Virtual resource processing method, device, equipment, computer program and storage medium

By building a social graph network and performing vectorization processing, combining virtual resource operation characteristics and social relationship data, the accuracy of virtual resource demand prediction is improved, and the problem of insufficient accuracy of virtual resource demand prediction in the prior art is solved.

CN114470790BActive Publication Date: 2024-12-06TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210122353.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-02-09
Publication Date
2024-12-06
Estimated Expiration
2042-02-09

AI Technical Summary

Technical Problem

The prior art has low accuracy in virtual resource demand prediction in virtual scenarios and is susceptible to subjective experience and random factors.

Method used

By building a social graph network, vectorized processing is performed, virtual resource operation characteristics are extracted, and virtual resource demand prediction is carried out in combination with social relationship data.

Benefits of technology

Improve the accuracy of virtual resource demand prediction and reduce subjective impact and random error.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114470790B_ABST
    Figure CN114470790B_ABST
Patent Text Reader

Abstract

The present application provides a virtual resource processing method, device, equipment, computer program and storage medium. The method includes: constructing a relationship graph network based on at least one type of social relationship data in a virtual scene to obtain at least one social graph network; performing vectorization processing on at least one social graph network to obtain at least one vectorization result; for each account included in at least one social graph network, extracting the virtual resource operation data corresponding to each account, and generating the resource operation characteristics corresponding to each account based on the virtual resource operation data; predicting the virtual resource demand based on the resource operation characteristics and at least one vectorization result to obtain the virtual resource demand quantity of each account. Through this application, the accuracy of virtual resource demand prediction is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a virtual resource processing method, device, equipment, computer program and storage medium. Background Art

[0002] At present, related technologies are usually based on empirical rules to predict users' demand for virtual resources in virtual scenes, such as predicting players' game needs based on their historical consumption behaviors in the game over a certain period of time. Alternatively, they are based on conventional machine learning regression algorithms, which use regression algorithms to learn players' consumption habits in the game and then predict players' consumption behaviors. It can be seen that the prediction method based on empirical rules is easily affected by subjective experience and has low accuracy, while the prediction method based on conventional machine learning theory is easily affected by random factors and has large errors, which also reduces the accuracy of virtual resource demand prediction. Summary of the invention

[0003] Embodiments of the present application provide a virtual resource processing method, apparatus, device, computer program, and storage medium, which can improve the accuracy of virtual resource demand prediction.

[0004] The technical solution of the embodiment of the present application is implemented as follows:

[0005] The present application provides a method for processing virtual resources, including:

[0006] Building a relationship graph network based on at least one type of social relationship data in the virtual scene to obtain at least one social graph network;

[0007] Performing vectorization processing on the at least one social graph network to obtain the at least one vectorization result;

[0008] For each account included in the at least one social graph network, extracting virtual resource operation data corresponding to each account, and generating resource operation features corresponding to each account based on the virtual resource operation data;

[0009] The virtual resource demand is predicted based on the resource operation characteristics and the at least one vectorization result to obtain the virtual resource demand quantity of each account.

[0010] The embodiment of the present application provides a virtual resource processing device, including:

[0011] A construction module, configured to construct a relationship graph network based on at least one type of social relationship data in a virtual scene, and obtain at least one social graph network;

[0012] A vectorization module, configured to perform vectorization processing on the at least one social graph network to obtain the at least one vectorization result;

[0013] a feature processing module, configured to extract, for each account included in the at least one social graph network, virtual resource operation data corresponding to each account, and generate resource operation features corresponding to each account based on the virtual resource operation data;

[0014] A prediction module is used to predict the virtual resource demand based on the resource operation characteristics and the at least one vectorization result to obtain the virtual resource demand quantity of each account.

[0015] In the above-mentioned device, the construction module is also used to obtain, for each type of social relationship data in the at least one type of social relationship data, the accounts in the each type of social relationship data as nodes, by connecting the nodes where this type of social relationship exists, a social graph network corresponding to the each type of social relationship data, thereby obtaining the at least one social graph network.

[0016] In the above-mentioned device, the vectorization module is also used to obtain the initial weight of each connection relationship in each social graph network in the at least one social graph network; the initial weight represents the degree of interaction between the accounts corresponding to each connection relationship; according to the initial weight of each connection relationship, the initial weight distribution corresponding to each connection relationship is calculated; vector initialization is performed on each node in each social graph network to obtain the initial vectorized representation of each node; weight distribution prediction is performed based on the initial vectorized representation of each node to obtain the predicted weight distribution corresponding to each connection relationship; the initial vectorized representation is iteratively updated by fitting the initial weight distribution and the predicted weight distribution until the preset fitting condition is met, and the vectorized representation of each node is obtained as the vectorized result of each social graph network, thereby obtaining the at least one vectorized result.

[0017] In the above-mentioned device, the at least one type of social relationship data includes at least one of the following: friendship relationship data, activity relationship data, organization relationship data and intimacy data; the vectorization module is also used to obtain the friendship experience value between the accounts corresponding to each connection relationship as the initial weight when each social graph network is constructed based on the friendship relationship data; obtain the number of times the accounts corresponding to each connection relationship jointly participate in virtual scene activities as the initial weight when each social graph network is constructed based on the activity relationship data; obtain the initial weight according to the organization information in which the accounts corresponding to each connection relationship participate when each social graph network is constructed based on the organization relationship data; obtain the intimacy value between the accounts corresponding to each connection relationship as the initial weight when each social graph network is constructed based on the intimacy relationship data.

[0018] In the above device, the vectorization module is also used to calculate the sum of the weights of each social graph network according to the initial weight of each connection relationship; and calculate the ratio of the initial weight of each connection relationship to the sum of the weights as the initial weight distribution corresponding to each connection relationship.

[0019] In the above-mentioned device, the vectorization module is also used to calculate the difference between the initial weight distribution and the predicted weight distribution using a preset target optimization function; when the difference does not meet the preset fitting condition, the initial vectorization representation of each node is updated to obtain the intermediate vectorization representation of each node; weight distribution prediction is performed based on the intermediate vectorization representation of each node, the predicted weight distribution is updated, and the difference is updated based on the updated predicted weight distribution and the initial weight distribution, until the difference meets the preset fitting condition, and the intermediate vectorization representation of each node is used as the vectorization representation of each node.

[0020] In the above device, the feature processing module is also used to obtain at least one resource quantity data of the virtual resources in each account in at least one quantity dimension within at least one preset time period as the virtual resource operation data.

[0021] In the above device, the feature processing module is also used to generate the resource operation feature corresponding to the at least one feature dimension by mapping the at least one quantity dimension to at least one feature dimension, based on mapping the at least one resource quantity data to at least one feature dimension space.

[0022] In the above device, the feature processing module is also used to extract at least one activity data of each account within at least one preset time period in the virtual scene; the at least one activity data is the data generated by the activities of each account in the virtual scene; and the resource operation feature is generated by performing feature dimension mapping on the at least one resource quantity data and the at least one activity data.

[0023] In the above-mentioned device, the prediction module is also used to obtain the associated resource data of each account in the associated virtual scene, and obtain the associated resource characteristics based on the associated resource data; the associated resource data is the data corresponding to the virtual resource operation performed by each account in the associated virtual scene; based on the associated resource characteristics, the resource operation characteristics and the at least one vectorization result are used to predict the virtual resource demand to obtain the virtual resource demand quantity of each account.

[0024] In the above device, the prediction module is also used to perform feature fusion on the associated resource characteristics, the resource operation characteristics and the at least one vectorization result to obtain a fusion result; and predict the virtual resource demand of the fusion result, the virtual resource demand quantity of each account.

[0025] An embodiment of the present application provides an electronic device, including:

[0026] A memory for storing executable instructions;

[0027] The processor is used to implement the virtual resource processing method provided in the embodiment of the present application when executing the executable instructions stored in the memory.

[0028] An embodiment of the present application provides a computer-readable storage medium storing executable instructions for causing a processor to execute and implement the virtual resource processing method provided in the embodiment of the present application.

[0029] An embodiment of the present application provides a computer program product, including a computer program or instructions, characterized in that when the computer program or instructions are executed by a processor, the virtual resource processing method provided in the embodiment of the present application is implemented.

[0030] The embodiments of the present application have the following beneficial effects:

[0031] By using the social relationship data between each account in the virtual scene, a social graph network is constructed and vectorized, so that social relationship information is introduced into the prediction task of virtual resource demand, and the virtual resource demand is predicted by using the vectorization results of social relationship data and combining the account's operation behavior on virtual resources. Since in actual applications, the account's demand for virtual resources will be affected by social relationship factors, the embodiment of the present application combines quantified social relationships to predict the virtual resource demand of each account, which can greatly improve the accuracy of virtual resource demand prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is an optional structural diagram of the virtual resource processing system architecture provided by the embodiment of the present application;

[0033] Figure 2 It is an optional structural diagram of a virtual resource processing device provided in an embodiment of the present application;

[0034] Figure 3 It is an optional flow chart of the virtual resource processing method provided in the embodiment of the present application;

[0035] Figure 4 is an optional schematic diagram of the vectorization processing provided in the embodiment of the present application;

[0036] Figure 5 It is an optional flow chart of the virtual resource processing method provided in the embodiment of the present application;

[0037] Figure 6 It is an optional flow chart of the virtual resource processing method provided in the embodiment of the present application;

[0038] Figure 7 It is an optional flow chart of the virtual resource processing method provided in the embodiment of the present application;

[0039] Figure 8 It is an optional flow chart of the virtual resource processing method provided in the embodiment of the present application;

[0040] Fig. 9 It is an optional flow chart of the virtual resource processing method provided in the embodiment of the present application applied to an actual scenario;

[0041] Fig.10 It is an optional flow chart of the application of the virtual resource processing method provided in the embodiment of the present application to an actual scenario. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0043] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0044] In the following description, the terms "first\second\third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0045] The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C can represent including any one or more elements selected from the set consisting of A, B, and C.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0047] Before further describing the embodiments of the present application in detail, the nouns and terms involved in the embodiments of the present application are explained. The nouns and terms involved in the embodiments of the present application are subject to the following interpretations.

[0048] 1) Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0049] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. Basic artificial intelligence technologies generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operating / interactive systems, mechatronics and other technologies. Artificial intelligence software technologies mainly include computer vision technology, speech processing technology, natural language processing technology, as well as machine learning / deep learning, autonomous driving, smart transportation and other major directions.

[0050] 2) Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0051] 3) Multilayer Perceptron (MLP) is a feed-forward artificial neural network model that maps multiple input data sets to a single output data set.

[0052] 4) Graph Neural Networks (GNN) is a neural network that runs directly on a graph structure. The graph in GNN refers to a graph consisting of several nodes and edges connecting two nodes, which is used to describe the relationship between different nodes. In a given graph, each node has its own characteristics. The learning goal of GNN is to obtain the graph-aware hidden state of each node, where the hidden state contains information from neighboring nodes. In other words, the learning goal of GNN is to make each node aware of other nodes on the graph.

[0053] With the research and advancement of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, Internet of Vehicles, automatic driving, smart transportation, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0054] The embodiments of the present application provide a virtual resource processing method, apparatus, device, computer program and storage medium, which can improve the accuracy of virtual resource demand prediction. The exemplary application of the electronic device provided by the embodiments of the present application is described below. The electronic device provided by the embodiments of the present application can be implemented as various types of terminals or user terminals such as smart phones, smart watches, laptops, tablet computers, desktop computers, set-top boxes, mobile devices (e.g., mobile phones, portable music players, personal digital assistants, dedicated messaging devices, portable gaming devices), intelligent voice interaction devices, smart home appliances and vehicle-mounted terminals, and can also be implemented as servers.

[0055] Next, an exemplary application in which the electronic device is implemented as a server will be described.

[0056] See also Figure 1 , Figure 1 The terminal 400 (exemplarily showing the terminal 400-1 and the terminal 400-2) is connected to the server 200 via the network 300, which may be a wide area network or a local area network, or a combination of the two.

[0057] The terminal 400 is used to log in and run a virtual scene application, such as a client of an online game, through an account, and receive operations of a user, such as a game player, through a graphical interface 410 of the virtual scene application (graphic interface 410-1 and graphical interface 410-2 are shown as examples), and upload the user's operation data to the server 200 running the background service of the virtual scene application, so that the server 200 generates response data to the user's operation data, and realizes the account's activities in the virtual scene through the response data, such as operating virtual resources in the virtual scene, or conducting social interactions in the virtual scene, etc., thereby generating at least one type of social relationship data in the virtual scene and virtual resource operation data corresponding to each account. At least one type of social relationship data and virtual resource operation data can be stored in the database 500.

[0058] The server 200 is used to construct a relationship graph network from the database 500 based on at least one type of social relationship data in the virtual scene to obtain at least one social graph network; perform vectorization processing on the at least one social graph network to obtain at least one vectorization result; for each account included in the at least one social graph network, extract the virtual resource operation data corresponding to each account, and generate the resource operation characteristics corresponding to each account based on the virtual resource operation data; predict the virtual resource demand based on the resource operation characteristics of each account and at least one vectorization result to obtain the virtual resource demand quantity of each account.

[0059] In some embodiments, the server may be an independent physical server, or a server cluster or 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, and big data and artificial intelligence platforms. The terminal may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, an intelligent voice interaction device, a smart home appliance, and a vehicle-mounted terminal, etc., but is not limited thereto. The terminal and the server may be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiments of the present application.

[0060] See also Figure 2 , Figure 2 is a schematic diagram of the structure of the server 200 provided in an embodiment of the present application, Figure 2 The server 200 shown includes: at least one processor 210, a memory 250, at least one network interface 220 and a user interface 230. The various components in the server 200 are coupled together via a bus system 240. It is understood that the bus system 240 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 240 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the bus system 240 is not described in detail. Figure 2 Various buses are labeled as bus system 240 .

[0061] The processor 210 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., where the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0062] The user interface 230 includes one or more output devices 231 that enable presentation of media content, including one or more speakers and / or one or more visual display screens. The user interface 230 also includes one or more input devices 232, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

[0063] The memory 250 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard disk drives, optical disk drives, etc. The memory 250 may optionally include one or more storage devices that are physically remote from the processor 210.

[0064] The memory 250 includes a volatile memory or a non-volatile memory, and may also include both volatile and non-volatile memories. The non-volatile memory may be a read-only memory (ROM), and the volatile memory may be a random access memory (RAM). The memory 250 described in the embodiments of the present application is intended to include any suitable type of memory.

[0065] In some embodiments, memory 250 can store data to support various operations, examples of which include programs, modules, and data structures, or a subset or superset thereof, as exemplarily described below.

[0066] Operating system 251, including system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks;

[0067] A network communication module 252, used to reach other computing devices via one or more (wired or wireless) network interfaces 220, exemplary network interfaces 220 include: Bluetooth, Wireless Compatibility Certification (WiFi), and Universal Serial Bus (USB);

[0068] a presentation module 253 for enabling presentation of information via one or more output devices 231 (e.g., display screen, speaker, etc.) associated with the user interface 230 (e.g., a user interface for operating peripherals and displaying content and information);

[0069] The input processing module 254 is used to detect one or more user inputs or interactions from one of the one or more input devices 232 and translate the detected inputs or interactions.

[0070] In some embodiments, the device provided in the embodiments of the present application can be implemented in software. Figure 2 A virtual resource processing device 255 stored in the memory 250 is shown, which can be software in the form of programs and plug-ins, including the following software modules: a construction module 2551, a vectorization module 2552, a feature processing module 2553 and a prediction module 2554. These modules are logical and can therefore be arbitrarily combined or further split according to the functions implemented.

[0071] The functions of each module will be described below.

[0072] In other embodiments, the device provided in the embodiments of the present application can be implemented in hardware. As an example, the device provided in the embodiments of the present application can be a processor in the form of a hardware decoding processor, which is programmed to execute the virtual resource processing method provided in the embodiments of the present application. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs) or other electronic components.

[0073] In some embodiments, the terminal or server can implement the virtual resource processing method provided in the embodiment of the present application by running a computer program. For example, the computer program can be a native program or software module in the operating system; it can be a native application (APP, Application), that is, a program that needs to be installed in the operating system to run, such as a game APP or a social platform APP; it can also be a small program, that is, a program that can be run only by downloading it to a browser environment; it can also be a small program or a web client program that can be embedded in any APP. In short, the above-mentioned computer program can be an application, module or plug-in in any form.

[0074] The virtual resource processing method provided in the embodiment of the present application will be explained in combination with the exemplary application and implementation of the electronic device provided in the embodiment of the present application.

[0075] See also Figure 3 , Figure 3 is an optional flow chart of the virtual resource processing method provided in the embodiment of the present application, which will be combined with Figure 3 The steps shown are explained.

[0076] S101: construct a relationship graph network based on at least one type of social relationship data in a virtual scene to obtain at least one social graph network.

[0077] The virtual resource processing method provided in the embodiments of the present application can be applied to scenarios in which users' virtual resource demands are predicted in virtual scene applications. For example, the method can be applied to scenarios in which players' demands for virtual resources such as equipment, props, and redeemable resources in online games are predicted, or in scenarios in which users' demands for virtual resources such as virtual gifts and virtual content resources in social applications are predicted, etc. The specific method can be selected based on actual conditions and is not limited by the embodiments of the present application.

[0078] In an embodiment of the present application, the electronic device obtains at least one type of social relationship data in a virtual scene, and constructs a graph neural network based on the at least one type of social relationship data to obtain at least one social graph network.

[0079] In the embodiment of the present application, the virtual scene may include a virtual scene in which multiple accounts interact. For example, the virtual scene may be a game scene, in which game players may engage in a series of social behaviors such as making friends, becoming apprentices, and teamwork in the game. Here, the electronic device may obtain accounts that have social behaviors within a preset time period, and obtain corresponding at least one type of social relationship data according to at least one type of social behavior.

[0080] In the embodiment of the present application, each type of social relationship data may include the social relationship and interaction degree between accounts corresponding to the type of social behavior. For example, the interaction relationship may be the social relationship corresponding to social behaviors such as likes or gift giving, and the interaction degree may include the number of likes, the number of gift giving and the gift level, etc. The specific selection is based on the actual situation and is not limited in the embodiment of the present application.

[0081] In an embodiment of the present application, the electronic device can construct a relationship graph network based on at least one type of social relationship data obtained, wherein, for each type of social relationship data, a social graph network corresponding to the type of social relationship data is constructed, thereby obtaining at least one social graph network.

[0082] In an embodiment of the present application, for each type of social relationship data, the electronic device uses the accounts therein as nodes, and connects the nodes where such social relationships exist based on the social relationships between the accounts contained therein to construct a relationship graph network, thereby obtaining a graph neural network corresponding to each type of social relationship data, with accounts as nodes and social relationships between accounts as edges, as a social graph network corresponding to each type of social relationship data.

[0083] In some embodiments, the electronic device can construct a graph neural network Line as a social graph network based on social relationship data, or other types of graph neural networks. The specific selection is based on actual conditions and is not limited in the embodiments of the present application.

[0084] It can be understood that by using graph neural networks to represent the social network between accounts, the graph structure can be used to represent the relationship between nodes, thereby improving the accuracy of the network representation of unstructured information.

[0085] S102: Perform vectorization processing on at least one social graph network to obtain at least one vectorization result.

[0086] In an embodiment of the present application, the electronic device performs vectorization processing on each social graph network in at least one social graph network to obtain a vector representation (Embedding) of each node in each social graph network as the vectorization result of each social graph network, thereby obtaining at least one vectorization result.

[0087] In the embodiment of the present application, the process of the electronic device performing vectorization processing on each social graph network is a process of converting the graphic data corresponding to the social graph network into a feature dimension space and retaining the graphic structure information and graphic attributes therein.

[0088] In the embodiment of the present application, the goal of the electronic device to perform vector processing is to encode or project the nodes in the social graph network into the latent space, and obtain a feature vector that can express the structural information of the original network as a vector representation of the node. Here, the geometric relationship in the latent space corresponds to the interaction relationship in the social graph network (such as the edge between nodes).

[0089] In some embodiments, the electronic device performs vectorization processing, and the vector representation of each node obtained may contain structural information of the social graph network and local similarity information of adjacent nodes.

[0090] In some embodiments, the electronic device can obtain the node similarity in the original space based on the connection relationship of the nodes in the social graph network, and the electronic device can obtain the vector representation of the node by iterative fitting with the goal of making the node similarity in the original space consistent with the node similarity in the vectorized feature expression space. Then, the downstream task can be executed according to the vector representation of the node, that is, the virtual resource demand prediction.

[0091] In some embodiments, the electronic device may use the Line first-order similarity algorithm to obtain a vector representation of each node, such as Figure 4 As shown, it is the vectorization result of the social graph network; Random Walk, Graphsage and other algorithms can also be used to vectorize the social graph network to obtain the vectorization result of the social graph network. The specific selection is made according to the actual situation, and the embodiments of this application are not limited.

[0092] S103: For each account included in at least one type of social relationship data, extract virtual resource operation data corresponding to each account, and generate resource operation features corresponding to each account based on the virtual resource operation data.

[0093] In an embodiment of the present application, for each account included in at least one type of social relationship data, the electronic device can obtain the data generated by each account operating virtual resources in a virtual scene as virtual resource operation data. The electronic device performs characterization processing on the virtual resource operation data to obtain resource operation characteristics corresponding to each account.

[0094] In some embodiments, the electronic device may obtain at least one operation data corresponding to at least one virtual resource operation behavior according to at least one virtual resource operation behavior of each account in a virtual scene as virtual resource operation data. Exemplarily, the virtual resource operation data may include data generated by virtual resource operation behaviors such as acquisition or consumption of virtual resources by each account in a virtual scene. Exemplarily, it may include data on the number of virtual resource recharges, virtual resource balances, and resource consumption quantities generated by game players performing operations such as recharge, exchange, or purchase of virtual items in the game, etc. The specific selection is made according to the actual situation, and the embodiments of the present application are not limited thereto.

[0095] In an embodiment of the present application, the electronic device can extract virtual resource operation data corresponding to an account from at least one dimension, and convert the dimension used to extract the virtual resource operation data into a feature dimension, thereby mapping the virtual resource operation data of at least one dimension into a feature vector containing at least one feature dimension to obtain a resource operation feature.

[0096] For example, the electronic device can extract multiple data such as the amount and number of resource consumption, the amount and number of virtual resource recharges, and the remaining amount of virtual resources of each account within a preset time period from dimensions such as resource consumption frequency, resource recharge frequency, and remaining amount of virtual resources as virtual resource operation data. The electronic device maps each of the data to a dimension in the multidimensional feature to obtain a multidimensional feature vector as a resource operation feature.

[0097] S104: Predict virtual resource demand based on resource operation characteristics and at least one vectorization result to obtain the virtual resource demand quantity of each account.

[0098] In an embodiment of the present application, the electronic device can predict the demand for virtual resources of each account based on the resource operation characteristics generated at the resource operation behavior level and at least one vectorized result generated at the social behavior level, using a neural network combined with the above two types of data to obtain the virtual resource demand quantity of each account.

[0099] In some embodiments, the electronic device may perform feature fusion of the resource operation feature with at least one vectorization result, and use a neural network to predict the feature fusion result to obtain the virtual resource demand quantity of each account.

[0100] For example, feature fusion can be achieved by summing the resource operation feature and at least one vectorized result through a Sum Pooling network layer, and the neural network can use a neural network perceptron to perform nonlinear calculations on the summed vector to increase the nonlinearity of the model. As shown in formula (1):

[0101] y=relu(relu(XW1+bias1)W2+bias2) (1)

[0102] In formula (1), X is the feature fusion result after Sum pooling, and W1, W2, bias1, and bias2 are all network parameters trained in the neural network perceptron.

[0103] It can be understood that by constructing a social graph network and vectorizing the social relationship data between various accounts in the virtual scene, social relationship information is introduced into the virtual resource demand prediction task, and the virtual resource demand is predicted by using the vectorization results of the social relationship data and combining the account's operation behavior on the virtual resources. Since in actual applications, the account's demand for virtual resources will be affected by social relationship factors, the embodiment of the present application combines quantified social relationships to predict the virtual resource demand of each account, which can greatly improve the accuracy of virtual resource demand prediction.

[0104] In some embodiments, based on Figure 3 ,like Figure 5 As shown, S102 can be implemented through S1021-S1025, which will be explained in combination with each step.

[0105] S1021. For each social graph network in at least one social graph network, obtain an initial weight of each connection relationship in each social graph network; the initial weight represents the degree of interaction between accounts corresponding to each connection relationship.

[0106] In an embodiment of the present application, the electronic device can vectorize each social graph network in at least one social graph network. For each social graph network, the electronic device obtains the initial weight of each connection relationship, that is, each edge. Here, the initial weight represents the degree of social interaction between the two nodes corresponding to each connection relationship, that is, the two accounts.

[0107] In some embodiments, at least one type of social relationship data may include: friendship relationship data; when a social graph network is constructed based on friendship relationship data, the electronic device can obtain the accounts corresponding to each connection relationship in the social graph network, that is, the experience value of the friendship relationship between the accounts that form the friendship, such as the length of friendship, degree of interaction, etc., as the initial weight of each connection relationship.

[0108] In some embodiments, at least one type of social relationship may include: activity relationship data. When a social graph network is constructed based on activity relationship data, the electronic device can obtain the number of times the accounts corresponding to each connection have jointly participated in virtual scene activities in the virtual scene, such as the number of times they have jointly teamed up to fight, jointly participated in guild battles or group battles, as the initial weight of each connection relationship.

[0109] In some embodiments, at least one type of social relationship may include: organizational relationship data. When the social graph network is constructed based on organizational relationship data, the electronic device may obtain an initial weight based on the organizational information in which each account corresponding to the connection participates. Exemplarily, the organizational relationship data may be the master-apprentice relationship data in the game scene. When the two accounts corresponding to each connection relationship have established a master-apprentice relationship, the electronic device may obtain an initial weight of 1 for the connection relationship. When the master-apprentice relationship has not been established, the initial weight of the connection relationship may be 0. Alternatively, the organizational relationship data may also be the sects joined by the accounts corresponding to each connection relationship in the game. When the two accounts corresponding to the connection relationship join the same organization, such as the same sect, the initial weight of the connection relationship may be 1, otherwise it may be 0. Of course, it may also be set to other weight values ​​according to the actual situation. The specific selection is made according to the actual situation, and the embodiments of the present application are not limited thereto.

[0110] In some embodiments, at least one type of social relationship may include: intimacy data. When a social graph network is constructed based on intimacy data, the electronic device may obtain the intimacy value between the accounts corresponding to each connection as the initial weight of each connection.

[0111] S1022. Calculate the initial weight distribution corresponding to each connection relationship according to the initial weight of each connection relationship.

[0112] In the embodiment of the present application, for each social graph network, the electronic device can calculate the numerical distribution of the initial weight on each connection relationship based on the initial weight of each connection relationship therein and the initial weight in the entire social graph network as the initial weight distribution corresponding to each connection relationship.

[0113] In some embodiments, the electronic device may calculate the sum of weights of each social graph network based on the initial weight of each connection relationship; and calculate the ratio of the initial weight of each connection relationship to the sum of weights as the initial weight distribution corresponding to each connection.

[0114] In an exemplary embodiment, a social graph network can be represented as G=(V, E, W), where V represents a node, i.e., an account, E represents an edge, i.e., a connection relationship, and W represents an initial weight. i ,v j )∈E, the empirical probability distribution of the connection relationship before vectorization (before embedding) can be calculated by formula (2) and formula (3) as the initial weight distribution of the connection relationship, as follows:

[0115]

[0116]

[0117] In formula (2), w ij For the connection relationship (v i ,v j ), W is the weight sum obtained by summing the initial weights of each connection relationship in the social graph network through formula (3). The electronic device calculates the connection relationship (v i ,v j )’s initial weight to the total weight, as the connection relationship (v i ,v j ) corresponds to the initial weight distribution.

[0118] S1023. Perform vector initialization on each node in each social graph network to obtain an initial vectorized representation of each node.

[0119] In the embodiment of the present application, for each social graph network to be vectorized, the electronic device performs vector initialization on each node therein to obtain an initial vectorized representation of each node.

[0120] In some embodiments, the electronic device may use a random value to initialize the vector of each node, or may initialize the vector according to a preset value. The specific selection is made based on actual conditions and is not limited in the embodiments of the present application.

[0121] S1024: Perform weight distribution prediction based on the initial vectorized representation of each node to obtain a predicted weight distribution corresponding to each connection relationship.

[0122] In an embodiment of the present application, the initial vectorized representation of each node represents the initial value of each node after vectorization. The electronic device can predict the weight distribution after vectorization (after embedding) based on the initial vectorized representation of each node to obtain the predicted weight distribution corresponding to each connection relationship.

[0123] In some embodiments, node v i With node v j The corresponding initial vectorized representation can be u i ,u j ∈R d , the electronic device can use a trained neural network, such as a multi-layer perceptron, to perform reasoning based on the initialization vector representation of the nodes connected to each connection relationship, predict the weight distribution of each connection relationship after vectorization, and obtain the predicted weight distribution corresponding to each connection relationship. For example, as shown in formula (4), as follows:

[0124] p1(v i ,v j )=D(u i ,u j ) (4)

[0125] In formula (4), p1(v i ,v j ) is the connection relationship (v i ,v j ) The corresponding prediction weight distribution D can be a neural network model using the LINE first-order algorithm, as shown in formula (5):

[0126] D(u,v)=σ(u·v) (5)

[0127] It can be seen that based on formula (4) and formula (5), formula (6) can be obtained as follows:

[0128]

[0129] It can be understood that for each connection relationship (v i ,v j ), the electronic device can connect to each node v i With v j The corresponding initial vector representation u i with u j , using the neural network D, the empirical probability distribution is predicted by the algorithm of formula (6) to obtain each connection relationship (v i ,v j ) corresponds to the predicted weight distribution.

[0130] S1025. Iteratively update the initial vectorized representation by fitting the initial weight distribution and the predicted probability distribution until a preset fitting condition is met, thereby obtaining a vectorized representation of each node as a vectorized result of each social graph network, thereby obtaining at least one vectorized result.

[0131] In an embodiment of the present application, the initial weight distribution represents the weight distribution before vectorization, and the predicted probability distribution represents the probability distribution after vectorization. In order to maintain the first-order similarity, the initial weight distribution and the predicted weight distribution should be as close as possible. Therefore, the electronic device can update the initial vectorized representation with the goal of fitting the initial weight and the predicted weight distribution to obtain an intermediate vectorized representation. The electronic device uses the intermediate vectorized representation to re-predict the predicted weight distribution corresponding to each connection relationship so that the predicted weight distribution gradually approaches the initial weight distribution. This iteration is performed until the preset fitting conditions are met, and the intermediate vectorized representation is used as the vectorized representation of each node.

[0132] In some embodiments, S1025 can be implemented by executing S201-S203, which will be described in conjunction with each step.

[0133] S201. Calculate the difference between the initial weight distribution and the predicted weight distribution using a preset target optimization function.

[0134] In the embodiment of the present application, the preset target optimization function is used to fit the initial weight distribution and the predicted weight distribution by calculating the difference between the initial weight distribution and the predicted weight distribution. Exemplarily, the difference between the initial weight distribution and the predicted weight distribution can be calculated by KL divergence (Kullback-Leibler divergence), as shown in formula (7), as follows:

[0135]

[0136] In formula (7), O is the difference, and C can be calculated by formula (8), as follows:

[0137]

[0138] It can be seen that based on formula (7), setting the optimization goal to reduce the difference O can fit the difference between the initial weight distribution and the predicted weight distribution. The calculated constant, therefore, formula (9) can achieve the same optimization goal as formula (7), as follows:

[0139]

[0140] Combined with formula (2) Compared with formula (9), since W in formula (2) is also a constant, the preset target optimization function can be obtained as shown in formula (10), as follows:

[0141]

[0142] It can be understood that the electronic device can calculate the difference O between the initial weight distribution and the predicted weight distribution through formula (10), and can achieve the optimization goal of fitting the initial weight distribution and the predicted weight distribution by optimizing the gradient of the difference O.

[0143] S202: When the difference does not satisfy a preset fitting condition, the initial vectorized representation of each node is updated to obtain an intermediate vectorized representation of each node.

[0144] In an embodiment of the present application, when the difference calculated according to the preset target optimization function does not meet the preset fitting conditions, it means that the similarity between the weight distribution before vectorization and the weight distribution after vectorization is not high. The electronic device updates the initial vectorization representation to obtain the intermediate vectorization representation of each node, so as to update the prediction of the weight distribution after vectorization through the intermediate vectorization, which will be explained in the following steps.

[0145] S203, predicting the weight distribution based on the intermediate vectorized representation of each node, updating the predicted weight distribution, and recalculating the difference based on the updated predicted weight distribution, until the difference meets the preset fitting condition, and using the intermediate vectorized representation of each node as the vectorized representation of each node.

[0146] In an embodiment of the present application, the electronic device uses the method in S1024 to predict the weight distribution based on the updated intermediate vectorized representation of each node, that is, the intermediate vectorized representation is used as the input of the neural network, the prediction result of the neural network is updated, and an updated predicted weight distribution is obtained.

[0147] In an embodiment of the present application, the electronic device can use a preset target optimization function to calculate the difference between the updated predicted weight distribution and the initial weight distribution, and update the value of the difference.

[0148] In some embodiments, for the preset target optimization function shown in formula (10), the electronic device may use the updated predicted weight distribution as p1(v i ,v j ), substitute it into formula (10) to calculate, get the value of O, and update the difference.

[0149] In an embodiment of the present application, when the difference does not meet the preset fitting condition, the electronic device continues to iterate the process in S202-S203 above, and gradually updates the intermediate vectorized representation until the difference meets the preset fitting condition, indicating that the similarity between the weight distribution before vectorization and the weight distribution after vectorization meets the expected requirements, and the electronic device uses the iterative intermediate vectorized representation as the vectorized representation.

[0150] In some embodiments, the preset fitting condition may be that the difference is less than a preset difference threshold, or the condition may be set according to the needs of the actual project. The specific selection is made according to the actual situation, and the embodiments of the present application are not limited thereto.

[0151] In some embodiments, based on Figure 3 or Figure 5 ,like Figure 6 As shown, S103 can be implemented by executing S1031-S1032, which will be explained in combination with each step.

[0152] S1031. For each account included in at least one social graph network, obtain at least one resource data of a virtual resource in each account in at least one quantity dimension within at least one preset time period as virtual resource operation data.

[0153] In an embodiment of the present application, for each account included in at least one social graph network, each account has its own corresponding virtual resources, that is, the virtual resources in each account. During the activity of the account in the virtual scene, operations such as expansion or consumption of virtual resources will occur, so the number of virtual resources may change in different time periods. The electronic device can obtain data corresponding to the virtual resource in each time period of at least one time period, and for each time period, the data is obtained from at least one quantity dimension to obtain at least one resource data of the virtual resource in at least one time period and at least one quantity dimension as virtual resource operation data.

[0154] Exemplarily, the preset time period may be one week, and at least one quantity dimension may include: a dimension of resource consumption and a dimension of resource remaining. The electronic device may obtain the quantity of virtual resources consumed by each account in the first week, the second week, the third week, and the fourth week, as well as the quantity of remaining resources in the first week, the second week, the third week, and the fourth week as virtual resource operation data.

[0155] S1032. Map at least one resource data to at least one feature dimension space to generate resource operation features.

[0156] In an embodiment of the present application, the electronic device can map at least one resource data to at least one feature dimension space, generate a feature dimension vector for each resource data, thereby converting at least one resource data into a multidimensional space vector and generating a resource operation feature.

[0157] Exemplarily, the at least one resource data may be 100 data, and the electronic device may map the 100 data into a 100-dimensional vector, where each data corresponds to one dimension of the vector.

[0158] In some embodiments, based on Figure 6 , after S1031. The electronic device can also obtain the activity data of each account to generate resource operation characteristics, such as Figure 7 As shown, each step will be described in combination.

[0159] S1033. Extract at least one kind of activity data from each account within at least one preset time period; the at least one kind of activity data is data generated by each account's activities in a virtual scene.

[0160] S1034. Perform feature dimension mapping based on at least one resource data and at least one activity data to generate resource operation features.

[0161] In an embodiment of the present application, the electronic device can extract at least one activity data generated by each account performing activities in a virtual scene within at least one preset time period. In some embodiments, at least one activity data may include data on social interactions of the account in the virtual scene, such as the number of friends in the game and the number of interactions with friends. Or it may also include activity status data of the account in the virtual scene, such as the online time and active time of the game, etc. The specific selection is made according to the actual situation, and the embodiment of the present application is not limited.

[0162] Exemplarily, the virtual scene may be a game scene, and at least one activity data may include: the number of friends in the game; the online time in the first week, the online time in the second week, the online time in the third week, the online time in the fourth week, the number of interactions with friends in the first week, the number of interactions with friends in the second week, the number of interactions with friends in the third week, and the number of interactions with friends in the fourth week, etc. The specific selection shall be made according to the actual situation and shall not be limited in the embodiments of the present application.

[0163] In the embodiment of the present application, the electronic device can perform feature dimension mapping on at least one resource data and at least one activity data synchronously in combination with at least one activity data, and generate a resource operation feature whose feature dimension corresponds to the sum of the quantity of at least one resource data and at least one activity data. Here, the electronic device performs feature dimension mapping based on at least one resource data and at least one activity data to generate a resource operation feature, which is consistent with the process description of S1032 and will not be repeated here.

[0164] It can be understood that by combining at least one activity data to generate resource operation features, more dimensional features can be introduced for prediction in subsequent virtual resource demand prediction tasks, thereby improving the accuracy of virtual resource demand prediction.

[0165] In some embodiments, based on Figure 7 ,like Figure 8 As shown, S104 can be implemented by executing S1041-S1042, which will be explained in combination with each step.

[0166] S1041. Obtain associated resource data of each account in an associated virtual scene, and obtain associated resource features based on the associated resource data; the associated resource data is data corresponding to virtual resource operations performed by each account in the associated virtual scene.

[0167] In the embodiment of the present application, there may be accounts in the current virtual scene, and the electronic device can obtain the data generated by each account performing virtual resource operations in the associated virtual scene as the associated resource data. Here, the associated virtual scene can be another virtual scene that has an associated relationship with the virtual scene, such as a data sharing relationship; the electronic device can obtain the data of each account in the current virtual scene performing virtual resource operations in the associated virtual scene as the associated resource data based on the associated relationship.

[0168] For example, an account can log in to or use multiple games. For related games that have a data sharing relationship, the electronic device can obtain data corresponding to the account's operations on virtual resources in other related games, such as data on the account in game A acquiring and consuming virtual resources in game B, as related resource data.

[0169] In some embodiments, the electronic device can acquire the associated resource data from at least one quantity dimension according to at least one time period in the same way as acquiring the virtual resource operation data. For example, the electronic device can acquire the maximum number of virtual resource acquisitions per day in the past year, the minimum number of virtual resource acquisitions per day in the past year, the total number of virtual resource acquisitions in the past year, etc. for each account in the associated virtual scene as the associated resource data.

[0170] In the embodiment of the present application, the electronic device can map the associated resource data to the feature dimension to obtain the associated resource feature. Here, the process of mapping the feature dimension according to the associated resource data is consistent with the description in S1032, and will not be repeated here.

[0171] S1042: Predict virtual resource demand based on associated resource characteristics, resource operation characteristics and at least one vectorization result to obtain the virtual resource demand quantity of each account.

[0172] In an embodiment of the present application, the electronic device can use a neural network to predict virtual resource demand and output the virtual resource demand quantity of each account based on the associated resource characteristics and resource operation characteristics that respectively characterize the virtual resource operation data in the user virtual scene and the associated virtual scene, and at least one vectorized result that characterizes the social relationship data of the account.

[0173] In some embodiments, based on Figure 8 ,like Fig. 9 As shown, S1042 can be implemented by executing S301-S302, which will be explained in combination with each step.

[0174] S301: perform feature fusion on associated resource features, resource operation features and at least one vectorization result to obtain a fusion result.

[0175] S302: Predict virtual resource requirements based on the fusion results to obtain the virtual resource requirements of each account.

[0176] In the embodiment of the present application, the electronic device performs feature fusion on the associated resource feature, the resource operation feature and the at least one vectorized result, and illustratively, sums them through a Sum Pooling layer to obtain a fusion result. The electronic device can use a neural network sensor as shown in formula (1) to predict the virtual resource demand of each account based on the fusion result.

[0177] It can be understood that in the embodiment of the present application, by combining the associated resource characteristics, resource operation characteristics and at least one vectorized result to predict the virtual resource demand, multiple factors affecting the account's virtual resource demand are simultaneously introduced, thereby improving the accuracy of virtual resource prediction.

[0178] Next, we will combine Fig.10 , illustrating an exemplary application of an embodiment of the present application in a practical application scenario.

[0179] The embodiment of the present application can be applied to the scenario of predicting the virtual resource demand of players in online game A. Here, online game A corresponds to the above virtual scenario, and the players correspond to the above accounts. Exemplarily, the social behaviors in online game A can include the following four types: making friends social behavior, guild war social behavior, master-apprentice relationship social behavior, and intimacy social behavior.

[0180] S01. Build friendship networks, guild battle networks, mentor-apprentice networks, and intimacy networks.

[0181] In S01, the electronic device obtains social relationship data corresponding to each category of social behavior, and forms a friendship network, a guild war network, a master-apprentice network and an intimacy network according to each category of social relationship data, as at least one social graph network.

[0182] Among them, the electronic device can select players who have established friendships in the game within a preset time period, use the player ID as a node, the friendships between the players as an edge, and the experience value of the friendships reported in the game log as the weight of the edge to construct a friendship network; select players who form a team together to participate in a guild war within a preset time period, use the player ID as a node, the relationship between the players who have participated in the guild war team together as an edge, and the number of times the two nodes have participated in the guild war together as the weight of the edge to construct a guild war network; select players who have established a master-apprentice relationship in the game within a period of time, use the player ID as a node, and the master-apprentice relationship that has been established between the two as an edge, if the master-apprentice relationship exists, the weight of the edge is 1, and if the master-apprentice relationship does not exist, the weight becomes 0, to construct a master-apprentice network; select players who have jointly established and improved intimacy in the game within a period of time, use the player ID as a node, the behavior of improving the intimacy of friends between the nodes as an edge, and the game intimacy value reported in the game log as the weight of the edge to construct an intimacy network.

[0183] S02. Obtain market data.

[0184] In S02, for each player included in the above four types of social relationship data, the electronic device obtains the operation data of each player on virtual resources in other games, such as online game B and online game C, which may include: the maximum payment amount per day within a year, the minimum payment amount per day within a year, the maximum payment amount per week within a year, the minimum payment amount per week within a year, the total payment amount within a year, the number of payment days within a year, the number of payment weeks within a year, the maximum payment amount per day for a single game within a year, the number of payment days for a single game within a year, the payment amount for a single game within a year, the mode of the payment interval per day within a year, the mode of the payment interval per day for a single game within a year, the average number of payment days per week for a single game within a year, the number of days between the last payment date of a single game within a year, the maximum number of consecutive payment days for a single game within a year, the payment amount for a single game within a month, etc., as the market data. Here, the market data is equivalent to the above-mentioned associated resource data.

[0185] S03. Obtain in-game data.

[0186] In S03, the electronic device obtains the number of friends of each player in the online game A, the online time in the first week (in hours), the online time in the second week, the online time in the third week, the online time in the fourth week, the number of interactions with friends in the first week, the number of interactions with friends in the second week, the number of interactions with friends in the third week, the number of interactions with friends in the fourth week, the number of virtual resources consumed in the first week, the number of virtual resources consumed in the second week, the number of virtual resources consumed in the third week, the number of virtual resources consumed in the fourth week, the remaining number of virtual resources in the first week, the remaining number of virtual resources in the second week, the remaining number of virtual resources in the third week, and the remaining number of virtual resources in the fourth week, etc., as in-game data. Here, the in-game data is equivalent to virtual resource operation data.

[0187] S04. Vectorize the social network to obtain a vectorized result.

[0188] In S04, the electronic device uses the first-order algorithm of the graph network Line to vectorize the friendship network, guild war network, mentor-apprentice network and intimacy network respectively, and obtains the vectorization results of the above four social networks respectively.

[0189] S05. Map feature dimensions based on the market data and the in-game data to obtain market features and in-game features.

[0190] In S05, the electronic device performs feature dimension mapping on the large-scale data and the in-game data respectively, and obtains a multi-dimensional space vector corresponding to the large-scale data as a large-scale feature, and a multi-dimensional space vector corresponding to the in-game data as an in-game feature.

[0191] S06. Feature fusion.

[0192] In S06, the electronic device sums the input disk features, in-game features, and Line first-order vectorization results of the four social networks through Sum pooling to obtain a feature fusion result.

[0193] S07. Virtual resource demand forecasting.

[0194] In S07, the electronic device uses the MLP network to predict the virtual resource demand based on the feature fusion result to obtain the virtual resource demand quantity of each account.

[0195] The applicant compared the virtual resource processing method provided in the embodiment of the present application with the method for predicting the number of virtual resource requirements in the related art, and the results are as follows:

[0196] Using the business rules of the relevant technology, the demand for virtual resources is estimated based on the consumption of virtual resources. The cumulative virtual resource consumption in the game for nearly a year is arranged in descending order, and the corresponding top 1,000 accounts are selected as the first account group. Using the virtual resource processing method of the embodiment of the present application, the predicted number of virtual resource demands is arranged in descending order, and the top 1,000 accounts are selected as the second account group. It can be understood that the first account group is a high-value account screened out by the relevant technology, and the second account group is a high-value account screened out by the method of the embodiment of the present application. The applicant continued to count the actual number of virtual resource demands generated by the first account group and the second account group in subsequent games, and the comparison results are as follows:

[0197] The actual demand for virtual resources generated by the second account group increased by 60% compared with the first account group, and the per capita demand for virtual resources in the second account group increased by 47% compared with the per capita demand for virtual resources in the first account group. In addition, the per capita demand for virtual resources in the second account group accounted for 87% of the per capita demand for virtual resources in the game accounts, while the per capita demand for virtual resources in the first account group accounted for only 44% of the per capita demand for virtual resources in the game accounts during the same period.

[0198] It can be seen that the embodiment of the present application vectorizes the social behavior in the game through Line, combines the user's game characteristics and in-game characteristics, and uses neural networks to predict the number of virtual resource requirements of players in the game in the future. In this way, the social relationship in the game is innovatively abstracted into a graph network, and applied to the high-value game player mining scenario. The frontier theory of graph network Line is used to solve the social relationship quantification problem in the first order, and the social relationship ability in the game is converted into virtual resource demand, thereby reducing the subjective blindness of using experience to judge which players in the game will become high-value game players, and improving the accuracy of the prediction of the number of virtual resource requirements. When applied to the operation of virtual scenes such as games, it can greatly improve the conversion rate of high-value users and the income of virtual resources.

[0199] The following is a description of an exemplary structure of the virtual resource processing device 255 provided in the embodiment of the present application implemented as a software module. In some embodiments, Figure 2 As shown, the software modules stored in the virtual resource processing device 255 of the memory 250 may include:

[0200] A construction module 2551 is used to construct a relationship graph network based on at least one type of social relationship data in the virtual scene to obtain at least one social graph network;

[0201] A vectorization module 2552, configured to perform vectorization processing on the at least one social graph network to obtain the at least one vectorization result;

[0202] A feature processing module 2553 is used to extract virtual resource operation data corresponding to each account included in the at least one social graph network, and generate resource operation features corresponding to each account based on the virtual resource operation data;

[0203] The prediction module 2554 is used to predict the virtual resource demand based on the resource operation characteristics and the at least one vectorization result to obtain the virtual resource demand quantity of each account.

[0204] In some embodiments, the construction module 2551 is also used to obtain, for each type of social relationship data in the at least one type of social relationship data, the accounts in the each type of social relationship data as nodes, and obtain the social graph network corresponding to the each type of social relationship data by connecting the nodes where the social relationship exists, thereby obtaining the at least one social graph network.

[0205] In some embodiments, the vectorization module 2552 is also used to obtain the initial weight of each connection relationship in each social graph network in the at least one social graph network; the initial weight represents the degree of interaction between the accounts corresponding to each connection relationship; based on the initial weight of each connection relationship, the initial weight distribution corresponding to each connection relationship is calculated; vector initialization is performed on each node in each social graph network to obtain the initial vectorized representation of each node; weight distribution prediction is performed based on the initial vectorized representation of each node to obtain the predicted weight distribution corresponding to each connection relationship; the initial vectorized representation is iteratively updated by fitting the initial weight distribution with the predicted weight distribution until the preset fitting condition is met, and the vectorized representation of each node is obtained as the vectorization result of each social graph network, thereby obtaining the at least one vectorization result.

[0206] In some embodiments, the at least one type of social relationship data includes at least one of the following: friendship relationship data, activity relationship data, organizational relationship data and intimacy data; the vectorization module 2552 is also used to obtain the friendship experience value between the accounts corresponding to each connection relationship as the initial weight when each social graph network is constructed based on the friendship relationship data; obtain the number of times the accounts corresponding to each connection relationship jointly participate in virtual scene activities as the initial weight when each social graph network is constructed based on the activity relationship data; obtain the initial weight according to the organization information in which the accounts corresponding to each connection relationship participate when each social graph network is constructed based on the organizational relationship data; obtain the intimacy value between the accounts corresponding to each connection relationship as the initial weight when each social graph network is constructed based on the intimacy relationship data.

[0207] In some embodiments, the vectorization module 2552 is further used to calculate the sum of the weights of each social graph network according to the initial weight of each connection relationship; and calculate the ratio of the initial weight of each connection relationship to the sum of the weights as the initial weight distribution corresponding to each connection relationship.

[0208] In some embodiments, the vectorization module 2552 is also used to calculate the difference between the initial weight distribution and the predicted weight distribution using a preset target optimization function; when the difference does not meet the preset fitting condition, the initial vectorization representation of each node is updated to obtain the intermediate vectorization representation of each node; weight distribution prediction is performed based on the intermediate vectorization representation of each node, the predicted weight distribution is updated, and the difference is updated based on the updated predicted weight distribution and the initial weight distribution, until the difference meets the preset fitting condition, and the intermediate vectorization representation of each node is used as the vectorization representation of each node.

[0209] In some embodiments, the feature processing module 2553 is further used to obtain at least one resource quantity data of the virtual resources in each account in at least one quantity dimension within at least one preset time period as the virtual resource operation data.

[0210] In some embodiments, the feature processing module 2553 is also used to generate the resource operation feature corresponding to the at least one feature dimension by mapping the at least one quantity dimension to at least one feature dimension, based on mapping the at least one resource quantity data to at least one feature dimension space.

[0211] In some embodiments, the feature processing module 2553 is also used to extract at least one activity data of each account within at least one preset time period in the virtual scene; the at least one activity data is the data generated by the activities of each account in the virtual scene; and the resource operation feature is generated by performing feature dimension mapping between the at least one resource quantity data and the at least one activity data.

[0212] In some embodiments, the prediction module 2554 is also used to obtain the associated resource data of each account in the associated virtual scene, and obtain associated resource characteristics based on the associated resource data; the associated resource data is the data corresponding to the virtual resource operation performed by each account in the associated virtual scene; based on the associated resource characteristics, the resource operation characteristics and the at least one vectorization result are used to predict the virtual resource demand to obtain the virtual resource demand quantity of each account.

[0213] In some embodiments, the prediction module 2554 is further used to perform feature fusion on the associated resource characteristics, the resource operation characteristics and the at least one vectorization result to obtain a fusion result; and to predict the virtual resource demand for the fusion result, the virtual resource demand quantity for each account.

[0214] It should be noted that the description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the device embodiment of the present application, please refer to the description of the method embodiment of the present application for understanding.

[0215] The present application embodiment provides a computer-readable storage medium storing executable instructions, wherein the executable instructions are stored. When the executable instructions are executed by a processor, the processor will be caused to execute the method provided by the present application embodiment, for example, Figure 3-Figure 10 The method shown in .

[0216] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface storage, optical disk, or CD-ROM; or it may be various devices including one or any combination of the above memories.

[0217] In some embodiments, executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0218] As an example, executable instructions may, but need not, correspond to a file in a file system, may be stored as part of a file that stores other programs or data, such as in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions).

[0219] By way of example, executable instructions may be deployed to be executed on one computing device, or on multiple computing devices located at one site, or on multiple computing devices distributed across multiple sites and interconnected by a communication network.

[0220] In summary, by using the social relationship data between each account in the virtual scene, a social graph network is constructed and vectorized, so that social relationship information is introduced into the task of predicting virtual resource demand, and the virtual resource demand is predicted by using the vectorization results of social relationship data and combining the account's operation behavior on virtual resources. Since in actual applications, the account's demand for virtual resources will be affected by social relationship factors, the embodiment of the present application combines quantified social relationships to predict the virtual resource demand of each account, which can greatly improve the accuracy of virtual resource demand prediction.

[0221] The above is only an embodiment of the present application and is not intended to limit the protection scope of the present application. Any modifications, equivalent substitutions and improvements made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. A virtual resource processing method, characterized in that: include: Building a relationship graph network based on at least one type of social relationship data in the virtual scene to obtain at least one social graph network; The virtual scene is a game scene; For each social graph network, obtaining an initial weight of each connection relationship in each social graph network; The initial weight represents the degree of interaction between the accounts corresponding to each connection relationship; Calculating the initial weight distribution corresponding to each connection relationship according to the initial weight of each connection relationship; Performing vector initialization on each node in each social graph network to obtain an initial vectorized representation of each node; Performing weight distribution prediction based on the initial vectorized representation of each node to obtain a predicted weight distribution corresponding to each connection relationship; By fitting the initial weight distribution and the predicted weight distribution, the initial vectorized representation is iteratively updated until a preset fitting condition is satisfied, and a vectorized representation of each node is obtained as a vectorized result of each social graph network, thereby obtaining at least one vectorized result; For each account included in the at least one social graph network, extracting virtual resource operation data corresponding to each account, and generating resource operation features corresponding to each account based on the virtual resource operation data; Acquire the associated resource data of each account in the associated virtual scene, and obtain the associated resource characteristics based on the associated resource data; the associated resource data is data corresponding to the virtual resource operation performed by each account in the associated virtual scene; the associated virtual scene is other game scenes of the associated game that have a data sharing relationship with the virtual scene; The virtual resource demand is predicted according to the associated resource characteristics, the resource operation characteristics and the at least one vectorization result to obtain the virtual resource demand quantity of each account.

2. The method according to claim 1, characterized in that The step of constructing a relationship graph network based on at least one type of social relationship data in the virtual scene to obtain at least one social graph network includes: For each type of social relationship data in the at least one type of social relationship data, taking the accounts in the each type of social relationship data as nodes, the social graph network corresponding to the each type of social relationship data is obtained by connecting the nodes with this type of social relationship, thereby obtaining the at least one social graph network.

3. The method according to claim 2, characterized in that The at least one type of social relationship data includes at least one of the following: friendship relationship data, activity relationship data, organization relationship data, and close relationship data; The obtaining of the initial weight of each connection relationship in each social graph network includes: In the case where each social graph network is constructed based on the friendship relationship data, obtaining friendship experience values ​​between accounts corresponding to each connection relationship as the initial weight; In the case where each social graph network is constructed based on the activity relationship data, obtaining the number of times that the accounts corresponding to each connection relationship jointly participate in the virtual scene activity as the initial weight; In the case where each social graph network is constructed based on the organizational relationship data, the initial weight is obtained according to the organizational information in which the account corresponding to each connection relationship participates; In the case where each social graph network is constructed based on the intimacy data, an intimacy value between the accounts corresponding to each connection relationship is obtained as the initial weight.

4. The method according to claim 1, characterized in that The calculating, according to the initial weight of each connection relationship, the initial weight distribution corresponding to each connection relationship comprises: Calculating the sum of weights of each social graph network according to the initial weight of each connection relationship; A ratio of the initial weight of each connection relationship to the total weight is calculated as the initial weight distribution corresponding to each connection relationship.

5. The method according to claim 1, characterized in that The step of iteratively updating the initial vectorized representation by fitting the initial weight distribution and the predicted weight distribution until a preset fitting condition is satisfied to obtain the vectorized representation of each node comprises: Calculating the difference between the initial weight distribution and the predicted weight distribution using a preset target optimization function; When the difference does not satisfy the preset fitting condition, updating the initial vectorized representation of each node to obtain an intermediate vectorized representation of each node; A weight distribution prediction is performed based on the intermediate vectorized representation of each node, the predicted weight distribution is updated, and the difference is updated based on the updated predicted weight distribution and the initial weight distribution, until the difference satisfies a preset fitting condition, and the intermediate vectorized representation of each node is used as the vectorized representation of each node.

6. The method according to any one of claims 1 to 5, characterized in that: The extracting the virtual resource operation data corresponding to each account includes: At least one resource data of the virtual resources in each account in at least one quantity dimension within at least one preset time period is obtained as the virtual resource operation data.

7. The method according to claim 6, characterized in that The generating the resource operation feature corresponding to each account based on the virtual resource operation data includes: The at least one resource data is mapped to at least one feature dimension space to generate the resource operation feature.

8. The method according to claim 6, characterized in that The method further comprises: Extracting at least one kind of activity data of each account within at least one preset time period; the at least one kind of activity data is data generated by each account in the virtual scene; The resource operation feature is generated by performing feature dimension mapping between the at least one resource data and the at least one activity data.

9. The method according to claim 1, characterized in that: The predicting of virtual resource demand according to the associated resource characteristics, the resource operation characteristics and the at least one vectorization result to obtain the virtual resource demand quantity of each account includes: Performing feature fusion on the associated resource feature, the resource operation feature and the at least one vectorization result to obtain a fusion result; The virtual resource demand is predicted based on the fusion result, and the virtual resource demand quantity of each account is predicted.

10. A virtual resource processing device, characterized in that: include: A construction module, configured to construct a relationship graph network based on at least one type of social relationship data in a virtual scene, and obtain at least one social graph network; The virtual scene is a game scene; A vectorization module, used for obtaining, for each social graph network, an initial weight of each connection relationship in the social graph network; The initial weight represents the degree of interaction between the accounts corresponding to each connection relationship; According to the initial weight of each connection relationship, the initial weight distribution corresponding to each connection relationship is calculated; each node in each social graph network is vectorized and initialized to obtain the initial vectorized representation of each node; weight distribution is predicted based on the initial vectorized representation of each node to obtain the predicted weight distribution corresponding to each connection relationship; the initial vectorized representation is iteratively updated by fitting the initial weight distribution and the predicted weight distribution until a preset fitting condition is met, and the vectorized representation of each node is obtained as the vectorized result of each social graph network, thereby obtaining at least one vectorized result; a feature processing module, configured to extract, for each account included in the at least one social graph network, virtual resource operation data corresponding to each account, and generate resource operation features corresponding to each account based on the virtual resource operation data; A prediction module is used to obtain the associated resource data of each account in the associated virtual scene, and obtain associated resource characteristics based on the associated resource data; the associated resource data is the data corresponding to the virtual resource operation performed by each account in the associated virtual scene; the associated virtual scene is other game scenes of the associated game that have a data sharing relationship with the virtual scene; the virtual resource demand is predicted based on the associated resource characteristics, the resource operation characteristics and the at least one vectorization result to obtain the virtual resource demand quantity of each account.

11. An electronic device, characterized in that: include: A memory for storing executable instructions; A processor, configured to implement the method according to any one of claims 1 to 9 when executing the executable instructions stored in the memory.

12. A computer-readable storage medium, characterized in that: Executable instructions are stored, and when executed by a processor, the method described in any one of claims 1 to 9 is implemented.

13. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the method according to any one of claims 1 to 9 is implemented.