaluation information of the game object in the business channel can be determined more accurately, and thus, based on the more accurate business evaluation information, effective business processing can be performed on the business object to avoid waste of resources involved in business processing.

[0089] In a specific implementation, the business processing method provided in the embodiment of the present application can be executed by a computer device. The computer device can be a business processing system composed of a terminal and a server. Figure 1 As shown, a business channel may be running in the terminal 101, and the server 102 may be a backend server of the business channel. The embodiment of the present application does not limit the connection method between the terminal 101 and the server 102. The terminal 101 and the server 102 may establish a direct communication connection through wired communication, or an indirect communication connection may be established between the terminal 101 and the server 102 through wireless communication.

[0090] In the business processing system composed of the terminal 101 and the server 102, the server 102 can obtain the channel characteristics of the game object related to the game in the business channel, the game characteristics of the game object in the game, and the relationship characteristics of the game object in the game, and determine the business evaluation information of the game object in the business channel based on these characteristics, and can generate business processing instructions for the game object in the business channel based on the business evaluation information of the game object in the business channel. The business processing instructions can be sent to the terminal 101; the terminal 101 can execute the business processing instructions and perform business processing on the game object in the business channel.

[0091] The terminal 101 may include, but is not limited to, any of the following: a smartphone, a tablet computer, a laptop computer, a desktop computer, a smartwatch, a smart home appliance, a smart car terminal, and an aircraft. The server 102 may be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0092] Alternatively, the computer device may be a separate device, for example, a terminal of a game object. In this case, the terminal of each game object may determine business evaluation information and perform business processing for each game object. Specifically, the terminal of the game object may obtain channel characteristics related to the game within the business channel, game characteristics of the game object within the game, and relationship characteristics of the game object within the game, and determine business evaluation information of the game object within the business channel based on these characteristics. Furthermore, business processing may be performed on the game object within the business channel based on the business evaluation information of the game object within the business channel.

[0093] The above describes the general process and execution entities of the business processing method. The following describes its application scenarios. The business processing method can be applied to any scenario requiring business evaluation information, such as channel game promotion, channel game reward feedback, and channel churn game object recall. The following describes how business evaluation information is used in these scenarios.

[0094] (1) Channel game promotion scenarios:

[0095] The channel game promotion scenario refers to a scenario in which a game object is promoted within a business channel based on the business evaluation information of the game object within the business channel. The application of business evaluation information in the channel game promotion scenario may specifically include: after obtaining the business evaluation information of the game object within the business channel, if the business evaluation information of the game object is higher than the benchmark evaluation information (the benchmark evaluation information can be set based on experience), or if the business evaluation information of the game object is arranged at the benchmark position (the benchmark position can be set based on experience) or the benchmark position, then game promotion content may be delivered to the game object within the business channel. Figure 2The product interface of the business channel shown can be used to place daily game advertisements 201 for game objects within the business channel, for example, one game advertisement 201 can be placed every day. Based on the business processing method provided in the embodiment of the present application, the business evaluation information of the game object within the business channel can be determined more accurately, so that the game object can be effectively promoted in the channel game promotion scene, avoiding the waste of game promotion resources (for example, financial resources for promotion investment, multimedia resources for promotional copy, network resources for promotion, etc.).

[0096] (2) Channel game reward feedback scenario:

[0097] The channel game reward and feedback scenario refers to a scenario where game rewards are awarded to game subjects within a business channel based on their business evaluation information within the business channel. The application of business evaluation information in the channel game reward and feedback scenario is similar to its application in the channel game promotion scenario. This may include: after obtaining the game subject's business evaluation information within the business channel, if the game subject's business evaluation information is higher than the benchmark evaluation information, or if the game subject's business evaluation information is ranked at or before the benchmark position, then the game subject may be awarded a game reward within the business channel.

[0098] Furthermore, game reward packages can be distributed to game objects in the business channel by means of some special time nodes (for example, special time nodes can include holidays, anniversaries of game objects registering games, anniversaries of game objects registering business channels, and birthdays of game objects, etc.); Figure 3 The product interface of the business channel shown can distribute a birthday reward package 301 to the game object within the business channel on the game object's birthday. Based on the business processing method provided by the embodiment of the present application, the business evaluation information of the game object within the business channel can be determined more accurately. Therefore, in the channel game reward feedback scenario, game rewards can be effectively provided to the game object, avoiding the waste of game reward resources.

[0099] (3) Channel Loss Game Object Recall Scenario:

[0100] The channel churn recall scenario involves recalling churned game objects within a business channel based on their business evaluation information within the business channel. The application of this business evaluation information in the channel churn recall scenario may include: after obtaining the business evaluation information of a churned game object within a business channel, if the churned game object's business evaluation information is higher than the benchmark evaluation information, or if the game object's business evaluation information is ranked at or before the benchmark position, the game object may be recalled within the business channel.

[0101] Furthermore, game recall information can be pushed to game objects within the business channel with the help of some special time nodes. Figure 4 The product interface of the business channel shown in the figure can push game recall information 401 to the game object in the business channel during the holiday period. The game recall information 401 can be used to notify the lost game object that it can obtain the game return reward after returning to the game. Based on the business processing method provided in the embodiment of the present application, the business evaluation information of the lost game object in the business channel can be determined more accurately, so that the game object can be effectively recalled in the channel lost game object recall scenario, avoiding the waste of recall resources (for example, the financial resources invested in the recall, the multimedia resources of the recall copy, and the network resources consumed by the recall).

[0102] It is worth noting that the collection and processing of relevant data in the embodiments of the present application should be strictly in accordance with the requirements of relevant laws and regulations. The acquisition of personal information requires the knowledge or consent of the individual subject (or the legal basis for obtaining the information), and the subsequent use and processing of data shall be carried out within the scope of authorization of laws and regulations and the subject of personal information. For example, when the embodiments of the present application are applied to specific products or technologies, for example, when obtaining the business characteristics of a game object and obtaining the relationship network of a game object, it is necessary to obtain the permission or consent of the game object, and the collection, use and processing of relevant data (for example, feature fusion learning of business characteristics and relationship characteristics, etc.) need to comply with the relevant laws, regulations and standards of the relevant region.

[0103] The above content introduces the general process of the business processing method, the execution subject of the business processing method, and the application scenarios of the business processing method. The following is a detailed introduction to the technical process of the business processing method with reference to the accompanying drawings.

[0104] The embodiment of the present application provides a business processing method, which includes the content of channel features, game features and relationship features, as well as the process of feature fusion learning of channel features, game features and relationship features. The business processing method can be executed by a computer device, such as Figure 5 As shown, the business processing method may include but is not limited to the following steps S501 to S504:

[0105] S501: Acquire business characteristics of a first game object, where the business characteristics include channel characteristics of the first game object in a business channel and game characteristics of the first game object in a game.

[0106] The business channel can be used to provide game-related business services to the first game object outside the game; the business channel can support the provision of game-related business services for one or more games, and the game mentioned here is any game that the business channel supports to provide game-related business services.

[0107] The business characteristics of the first game object may include channel characteristics of the first game object within the business channel and game characteristics of the first game object within the game. The channel characteristics of the first game object within the business channel can be used to characterize the first game object's access to the game within the business channel; the game characteristics of the first game object within the game can be used to characterize the first game object's access to the game.

[0108] Specifically, the channel characteristics of the first game object in the business channel may include at least one of the following: channel activity characteristics, channel interaction characteristics, and channel consumption characteristics.

[0109] The channel activity feature can be used to characterize the activity of the first game object in the business channel, which is related to the game. For example, the channel activity feature may include the number of channel active days, the number of channel activity participations, and the number of channel activity interactions of the first game object in each of the most recent P (P is a positive integer) statistical time periods, etc. The statistical time period can be, for example, one day or one week. The number of channel active days refers to the number of active days (for example, login days) of the first game object in the business channel related to the game; the number of channel activity participations refers to the number of game-related activity participations of the first game object in the business channel; the number of channel activity interactions refers to the number of game-related activity interactions of the first game object in the business channel, and activity interactions can include, for example, activity browsing and activity participation.

[0110] Channel interaction features can be used to characterize a first player's interactions with the game within a business channel. For example, channel interaction features may include the number of channel-generated team formations, the number of completed team formations, and the number of successful gift pack redemptions for a team formation, etc., for the first game object within each of the last P statistical time periods. The number of channel-generated team formations refers to the number of times the first game object has formed a game-related team within the business channel; the number of completed team formations refers to the number of times the first game object has successfully formed a game-related team within the business channel; and the number of successful gift pack redemptions refers to the number of times the first game object has received a gift pack after successfully forming a game-related team within the business channel.

[0111] The channel consumption feature can be used to characterize the game-related resource consumption of the first game object within the business channel. For example, the channel consumption feature may include the channel consumption of the first game object in each statistical time period of the most recent P statistical time periods, the channel consumption item ranking, and the channel penetration parameter, etc. Channel consumption refers to the consumption of game resources related to the game by the first game object within the business channel; the channel consumption item ranking refers to the ranking of the consumption of game props related to the game by the first game object within the business channel; the channel penetration parameter refers to the proportion of the game resource consumption of the first game object within the business channel to the overall game resource consumption of the first game object. Channel penetration parameter = (game resource consumption of the first game object within the business channel / overall game resource consumption) * 100%.

[0112] That is to say, when obtaining channel characteristics, channel characteristics from multiple different angles can be obtained. Channel characteristics from different angles can characterize the first game object's access to the game within the business channel from different angles. This is conducive to improving the prediction accuracy of business evaluation information when predicting the business evaluation information of the first game object.

[0113] The game characteristics of the first game object in the game may include at least one of the following: game activity characteristics, game interaction characteristics, and game consumption characteristics. Among them:

[0114] The game activity feature can be used to characterize the activity of the first game object within the game. For example, the game activity feature may include the online game duration of the first game object in each of the last P statistical time periods. Taking a statistical time period of one week as an example, the game activity feature may include the online game duration of the first game object in the first week, the online game duration of the second week, the online game duration of the third week, and the online game duration of the fourth week within the last four weeks.

[0115] Game interaction features can be used to characterize the in-game interactions of the first game object. For example, the game interaction features may include the number of in-game friends of the first game object and the number of friend interactions of the first game object in each of the last P statistical time periods. For example, taking a statistical time period of one week, the game interaction features may include the number of in-game friends of the first game object and the number of friend interactions of the first game object in the first week, the number of friend interactions of the second week, the number of friend interactions of the third week, and the number of friend interactions of the fourth week in the last four weeks.

[0116] The game consumption feature is used to characterize the resource consumption of the first game object within the game. For example, the game consumption feature may include the game consumption of the first game object within each of the last P statistical time periods. Game consumption refers to the amount of game resources consumed by the first game object within the game. For example, if the statistical time period is one week, the game consumption feature may include the first game object's game consumption within the last four weeks, the second week's game consumption, the third week's game consumption, and the fourth week's game consumption.

[0117] That is to say, when obtaining game features, game features from multiple different angles can be obtained. Game features from different angles can characterize the access situation of the first game object in the game from different angles. This is conducive to improving the prediction accuracy of business evaluation information when predicting the business evaluation information of the first game object.

[0118] S502: Generate a relationship feature of the first game object based on the relationship network of the first game object in the game.

[0119] The relationship network of the first game object within the game can reflect the relationship between the first game object and other game objects within the game. The relationship network can include one or more relationship sub-networks, each of which can correspond to a different relationship type. For any relationship sub-network, the relationship sub-network can be a graph network consisting of nodes, edges between each node, and weights corresponding to the edges. Nodes can represent game objects; edges between each node can represent the relationship between each game object; weights corresponding to edges can represent the relationship quantification value between each player. The relationship quantification value can be used to indicate the degree of intimacy between the two game objects. The higher the intimacy, the greater the relationship quantification value and the greater the weight. Conversely, the lower the intimacy, the smaller the relationship quantification value and the smaller the weight. Figure 6 The relationship network of the first game object is shown. It can be seen that the relationship network of the first game object may include three relationship sub-networks, namely relationship sub-network 601, relationship sub-network 602 and relationship sub-network 603. The three relationship sub-networks correspond to three different relationship types.

[0120] Furthermore, in order to more clearly represent the relationship network, the number of relationship sub-networks included in the relationship network can be expressed as N, that is, the relationship network includes N relationship sub-networks; any relationship sub-network in the N relationship sub-networks is expressed as the i-th relationship sub-network, and the relationship type corresponding to the i-th relationship sub-network is expressed as the i-th relationship type, that is, the i-th relationship sub-network corresponds to the i-th relationship type; wherein i and N are both positive integers, and i is less than or equal to N.

[0121] For the i-th relationship sub-network, such as Figure 7As shown, the i-th relationship subnetwork consists of nodes, edges between each node, and edge weights. The i-th relationship subnetwork may include at least a first node and a second node, where the first node may represent a first game object and the second node may represent a second game object in the game. The edge between the first node and the second node may represent a relationship between the first game object and the second game object under the i-th relationship type. The weight corresponding to the edge between the first node and the second node may represent a quantitative value of the relationship between the first game object and the second game object; the quantitative value of the relationship may be used to indicate the closeness of the relationship between the first game object and the second game object under the i-th relationship type.

[0122] For example, a relationship network can include four relationship sub-networks: an alliance relationship network, a gift relationship network, a friend relationship network, and a guardian relationship network. The relationship type corresponding to the alliance relationship network is an alliance relationship, the relationship type corresponding to the gift relationship network is a gift relationship, the relationship type corresponding to the friend relationship network is a friend relationship, and the relationship type corresponding to the guardian relationship network is a guardian relationship. The alliance relationship network is a relationship sub-network composed of in-game objects that have an alliance relationship. Game objects that have formed an alliance relationship within a certain period of time can be selected to construct the alliance relationship network. In the alliance relationship network, nodes represent game objects, edges between two nodes represent the existence of an alliance relationship between two game objects, and the weights corresponding to the edges represent the experience value of the alliance relationship between two game objects. The gift relationship network is a relationship sub-network composed of in-game objects that have gifted each other. Game objects that have formed a gift relationship within a certain period of time can be selected to construct the gift relationship network. In the gift relationship network, nodes represent game objects, edges between two nodes represent the existence of a gift relationship between two game objects, and the weights corresponding to the edges represent the number of gift exchanges between two game objects. The Friendship Network is a sub-network of in-game objects that have friendships. Game objects that have established friendships within a certain period of time can be selected to form a Friendship Network. Nodes represent game objects, and edges between any two nodes represent friendships between any two game objects. The Guardianship Network is a sub-network of in-game objects that have guardianships. Game objects that have jointly established and enhanced guardianships within a certain period of time can be selected to form a Guardianship Network. Nodes represent game objects, and edges between any two nodes represent actions that enhance guardianship between any two game objects.

[0123] As described above, the relationship network of the first game object may include relationship sub-networks of different relationship types. In the process of generating the relationship features of the first game object based on the relationship network of the first game object in the game, the relationship features under the relationship type corresponding to each relationship sub-network may be generated respectively. Specifically, the relationship features of the first game object under the corresponding relationship type may be generated based on each relationship sub-network; the relationship features under each relationship type may be determined as the relationship features of the first game object. It can be seen that the relationship features of the first game object can be used to characterize the relationship characteristics between the first game object and other game objects. Through the relationship networks corresponding to a variety of different relationship types, the relationship features of the first game object under different relationship types may be generated. The relationship features under different relationship types may characterize the relationship establishment preferences of the first game object in the game from different perspectives. This is conducive to improving the prediction accuracy of the business evaluation information when predicting the business evaluation information of the first game object.

[0124] S503: Perform feature fusion learning on the channel features, game features, and relationship features to predict business evaluation information of the first game object in the business channel.

[0125] In an embodiment of the present application, by performing feature fusion learning on channel features, game features, and relationship features, the business evaluation information of the first game object in the business channel can be predicted.

[0126] Feature fusion learning refers to the process of fusing channel features, game features, and relationship features, and then performing feature learning on the fused features. Specifically, this process may include: semantically understanding the channel features to obtain the corresponding channel semantic vectors; semantically understanding the game features to obtain the corresponding game semantic vectors; fusing the channel semantic vectors, game semantic vectors, and relationship features to obtain the fused features; and performing feature learning on the fused features, using the feature learning results of the fused features as the business evaluation information for the first game object.

[0127] In the feature fusion learning process described above, the channel semantic vector is a vector that represents the semantic meaning of channel features. Semantic understanding of channel features involves converting them into multidimensional space vectors. In other words, the channel semantic vector is a multidimensional space vector. There can be multiple channel features, each corresponding to a dimension of the channel semantic vector. For example, if there are 100 channel features, they can be mapped to a 100-dimensional vector. When a channel feature has a eigenvalue, the corresponding dimension in the vector is set to 1; otherwise, it is set to 0.

[0128] Similarly, a game semantic vector is a vector that represents the semantic meaning of game features. Semantic understanding of game features involves converting them into multidimensional space vectors. In other words, a game semantic vector is a multidimensional space vector. There can be multiple game features, each corresponding to a dimension of the game semantic vector. For example, if there are 100 game features, they can be mapped to a 100-dimensional vector. When a game feature has a value, the corresponding dimension in the vector is set to 1; otherwise, it is set to 0.

[0129] In the above feature learning process, relationship features can include relationship features under the relationship types corresponding to each relationship sub-network. The relationship features under the relationship types corresponding to each relationship sub-network are also essentially vectors. Therefore, the channel semantic vector, game semantic vector, and relationship features are fused. Specifically, the channel semantic vector, game semantic vector, and relationship features are vector-summed to obtain the fused feature. The fused feature is also essentially a vector. Feature learning of the fused feature involves performing nonlinear calculations on the fused feature.

[0130] The contents of the above steps S501 to S503 can be executed by calling a trained business evaluation model. The structure of the business evaluation model is introduced below, and the combination of the business evaluation model and steps S501 to S503 is introduced in combination with the structure of the business evaluation model.

[0131] Figure 8 The model structure of the business evaluation model is shown. The business evaluation model may include a first vector conversion layer (Embedding lookup), a second vector conversion layer (Embedding lookup), and an evaluation prediction layer (Baseline prediction layer); the evaluation prediction layer may include a vector summation layer (Sum pooling) and a neural network perception layer (MLP). Figure 9 As shown, the perception layer of the neural network is composed of several inputs (e.g. Figure 9 An artificial neural network structure consists of input 1, input 2, and input 3) and an output; an artificial neural network is an algorithmic mathematical model that imitates the behavioral characteristics of animal neural networks and performs distributed parallel information processing. This network relies on the complexity of the system and adjusts the interconnected relationships between a large number of internal nodes to achieve the purpose of processing information.

[0132] The combination of the business evaluation model and steps S501 to S503 is as follows: Figure 10 As shown:

[0133] ① The first vector conversion layer can be used to execute the feature engineering phase. The execution logic of the feature engineering phase may include: obtaining channel features, performing semantic understanding on the channel features, and obtaining the corresponding channel semantic vectors; and obtaining game features, performing semantic understanding on the game features, and obtaining the corresponding channel semantic vectors.

[0134] ② The second vector conversion layer can be used to execute the relationship feature generation phase. The execution logic of the relationship feature generation phase may include: generating relationship features based on the relationship network. Generating relationship features based on the relationship network can be performed by the second vector conversion layer calling the language processing network in the natural language processing model. The language processing network can be implemented using the DeepWalk algorithm; the DeepWalk algorithm is a graph structure data mining algorithm that combines random walks with natural language processing models. It aims to learn the hidden information of the graph network and express the nodes in the graph network.

[0135] ③ The evaluation prediction layer can be used to execute the evaluation prediction stage. The execution logic of the evaluation prediction stage may include: calling the vector summation layer to perform vector summation on the channel semantic vector, game semantic vector, and relationship features to obtain fusion features; calling the neural network perception layer to perform feature learning (i.e., nonlinear calculation) on the fusion features to obtain the feature learning results of the fusion features, and using the feature learning results of the fusion features as the business evaluation information of the first game object. Among them, the neural network perception layer can increase the nonlinearity of the business processing model. The nonlinear calculation of the neural network perception layer is shown in the following formula:

[0136] y=relu(relu(XW1+bias1)W2+bias2)

[0137] In the above formula, X represents the fused feature; Relu (Linear Rectification Function) is an activation function. W1, W2, bias1, and bias2 are trainable model parameters in the business evaluation model.

[0138] The above content introduces the business evaluation model and the execution logic of the business evaluation model. The following introduces the training process of the business evaluation model. The training process of the business evaluation model may include: obtaining sample business features of the sample game object, the sample business features include the sample channel features of the sample game object in the business channel, the game features of the sample game object in the game, and the sample relationship features of the sample game object; performing feature fusion learning on the sample channel features, sample game features, and sample relationship features to predict the sample business evaluation information of the sample game object in the business channel; predicting the business processing method for the sample game object based on the sample business evaluation information (for example, the predicted business processing method is to deliver promotional content to the sample game object); calculating the loss of the sample business evaluation information based on the difference between the predicted business processing method and the actual business processing method corresponding to the sample game object (for example, the actual business processing method is not delivering promotional content to the sample player), and obtaining the loss information of the business evaluation model; training the business evaluation model based on the loss information of the business evaluation model.

[0139] During one training process of the above-mentioned business evaluation model, the sample relationship features of the sample game objects are generated based on the relationship network of the sample game objects in the game. The generation process of the relationship features of the sample game objects in the game is similar to the generation process of the relationship features of the first game object in the game. For details, please refer to the generation process of the relationship features of the first game object in the game, which will not be repeated here. The feature fusion learning process of the sample channel features, sample game features and sample relationship features is similar to the feature fusion learning process of the channel features, game features and relationship features. For details, please refer to the feature fusion learning process of the channel features, game features and relationship features, which will not be repeated here. The loss calculation of the sample business evaluation information is based on the loss function calculated on the sample business evaluation information.

[0140] In a specific implementation, the business evaluation model can be trained multiple times according to the single training process of the business evaluation model described above. When the training termination condition is reached, a trained business evaluation model is obtained. The training termination condition may include any of the following situations: the loss information of the business evaluation model is less than or equal to a first loss threshold (the first loss threshold can be set based on experience), and the number of training times of the business evaluation model reaches a first training number threshold (the first training number threshold can be set based on experience).

[0141] S504: Perform business processing on the first game object within the business channel according to the business evaluation information, where the business processing is related to the game.

[0142] After obtaining the business evaluation information through feature fusion learning, game-related business processing can be performed on the first game object within the business channel based on the business evaluation information. For example, game promotional content can be delivered to the first game object within the business channel, game rewards can be distributed to the first game object within the business channel, and game recall information can be pushed to the first game object within the business channel.

[0143] In the embodiment of the present application, the channel characteristics of the game object in the business channel, the game characteristics of the game object in the game, and the relationship characteristics of the game object in the game can be comprehensively considered. By comprehensively considering the multi-angle characteristics of the game object in the business channel and in the game, the business evaluation information of the game object in the business channel can be determined more accurately. In addition, when obtaining the channel characteristics and game characteristics, multiple characteristics from different angles can be obtained. The channel characteristics from different angles can characterize the access of the first game object to the game in the business channel from different angles. The game characteristics from different angles can characterize the access of the first game object in the game from different angles. This is conducive to improving the accuracy of the business evaluation information when predicting the business evaluation information of the first game object. For the relationship network, relationship characteristics under different relationship types can be generated based on the relationship network under multiple different relationship types. The relationship characteristics under different relationship types can characterize the relationship establishment preferences of the first game object in the game from different angles. This is conducive to further improving the prediction accuracy of the business evaluation information when predicting the business evaluation information of the first game object.

[0144] The embodiment of the present application provides a business processing method, which includes the generation process of the dry relationship feature and the business processing method of the first game object. The business processing method can be executed by a computer device, such as Figure 11 As shown, the business processing method may include but is not limited to the following steps S1101 to S1105:

[0145] S1101: Acquire business characteristics of a first game object, where the business characteristics include channel characteristics of the first game object in a business channel and game characteristics of the first game object in a game.

[0146] In the embodiment of the present application, the execution process of step S1101 is the same as the above Figure 5 The execution process of step S501 in the embodiment shown is the same, and the details can be found in the above Figure 5 The execution process of step S501 in the illustrated embodiment will not be described in detail here.

[0147] S1102: Based on each relationship sub-network, generate a relationship feature of the first game object under the corresponding relationship type.

[0148] As described above, the relational sub-network can be processed using the DeepWalk algorithm. The DeepWalk algorithm generates relational features based on the relational sub-network. Figure 12 As shown, the goal of analyzing the relational subnetwork is to obtain a vectorized representation of each node in the relational subnetwork. The relational features of the first game object under the corresponding relation type in the relational subnetwork are composed of the vectorized representations of each node in the relational subnetwork. To achieve this goal, the proposed approach draws on the language processing network in natural language processing models; for example, the language processing network can be word2vec (word to vector, a tool that converts words into vectors). However, the training samples for the language processing network are text sentences. Therefore, it is necessary to generate training samples similar to text corpus based on the network topology of the relational subnetwork. Training samples similar to text corpus can be generated by random walks in the relational subnetwork. Specifically, a random walk treats each node in the relational subnetwork as a word and uses a traversal method (for example, depth-first traversal or breadth-first traversal) to obtain a node sequence of a specified length, which is similar to a sentence of text corpus. Based on this, the language processing network can be trained using text corpus, and the trained language processing model can be used to convert each node in the relational subnetwork into its corresponding vectorized representation.

[0149] As described above, a relationship network can include N relationship subnetworks. Any of these N relationship subnetworks can be represented as the i-th relationship subnetwork, where the i-th relationship subnetwork corresponds to the i-th relationship type. The following describes the processing of any of these N relationship subnetworks (i.e., the i-th relationship subnetwork) as an example.

[0150] The process of generating the relationship feature of the first game object under the i-th relationship type based on the i-th relationship sub-network may include:

[0151] ① Take each node in the i-th relationship sub-network as the starting point, traverse the i-th relationship sub-network, and obtain the node sequence corresponding to each node. In the specific implementation, the length of the traversed node sequence can be specified (the node sequence length refers to the number of nodes contained in the node sequence). For each starting point, after reaching the specified node sequence length, the traversal can be stopped to obtain the node sequence of the specified length. Figure 13 As shown, with node A as the starting point and the length of the node sequence specified as 4, the node sequences "ABCF", "ABCE", and "ABEF", etc. can be obtained.

[0152] ② Determine the relationship text corresponding to each node sequence; the relationship text corresponding to each node sequence is composed of the player identifier of the game object represented by the node in the node sequence, and the weight corresponding to the edge between each two nodes in the node sequence.

[0153] For example, in a gift-giving relationship network, the weight of the edge between any two nodes is the number of gift exchanges between any two game objects. Traversing the gift-giving relationship network yields a node sequence of "ABCF," where node A represents the first game object, node B represents the second game object, node C represents the third game object, and node F represents the sixth game object. The weight of the edge between node A and node B is 3, indicating that the number of gift exchanges between the first and second game objects is 3. The weight of the edge between node B and node C is 4, indicating that the number of gift exchanges between the second and third game objects is 4. The weight of the edge between node C and node F is 2, indicating that the number of gift exchanges between the third and sixth game objects is 2. The relationship text obtained by converting the node sequence "ABCF" is "First Game Object 3 Second Game Object 4 Third Game Object 2 Sixth Game Object."

[0154] For example, in a friend relationship network, the weight of each edge between any two nodes is 1. Traversing the friend relationship network yields a node sequence of "ABCD," where node A represents the first game object, node B represents the second game object, node C represents the third game object, and node D represents the fourth game object. The resulting relationship text from the node sequence "ABCD" is "first game object 1 second game object 1 third game object 1 fourth game object."

[0155] ③ Call the language processing network to perform semantic understanding on each relationship text, and obtain the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type.

[0156] In a specific implementation, the language processing network must first be trained on the text sets consisting of each relational text, so that the language processing network has the ability to semantically understand the text sets consisting of each relational text. Then, the trained language processing network can be called to convert the player ID of each game object into a corresponding semantic vector to obtain the relationship features of each game object. In other words, the language processing network can be trained based on each relational text; the trained language processing network can be called to semantically understand the player ID of each game object in the i-th relational sub-network, and the relationship features of each game object in the i-th relational sub-network under the i-th relationship type can be obtained.

[0157] like Figure 12As shown, the language processing network can be trained using a word vector representation algorithm. For example, the word vector representation algorithm can be the SkipGram algorithm (an algorithm used to train word vector representations in natural language processing). The SkipGram network training logic is to use word vectors to represent each word, and the vector of each word is learned from the ability to predict the surrounding context words (also called background words). In other words, the SkipGram algorithm learns word vectors by predicting the context words around the word. For ease of understanding, the number of relational texts can be represented as M, and any relational text among the M relational texts can be represented as the jth relational text. The jth relational text can include multiple words, and j and M are both positive integers, and j is less than or equal to M. Based on the network algorithm logic of the SkipGram algorithm, the following, combined with the structure of the language processing network, takes any relational text (i.e., the jth relational text) as an example to introduce the training process of the language processing network based on the jth relational text.

[0158] The structure of the language processing network is as follows Figure 14 As shown, the language processing network may include an input layer, a hidden layer, and an output layer. Among them: the input layer can be used to encode the words in the relational text to obtain the encoding vector of the word, and the encoding vector can be, for example, a one-hot encoding vector (single-hot encoding vector). The hidden layer can be used to perform semantic understanding on the encoding vector of the word in the relational text to obtain the semantic vector of the word. The output layer can be used to classify and predict the background words of the word based on the semantic vector of the word. Based on Figure 13 The structure of the language processing network shown in FIG. 1 is based on the j-th relation text. The training process of the language processing network may include:

[0159] First, each word in the j-th relational text can be taken as the central word in turn, and the marked background words corresponding to the central word can be determined; the marked background words corresponding to the central word can be the words located around the central word in the j-th relational text, and the marked background words corresponding to the central word can specifically include Q words adjacent to the central word and located above the central word, and Q words adjacent to the central word and located below the central word, where Q is a positive integer.

[0160] Secondly, the language processing network can be called to perform semantic understanding on the central word to obtain the semantic vector of the central word, and the language processing network can be called to perform semantic understanding on each other word in the j-th relation text to obtain the semantic vector of each other word, where each other word is a word other than the central word in the j-th relation text. Specifically, the input layer in the language processing network can be called to encode the central word to obtain the encoding vector of the central word, and the hidden layer in the language processing network can be called to perform semantic understanding on the encoding vector of the central word to obtain the semantic vector of the central word; similarly, the input layer in the language processing network can be called to encode other words to obtain the encoding vectors of other words, and the hidden layer in the language processing network can be called to perform semantic understanding on the encoding vectors of other words to obtain the semantic vectors of other words.

[0161] Then, the probability of each other word being the background word corresponding to the central word can be determined based on the semantic vector of the central word and the semantic vector of each other word. Specifically, the output layer in the language processing network can be called to determine the probability of each other word being the background word corresponding to the central word based on the semantic vector of the central word and the semantic vector of each other word.

[0162] Then, based on the difference between the labeled background word and each other word, and according to the probability of each other word as the center word corresponding to the background word, the loss information of the language processing network can be determined, and the language processing network can be trained based on the loss information of the language processing network.

[0163] In a specific implementation, the language processing network can be trained multiple times according to the above-described single training process of the language processing network. When a training termination condition is reached, a trained language processing network is obtained. The training termination condition may include any of the following situations: the loss information of the language processing network is less than or equal to a second loss threshold (the second loss threshold can be set based on experience), and the number of times the language processing network has been trained reaches a second training number threshold (the second training number threshold can be set based on experience).

[0164] ④ Determine the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type as the relationship features of the first game object under the i-th relationship type.

[0165] S1103: Determine the relationship features under each relationship type as the relationship features of the first game object.

[0166] S1104: Perform feature fusion learning on the channel features, game features, and relationship features to predict business evaluation information of the first game object in the business channel.

[0167] In the embodiment of the present application, the execution process of step S1104 is the same as the above Figure 5 The execution process of step S503 in the embodiment shown is the same, and the details can be found in the above Figure 5 The execution process of step S503 in the illustrated embodiment will not be described in detail here.

[0168] S1105 , performing business processing on the first game object within the business channel according to the business evaluation information, where the business processing is related to the game.

[0169] In a specific implementation, the process of performing business processing on the first game object within a business channel based on the business evaluation information may include: determining a business processing strategy corresponding to the first game object within the business channel; the business channel may provide one or more business processing strategies, each corresponding to a respective business processing standard, and the business processing strategy corresponding to the first game object is any one or more of the business processing strategies provided by the business channel. The business evaluation information may then be checked based on the business processing standard corresponding to the business processing strategy of the first game object; if the business evaluation information meets the business processing standard corresponding to the business processing strategy of the first game object, then performing business processing corresponding to the business processing strategy of the first game object within the business channel on the first game object.

[0170] Specifically, corresponding to the application scenario of the business processing method, the business processing strategies provided within the business channel may include any one or more of the following: a channel game promotion strategy, a channel game reward and feedback strategy, and a channel lapsed player recall strategy. The business processing standards corresponding to the business processing strategies may include benchmark evaluation information or a benchmark position. When the business processing standards corresponding to the business processing strategies include benchmark evaluation information, the business evaluation information conforming to the business processing standards corresponding to the business processing strategy of the first game object means that the business evaluation information of the first game object within the business channel is higher than the benchmark evaluation information. When the business processing standards corresponding to the business processing strategies include a benchmark position, the business evaluation information conforming to the business processing standards corresponding to the business processing strategy of the first game object means that the business evaluation information of the first game object within the business channel is arranged at or before the benchmark position.

[0171] Furthermore, the business processing strategy corresponding to the first game object in the business channel can be determined based on the historical access data of the first game object related to the game in the business channel. In detail, the historical access data of the first game object related to the game in the business channel can include any one or more of the following: the active data of the first game object related to the game in the business channel (for example, the time interval between the last time the first game object accessed the game in the business channel and the current time), the number of times the first game object viewed the game promotional content in the business channel, and the number of times the first game object received the game rewards in the business channel. Different business processing strategies can correspond to different determination criteria, and the business processing strategy that matches the historical access data can be determined in each business processing strategy based on the matching between the historical access data of the first game object and the determination criteria corresponding to each business processing strategy, and the matching business processing strategy can be determined as the business processing strategy corresponding to the first game object in the business channel.

[0172] For example, the criteria for determining a channel game promotion strategy may include a threshold for the number of times a game promotional content is viewed (the threshold can be set based on experience). If the number of times a first game object views game promotional content within a business channel exceeds the threshold, it can be determined that the first game object has a habit of viewing game promotional content within the business channel, and the channel game promotion strategy can be determined as the business processing strategy for the first game object. For another example, the criteria for determining a channel game reward feedback strategy may include a threshold for the number of times a game reward is received (the threshold can be set based on experience). If the number of times a first game object receives game rewards within a business channel exceeds the threshold, it can be determined that the first game object has a habit of receiving game rewards within the business channel, and the channel game reward feedback strategy can be determined as the business processing strategy for the first game object. For another example, the criteria for determining a channel churn player recall strategy may include a time interval threshold. If the time interval between the first game object's last game access within the business channel and the current time is greater than the time interval threshold (the time interval threshold can be set based on experience), it can be determined that the first game object has been inactive for a long time, and the channel churn player recall strategy can be determined as the business processing strategy for the first game object. In this way, the business processing strategy determined for the first game object matches the usage habits of the first game object in the business channel, so that the first game object can be provided with business services in the business channel that are more closely matched with the usage habits of the first game object, more targeted, and of higher quality.

[0173] In the embodiments of the present application, the channel characteristics of the game object within the business channel, the game characteristics of the game object within the game, and the relationship characteristics of the game object within the game can be comprehensively considered. By comprehensively considering the multi-angle characteristics of the game object within the business channel and within the game, the business evaluation information of the game object within the business channel can be more accurately determined. Moreover, by converting the relationship network, a graph network, into a node sequence and then into text, the relationship network can be processed using natural language processing methods, which can better mine the relationship characteristics in the relationship network. In addition, the historical access data of the first game object related to the game within the business channel can represent the usage habits of the first game object within the business channel. Based on the historical access data of the first game object related to the game within the business channel, the business processing strategy corresponding to the first game object is determined, and the business processing of the first game object is performed according to the determined business processing strategy. This can match the determined business processing strategy with the usage habits of the first game object within the business channel. In this way, the first game object can be provided with a more targeted and higher-quality business service within the business channel that matches the usage habits of the first game object.

[0174] In experimental verification, for the channel lost game object recall scenario, the embodiment of the present application was used to conduct a first experiment (the experimental scheme of the first experiment was: sorting the business evaluation information of each lost game object and pushing the game recall information to the top 1000 users); and a second experiment was conducted (the experimental scheme of the second experiment was: sorting the lost game objects according to the length of time they were lost and pushing the game recall information to the top 1000 users). The experiments showed that the return rate of lost game objects in the first experiment was higher than the return rate of lost game objects in the second experiment, and the resource consumption of game objects in the channel in the first experiment was higher than the resource consumption of game objects in the business channel in the second experiment.

[0175] The above describes in detail the method of the embodiment of the present application. In order to facilitate better implementation of the above scheme of the embodiment of the present application, the device of the embodiment of the present application is provided below accordingly.

[0176] See Figure 15 , Figure 15 This is a structural diagram of a business processing device provided in an embodiment of the present application. The business processing device can be set in the computer device provided in an embodiment of the present application. The computer device can be the terminal mentioned above, or a business processing system composed of a terminal and a server. Figure 15 The service processing device shown may be a computer program running on a computer device, and the service processing device may be used to execute Figure 5 or Figure 11 Some or all of the steps in the method embodiment shown. Figure 15, the service processing device may include the following units:

[0177] Acquisition unit 1501 is configured to acquire business characteristics of a first game object, the business characteristics including channel characteristics of the first game object within a business channel and game characteristics of the first game object within a game; the channel characteristics are used to characterize access to the game by the first game object within the business channel; the game characteristics are used to characterize access to the game by the first game object; and the business channel is used to provide game-related business services to the first game object outside the game.

[0178] The processing unit 1502 is configured to generate a relationship feature of the first game object based on the relationship network of the first game object within the game;

[0179] The processing unit 1502 is further configured to perform feature fusion learning on the channel features, the game features, and the relationship features to predict business evaluation information of the first game object in the business channel;

[0180] The processing unit 1502 is further configured to perform business processing on the first game object within the business channel according to the business evaluation information, where the business processing is related to the game.

[0181] In one implementation, the relationship network includes one or more relationship sub-networks, each of which corresponds to a different relationship type. The processing unit 1502 is configured to generate a relationship feature of the first game object based on the relationship network of the first game object in the game, specifically to perform the following steps:

[0182] Based on each relationship sub-network, generating a relationship feature of the first game object under the corresponding relationship type;

[0183] The relationship features under each relationship type are determined as the relationship features of the first game object.

[0184] In one implementation, a relationship network includes N relationship sub-networks, any one of the N relationship sub-networks is represented as the i-th relationship sub-network, the i-th relationship sub-network corresponds to the i-th relationship type, i and N are both positive integers, and i is less than or equal to N; the i-th relationship sub-network consists of nodes, edges between any two nodes, and edge weights;

[0185] Among them, the i-th relationship sub-network includes at least a first node and a second node; the first node represents the first game object, and the second node represents the second game object of the game; the edge between the first node and the second node indicates that the first game object and the second game object have a relationship under the i-th relationship type; the weight corresponding to the edge between the first node and the second node represents the relationship quantification value between the first game object and the second game object; the relationship quantification value is used to indicate the intimacy of the relationship between the first game object and the second game object under the i-th relationship type.

[0186] In one implementation, the processing unit 1502 is configured to generate, based on each relationship sub-network, a relationship feature of the first game object under the corresponding relationship type, by performing the following steps:

[0187] Take each node in the i-th relationship sub-network as the starting point, traverse the i-th relationship sub-network, and obtain the node sequence corresponding to each node;

[0188] Determine the relationship text corresponding to each node sequence; the relationship text corresponding to each node sequence is composed of the player identifier of the game object represented by the node in the node sequence, and the weight corresponding to the edge between each two nodes in the node sequence;

[0189] Call the language processing network to perform semantic understanding on each relation text, and obtain the relational features of each game object in the i-th relation sub-network under the i-th relation type;

[0190] The relationship feature of each game object in the i-th relationship sub-network under the i-th relationship type is determined as the relationship feature of the first game object under the i-th relationship type.

[0191] In one implementation, the processing unit 1502 is configured to call a language processing network to perform semantic understanding on each relational text, and to obtain the relational features of each game object in the i-th relational subnetwork under the i-th relational type, and specifically to perform the following steps:

[0192] Based on each relational text, the language processing network is trained;

[0193] Call the trained language processing network to perform semantic understanding on the player ID of each game object in the i-th relationship sub-network, and obtain the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type.

[0194] In one implementation, the number of relation texts is M, any one of the M relation texts is represented as the j-th relation text, j and M are both positive integers, and j is less than or equal to M; the j-th relation text includes multiple words; and the training process of the language processing network based on the j-th relation text includes:

[0195] Take each word in the j-th relation text as the central word in turn, and determine the marked background word corresponding to the central word;

[0196] Calling the language processing network to perform semantic understanding on the central word to obtain a semantic vector of the central word; and calling the language processing network to perform semantic understanding on each other word in the j-th relation text to obtain a semantic vector of each other word, where each other word is a word in the j-th relation text other than the central word;

[0197] According to the semantic vector of the central word and the semantic vector of each other word, the probability of each other word being the background word corresponding to the central word is determined;

[0198] Based on the difference between the labeled background word and each other word, the loss information of the language processing network is determined according to the probability of each other word being the center word corresponding to the background word;

[0199] The language processing network is trained based on the loss information of the language processing network.

[0200] In one implementation, the channel characteristics include at least one of the following: a channel activity characteristic, a channel interaction characteristic, and a channel consumption characteristic; the channel activity characteristic is used to characterize the activity of the first game object in the business channel and related to the game; the channel interaction characteristic is used to characterize the interaction of the first game object in the business channel and related to the game; the channel consumption characteristic is used to characterize the resource consumption of the first game object in the business channel and related to the game;

[0201] The game features include at least one of the following: game activity features, game interaction features, and game consumption features; the game activity features are used to characterize the activity of the first game object in the game; the game interaction features are used to characterize the interaction of the first game object in the game; and the game consumption features are used to characterize the resource consumption of the first game object in the game.

[0202] In one implementation, the processing unit 1502 is configured to perform feature fusion learning on the channel feature, the game feature, and the relationship feature to predict the business evaluation information of the first game object in the business channel, specifically to perform the following steps:

[0203] Perform semantic understanding on channel features to obtain the channel semantic vector corresponding to the channel features;

[0204] Perform semantic understanding on game features to obtain the game semantic vector corresponding to the game features;

[0205] Fuse the channel semantic vector, game semantic vector, and relationship features to obtain fused features;

[0206] Feature learning is performed on the fused features, and the feature learning results of the fused features are used as business evaluation information of the first game object.

[0207] In one implementation, feature fusion learning is performed by calling a trained business evaluation model. The training process of the business evaluation model includes:

[0208] Obtaining sample business features of the sample game object, the sample business features including sample channel features of the sample game object in the business channel, game features of the sample game object in the game, and sample relationship features of the sample game object;

[0209] Perform feature fusion learning on sample channel features, sample game features, and sample relationship features to predict sample business evaluation information of sample game objects in the business channel;

[0210] Predicting a business processing method for a sample game object based on sample business evaluation information;

[0211] According to the difference between the predicted business processing mode and the actual business processing mode corresponding to the sample game object, the loss of the sample business evaluation information is calculated to obtain the loss information of the business evaluation model;

[0212] The business evaluation model is trained based on its loss information.

[0213] In one implementation, the processing unit 1502 is configured to perform the following steps when performing business processing on the first game object in the business channel according to the business evaluation information:

[0214] Determine a business processing strategy corresponding to the first game object in the business channel; the business channel provides one or more business processing strategies, each business processing strategy corresponds to a respective business processing standard, and the business processing strategy corresponding to the first game object is any one or more business processing strategies provided by the business channel;

[0215] Checking the business evaluation information according to a business processing standard corresponding to the business processing strategy of the first game object;

[0216] If the business evaluation information meets the business processing standard corresponding to the business processing strategy of the first game object, the business processing corresponding to the business processing strategy of the first game object is performed on the first game object in the business channel.

[0217] According to another embodiment of the present application, Figure 15 The various units in the business processing device shown can be individually or all combined into one or several other units to form, or one (or some) of the units can be further divided into multiple functionally smaller units to form, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above-mentioned units are divided based on logical functions. In actual applications, the functions of one unit can also be implemented by multiple units, or the functions of multiple units can be implemented by one unit. In other embodiments of the present application, the business processing device may also include other units. In actual applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of multiple units.

[0218] According to another embodiment of the present application, the program can be executed by running on a general computing device such as a computer including a central processing unit (CPU), a random access memory (RAM), a read-only memory (ROM) and other processing elements and storage elements. Figure 5 or Figure 11 A computer program for each step involved in part or all of the method shown is constructed as follows Figure 15 The business processing device shown in and the business processing method of the embodiment of the present application are implemented. The computer program can be recorded on, for example, a computer-readable storage medium, and loaded into the above-mentioned computing device through the computer-readable storage medium and run therein.

[0219] In an embodiment of the present application, feature fusion learning can be performed on the channel characteristics of the game object in the business channel, the game characteristics of the game object in the game, and the relationship characteristics of the game object in the game to predict the business evaluation information of the game object in the business channel; that is, the channel characteristics of the game object in the business channel, the game characteristics of the game object in the game, and the relationship characteristics of the game object in the game can be comprehensively considered, and the multi-angle characteristics of the game object in the business channel and in the game can be comprehensively considered to more accurately determine the business evaluation information of the game object. Therefore, based on the more accurate business evaluation information, effective business processing can be performed on the business object to avoid waste of resources involved in business processing.

[0220] Based on the above method and device embodiments, the present application provides a computer device. Figure 16 , Figure 16 It is a structural diagram of a computer device provided in an embodiment of the present application. Figure 16The computer device shown includes at least a processor 1601, an input interface 1602, an output interface 1603, and a computer-readable storage medium 1604. The processor 1601, the input interface 1602, the output interface 1603, and the computer-readable storage medium 1604 may be connected via a bus or other means.

[0221] Computer-readable storage medium 1604 may be stored in a memory of a computer device. Computer-readable storage medium 1604 is used to store a computer program, which includes computer instructions. Processor 1601 is used to execute the computer program stored in computer-readable storage medium 1604. Processor 1601 (or CPU (Central Processing Unit)) is the computing and control core of the computer device and is suitable for implementing computer programs, specifically loading and executing computer programs to implement corresponding method processes or corresponding functions.

[0222] The embodiment of the present application also provides a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space that stores the operating system of the computer device. In addition, a computer program suitable for being loaded and executed by the processor is also stored in the storage space. It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory (Non-Volatile Memory), such as at least one disk memory; optionally, it can also be at least one computer-readable storage medium located away from the aforementioned processor.

[0223] The computer device may be the terminal mentioned above, or a business processing system composed of a terminal and a server. In a specific implementation, the processor 1601 may load and execute the computer program stored in the computer-readable storage medium 1604 to implement the above-mentioned Figure 5 or Figure 11 In a specific implementation, the computer program in the computer-readable storage medium 1604 is loaded by the processor 1601 and executes the following steps:

[0224] Obtaining business characteristics of the first game object, the business characteristics including channel characteristics of the first game object within the business channel and game characteristics of the first game object within the game; the channel characteristics are used to characterize the first game object's access to the game within the business channel; the game characteristics are used to characterize the first game object's access to the game; the business channel is used to provide game-related business services to the first game object outside the game;

[0225] generating a relationship feature of the first game object based on a relationship network of the first game object within the game;

[0226] Perform feature fusion learning on channel features, game features, and relationship features to predict business evaluation information of the first game object in the business channel;

[0227] According to the business evaluation information, business processing is performed on the first game object within the business channel, where the business processing is related to the game.

[0228] In one implementation, the relationship network includes one or more relationship sub-networks, each of which corresponds to a different relationship type. When the computer program in the computer-readable storage medium 1604 is loaded and executed by the processor 1601, the computer program generates a relationship feature of the first game object based on the relationship network of the first game object in the game, specifically for performing the following steps:

[0229] Based on each relationship sub-network, generating a relationship feature of the first game object under the corresponding relationship type;

[0230] The relationship features under each relationship type are determined as the relationship features of the first game object.

[0231] In one implementation, a relationship network includes N relationship sub-networks, any one of the N relationship sub-networks is represented as the i-th relationship sub-network, the i-th relationship sub-network corresponds to the i-th relationship type, i and N are both positive integers, and i is less than or equal to N; the i-th relationship sub-network consists of nodes, edges between any two nodes, and edge weights;

[0232] Among them, the i-th relationship sub-network includes at least a first node and a second node; the first node represents the first game object, and the second node represents the second game object of the game; the edge between the first node and the second node indicates that the first game object and the second game object have a relationship under the i-th relationship type; the weight corresponding to the edge between the first node and the second node represents the relationship quantification value between the first game object and the second game object; the relationship quantification value is used to indicate the intimacy of the relationship between the first game object and the second game object under the i-th relationship type.

[0233] In one implementation, when the computer program in the computer-readable storage medium 1604 is loaded and executed by the processor 1601 to generate a relationship feature of the first game object under the corresponding relationship type based on each relationship sub-network, the computer program is specifically configured to perform the following steps:

[0234] Take each node in the i-th relationship sub-network as the starting point, traverse the i-th relationship sub-network, and obtain the node sequence corresponding to each node;

[0235] Determine the relationship text corresponding to each node sequence; the relationship text corresponding to each node sequence is composed of the player identifier of the game object represented by the node in the node sequence, and the weight corresponding to the edge between each two nodes in the node sequence;

[0236] Call the language processing network to perform semantic understanding on each relation text, and obtain the relational features of each game object in the i-th relation sub-network under the i-th relation type;

[0237] The relationship feature of each game object in the i-th relationship sub-network under the i-th relationship type is determined as the relationship feature of the first game object under the i-th relationship type.

[0238] In one implementation, the computer program in the computer-readable storage medium 1604 is loaded and executed by the processor 1601 to call the language processing network to perform semantic understanding on each relationship text, and to obtain the relationship features of each game object in the i-th relationship subnetwork under the i-th relationship type, specifically to perform the following steps:

[0239] Based on each relational text, the language processing network is trained;

[0240] Call the trained language processing network to perform semantic understanding on the player ID of each game object in the i-th relationship sub-network, and obtain the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type.

[0241] In one implementation, the number of relation texts is M, any one of the M relation texts is represented as the j-th relation text, j and M are both positive integers, and j is less than or equal to M; the j-th relation text includes multiple words; and the training process of the language processing network based on the j-th relation text includes:

[0242] Take each word in the j-th relation text as the central word in turn, and determine the marked background word corresponding to the central word;

[0243] Calling the language processing network to perform semantic understanding on the central word to obtain a semantic vector of the central word; and calling the language processing network to perform semantic understanding on each other word in the j-th relation text to obtain a semantic vector of each other word, where each other word is a word in the j-th relation text other than the central word;

[0244] According to the semantic vector of the central word and the semantic vector of each other word, the probability of each other word being the background word corresponding to the central word is determined;

[0245] Based on the difference between the labeled background word and each other word, the loss information of the language processing network is determined according to the probability of each other word being the center word corresponding to the background word;

[0246] The language processing network is trained based on the loss information of the language processing network.

[0247] In one implementation, the channel characteristics include at least one of the following: a channel activity characteristic, a channel interaction characteristic, and a channel consumption characteristic; the channel activity characteristic is used to characterize the activity of the first game object in the business channel and related to the game; the channel interaction characteristic is used to characterize the interaction of the first game object in the business channel and related to the game; the channel consumption characteristic is used to characterize the resource consumption of the first game object in the business channel and related to the game;

[0248] The game features include at least one of the following: game activity features, game interaction features, and game consumption features; the game activity features are used to characterize the activity of the first game object in the game; the game interaction features are used to characterize the interaction of the first game object in the game; and the game consumption features are used to characterize the resource consumption of the first game object in the game.

[0249] In one implementation, the computer program in the computer-readable storage medium 1604 is loaded and executed by the processor 1601 to perform feature fusion learning on the channel features, the game features, and the relationship features, and to predict the business evaluation information of the first game object in the business channel, specifically for performing the following steps:

[0250] Perform semantic understanding on channel features to obtain the channel semantic vector corresponding to the channel features;

[0251] Perform semantic understanding on game features to obtain the game semantic vector corresponding to the game features;

[0252] Fuse the channel semantic vector, game semantic vector, and relationship features to obtain fused features;

[0253] Feature learning is performed on the fused features, and the feature learning results of the fused features are used as business evaluation information of the first game object.

[0254] In one implementation, feature fusion learning is performed by calling a trained business evaluation model. The training process of the business evaluation model includes:

[0255] Obtaining sample business features of the sample game object, the sample business features including sample channel features of the sample game object in the business channel, game features of the sample game object in the game, and sample relationship features of the sample game object;

[0256] Perform feature fusion learning on sample channel features, sample game features, and sample relationship features to predict sample business evaluation information of sample game objects in the business channel;

[0257] Predicting a business processing method for a sample game object based on sample business evaluation information;

[0258] According to the difference between the predicted business processing mode and the actual business processing mode corresponding to the sample game object, the loss of the sample business evaluation information is calculated to obtain the loss information of the business evaluation model;

[0259] The business evaluation model is trained based on its loss information.

[0260] In one implementation, the computer program in the computer-readable storage medium 1604 is loaded and executed by the processor 1601 to perform the following steps when performing business processing on the first game object in the business channel based on the business evaluation information:

[0261] Determine a business processing strategy corresponding to the first game object in the business channel; the business channel provides one or more business processing strategies, each business processing strategy corresponds to a respective business processing standard, and the business processing strategy corresponding to the first game object is any one or more business processing strategies provided by the business channel;

[0262] Checking the business evaluation information according to a business processing standard corresponding to the business processing strategy of the first game object;

[0263] If the business evaluation information meets the business processing standard corresponding to the business processing strategy of the first game object, the business processing corresponding to the business processing strategy of the first game object is performed on the first game object in the business channel.

[0264] In an embodiment of the present application, feature fusion learning can be performed on the channel characteristics of the game object in the business channel, the game characteristics of the game object in the game, and the relationship characteristics of the game object in the game to predict the business evaluation information of the game object in the business channel; that is, the channel characteristics of the game object in the business channel, the game characteristics of the game object in the game, and the relationship characteristics of the game object in the game can be comprehensively considered, and the multi-angle characteristics of the game object in the business channel and in the game can be comprehensively considered to more accurately determine the business evaluation information of the game object. Therefore, based on the more accurate business evaluation information, effective business processing can be performed on the business object to avoid waste of resources involved in business processing.

[0265] The present application also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the aforementioned service processing method.

[0266] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0267] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program that has a predetermined function and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0268] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When software is used for implementation, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrations. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0269] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A business processing method, characterized in that: include: Obtaining business characteristics of a first game object, the business characteristics including channel characteristics of the first game object within a business channel and game characteristics of the first game object within a game; the channel characteristics are used to characterize access of the first game object to the game within the business channel; the game characteristics are used to characterize access of the first game object within the game; the business channel is used to provide business services related to the game to the first game object outside the game; generating a relationship feature of the first game object based on a relationship network of the first game object within the game; Performing feature fusion learning on the channel features, the game features, and the relationship features to predict business evaluation information of the first game object in the business channel; According to the business evaluation information, business processing is performed on the first game object in the business channel, where the business processing is related to the game.

2. The method according to claim 1, wherein The relationship network includes one or more relationship sub-networks, each of which corresponds to a different relationship type. Generating the relationship feature of the first game object based on the relationship network of the first game object in the game includes: Based on each of the relationship sub-networks, generating a relationship feature of the first game object under a corresponding relationship type; The relationship features under each of the relationship types are determined as the relationship features of the first game object.

3. The method according to claim 2, wherein The relationship network includes N relationship sub-networks, any one of the N relationship sub-networks is represented as the i-th relationship sub-network, the i-th relationship sub-network corresponds to the i-th relationship type, i and N are both positive integers, and i is less than or equal to N; the i-th relationship sub-network consists of nodes, edges between any two nodes, and edge weights; Wherein, the i-th relationship sub-network includes at least a first node and a second node; the first node represents the first game object, and the second node represents the second game object of the game; the edge between the first node and the second node indicates that the first game object and the second game object have a relationship under the i-th relationship type; the weight corresponding to the edge between the first node and the second node represents the relationship quantification value between the first game object and the second game object; the relationship quantification value is used to indicate the intimacy of the relationship between the first game object and the second game object under the i-th relationship type.

4. The method according to claim 3, wherein The generating, based on each of the relationship sub-networks, a relationship feature of the first game object under a corresponding relationship type includes: Taking each of the nodes in the i-th relationship sub-network as a starting point, traversing the i-th relationship sub-network to obtain a node sequence corresponding to each node; Determine a relationship text corresponding to each node sequence; the relationship text corresponding to each node sequence is composed of a player identifier of a game object represented by a node in the node sequence, and a weight corresponding to an edge between two nodes in the node sequence; Calling a language processing network to perform semantic understanding on each of the relationship texts to obtain the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type; The relationship feature of each game object in the i-th relationship sub-network under the i-th relationship type is determined as the relationship feature of the first game object under the i-th relationship type.

5. The method according to claim 4, wherein The calling of the language processing network to perform semantic understanding on each of the relationship texts to obtain the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type includes: Training the language processing network based on each of the relational texts; The trained language processing network is called to perform semantic understanding on the player identifier of each game object in the i-th relationship sub-network, and obtain the relationship features of each game object in the i-th relationship sub-network under the i-th relationship type.

6. The method according to claim 5, wherein The number of the relationship texts is M, and any relationship text among the M relationship texts is represented as the j-th relationship text, where j and M are both positive integers, and j is less than or equal to M; The j-th relation text includes multiple words; The training process of the language processing network based on the j-th relation text includes: Taking each of the words in the j-th relation text as a central word in turn, and determining a marked background word corresponding to the central word; Calling the language processing network to perform semantic understanding on the central word to obtain a semantic vector of the central word; and calling the language processing network to perform semantic understanding on each other word in the j-th relationship text to obtain a semantic vector of each other word, where each other word is a word in the j-th relationship text other than the central word; Determining, based on the semantic vector of the central word and the semantic vector of each of the other words, the probability that each of the other words serves as a background word corresponding to the central word; Determining loss information of the language processing network based on a difference between the marked background word and each of the other words and a probability of each of the other words being the background word corresponding to the center word; The language processing network is trained based on the loss information of the language processing network.

7. The method according to claim 1, wherein The channel characteristics include at least one of the following: a channel activity characteristic, a channel interaction characteristic, and a channel consumption characteristic; the channel activity characteristic is used to characterize the activity of the first game object in the business channel and related to the game; the channel interaction characteristic is used to characterize the interaction of the first game object in the business channel and related to the game; the channel consumption characteristic is used to characterize the resource consumption of the first game object in the business channel and related to the game; The game characteristics include at least one of the following: game activity characteristics, game interaction characteristics, and game consumption characteristics; The game activity feature is used to characterize the activity of the first game object in the game; The game interaction feature is used to characterize the interaction of the first game object in the game; the game consumption feature is used to characterize the resource consumption of the first game object in the game.

8. The method according to claim 1 or 7, wherein: The performing feature fusion learning on the channel feature, the game feature, and the relationship feature to predict business evaluation information of the first game object in the business channel includes: Performing semantic understanding on the channel features to obtain a channel semantic vector corresponding to the channel features; Performing semantic understanding on the game features to obtain game semantic vectors corresponding to the game features; fusing the channel semantic vector, the game semantic vector, and the relationship feature to obtain a fused feature; Feature learning is performed on the fused features, and feature learning results of the fused features are used as business evaluation information of the first game object.

9. The method according to claim 1, wherein The feature fusion learning is performed by calling a trained business evaluation model; the training process of the business evaluation model includes: Acquire sample business features of a sample game object, the sample business features including sample channel features of the sample game object in the business channel, game features of the sample game object in the game, and sample relationship features of the sample game object; Performing feature fusion learning on the sample channel features, the sample game features, and the sample relationship features to predict sample business evaluation information of the sample game object in the business channel; Predicting a business processing method for the sample game object based on the sample business evaluation information; performing loss calculation on the sample business evaluation information according to a difference between the predicted business processing mode and the actual business processing mode corresponding to the sample game object to obtain loss information of the business evaluation model; The business evaluation model is trained according to the loss information of the business evaluation model.

10. The method according to claim 1, wherein The performing business processing on the first game object in the business channel according to the business evaluation information includes: Determining a business processing strategy corresponding to the first game object in the business channel; the business channel provides one or more business processing strategies, each of which corresponds to a respective business processing standard, and the business processing strategy corresponding to the first game object is any one or more business processing strategies provided by the business channel; checking the business evaluation information according to a business processing standard corresponding to the business processing policy of the first game object; If the service evaluation information meets the service processing standard corresponding to the service processing policy of the first game object, the service processing corresponding to the service processing policy of the first game object is performed on the first game object in the service channel.

11. A business processing device, characterized in that: include: an acquisition unit, configured to acquire business characteristics of a first game object, the business characteristics including channel characteristics of the first game object within a business channel and game characteristics of the first game object within a game; the channel characteristics are used to characterize access of the first game object to the game within the business channel; the game characteristics are used to characterize access of the first game object within the game; and the business channel is used to provide business services related to the game to the first game object outside the game; a processing unit, configured to generate a relationship feature of the first game object based on a relationship network of the first game object within the game; The processing unit is further configured to perform feature fusion learning on the channel feature, the game feature, and the relationship feature to predict business evaluation information of the first game object in the business channel; The processing unit is further configured to perform business processing on the first game object within the business channel according to the business evaluation information, where the business processing is related to the game.

12. A computer device, characterized in that: The computer device comprises: a processor suitable for implementing a computer program; A computer-readable storage medium storing a computer program, wherein the computer program is suitable for being loaded by the processor and executing the service processing method according to any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing the service processing method according to any one of claims 1 to 10.

14. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the service processing method according to any one of claims 1 to 10 is implemented.

Images