Information recommendation method, device, and electronic equipment of game program, and storage medium

Abstract

This application provides a method, apparatus, electronic device, and storage medium for information recommendation in a game program. The method includes: determining a first difficulty coefficient of a game level based on first player profile features and a first mapping rule; responding to a trigger operation for the game level, running the game level based on the first difficulty coefficient and acquiring player behavior data during the game level's execution; updating the first player profile features based on the player behavior data to obtain a second player profile feature; updating the first difficulty coefficient based on the second player profile features to obtain a second difficulty coefficient, and running the game level with the second difficulty coefficient; and responding to the end of the game level, predicting a target advertisement based on the game level's result, game level progress, and the second player profile features, wherein the target advertisement is intended to be displayed at the end of the game level. This application can improve the efficiency and accuracy of information recommendation in game programs.

Description

Technical Field

[0001] This application relates to computer technology, and more particularly to a method, apparatus, electronic device, and storage medium for recommending information in a game program. Background Technology

[0002] In the field of video game data processing and interactive control technology, player behavior analysis is often necessary to achieve refined operations. In related technologies, player profile features are generated using fixed logical judgments to produce static player data tags. This method results in low-frequency updates of these tags, making it difficult to reflect real-time changes in player status. Limited by this static profiling method, existing game level difficulty control typically employs pre-configured fixed parameter schemes. That is, during the level design phase, various difficulty coefficients are pre-set through configuration files. During level execution, it is impossible to dynamically adjust the difficulty coefficients based on the player's current real-time skill level or competitive state. This static difficulty mechanism easily leads to a mismatch between level difficulty and player ability. When the difficulty coefficient is too high or too low relative to the player's skill level, it can easily lead to user churn or reduced game stickiness.

[0003] Furthermore, in the advertising delivery process within games, related technologies typically employ a single delivery strategy, randomly displaying ads based solely on coarse-grained static tags, or showing the same ad content to players with different characteristics. Because this approach fails to consider players' specific level completion results, game progress, and real-time behavioral patterns for targeted matching, the displayed ad content has a low relevance to players' current potential needs, thus limiting the efficiency of ad resource utilization and conversion rates. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for recommending information in game programs, which can improve the efficiency and accuracy of information recommendation in game programs.

[0005] The technical solution of this application embodiment is implemented as follows: This application provides a method for recommending information in a game program, the method comprising: Based on the first player profile features and the first mapping rule, the first difficulty coefficient of the game level is determined, wherein the first player profile features are obtained before the game level is run in the game program; In response to a trigger operation on a game level, the game level is run based on the first difficulty coefficient, and player behavior data is acquired during the running of the game level. The first player profile features are updated based on the player behavior data to obtain the second player profile features; The first difficulty coefficient is updated based on the second player profile features to obtain the second difficulty coefficient, and the game level is run with the second difficulty coefficient. In response to the end of the game level, a target advertisement is predicted based on the result of the game level, the game level progress, and the second player profile characteristics, wherein the target advertisement is intended to be displayed when the game level ends.

[0006] This application provides an information recommendation device for game programs, including: The difficulty mapping module is used to determine the first difficulty coefficient of the game level based on the first player profile features and the first mapping rule, wherein the first player profile features are obtained before the game level is run in the game program; The data acquisition module is used to respond to the trigger operation for the game level, run the game level based on the first difficulty coefficient, and acquire player behavior data during the running of the game level; The data acquisition module is used to update the first player profile features based on the player behavior data to obtain the second player profile features; The difficulty mapping module is used to update the first difficulty coefficient based on the second player profile features to obtain a second difficulty coefficient, and run the game level with the second difficulty coefficient. An advertising recommendation module is used to predict target advertisements in response to the end of the game level, based on the result of the game level, the game level progress, and the characteristics of the second player profile, wherein the target advertisements are used to be displayed when the game level ends.

[0007] This application provides an electronic device, the electronic device comprising: Memory is used to store executable instructions or computer programs. When a processor executes computer-executable instructions or computer programs stored in the memory, it implements the game program information recommendation method provided in the embodiments of this application.

[0008] This application provides a computer-readable storage medium storing computer-executable instructions or computer programs, which, when executed by a processor, implement the information recommendation method for game programs provided in this application.

[0009] This application provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, they implement the information recommendation method for the game program provided in this application.

[0010] The embodiments of this application have the following beneficial effects: By collecting player behavior data in real time during level execution and adjusting the difficulty level from the first to the second, this system overcomes the limitations of traditional static difficulty settings that cannot handle fluctuations in player state. It dynamically adjusts the game difficulty based on current player behavior data, ensuring the difficulty curve dynamically matches the player's actual skill level and playability needs. This reduces the risk of player churn due to excessive difficulty or boredom due to excessively low difficulty. Targeted ad predictions not only rely on historical player profiles but also incorporate real-time contextual information at level completion (level results, progress) and the real-time adjusted second player profile. This multi-dimensional and highly timely feature input allows the ad recommendation logic to more accurately capture the player's psychological state and potential needs at the moment the game ends, thereby improving ad recommendation accuracy. Attached Figure Description

[0011] Figure 1 This is a schematic diagram illustrating the application mode of the information recommendation method for game programs provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application; Figure 3A This is a first flowchart illustrating the information recommendation method for game programs provided in this application embodiment; Figure 3B This is a second flowchart illustrating the information recommendation method for game programs provided in this application embodiment; Figure 4 This is a schematic diagram of the third process of the information recommendation method for game programs provided in the embodiments of this application.

[0012] It should be noted that the terms "first" and "second" mentioned above are only used to distinguish between different options and do not represent the degree of superiority or inferiority of the options or their priority in the implementation process. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is 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.

[0015] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0016] It should be noted that the data collection and processing in this application (e.g., user account behavior data in the game, user profiles) should strictly comply with the requirements of relevant laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0017] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0019] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0020] 1) User Profile: A user profile is a tagged model built based on user attributes, preferences, behaviors, and other information, used to abstractly describe the characteristics of a target user group. Essentially, a user profile integrates scattered user data (such as demographic information, consumption habits, device usage, etc.) into a structured set of tags, thus forming a virtual representative of the real user. Building a user profile typically relies on multi-dimensional data, including basic attributes (such as age, gender), behavioral data (such as browsing paths, access frequency), consumption characteristics, location information, etc. With technological advancements, in the digital age, user profiles, combined with algorithmic recommendations and semantic analysis, can dynamically update and integrate information such as social dynamics and emotional dimensions, achieving a shift from information output to cognitive co-construction. User profiles have wide applications, covering early product design, user research, service optimization, and other stages.

[0021] 2) Game Levels: These are the basic units that divide tasks or challenges within the game's progression, typically consisting of a map, objectives, and an interactive environment. Game levels can be understood as the space where the game takes place, such as battle scenes, mazes, or towns. They are the rules that aggregate game content, connecting elements like maps, tasks, NPCs, and items in a structured way to form the carrier of the player's experience. Essentially, game levels are containers for gameplay. Through goal-oriented design, they allow players to complete challenges within rules and system constraints, thereby gaining enjoyment and a sense of accomplishment. For example, in linear levels, levels may manifest as challenges in a strictly sequential order; while in open-world games, levels may be a collection of freely explorable areas.

[0022] 3) Difficulty coefficient: In the operational logic of a video game program, the difficulty coefficient refers to a set of numerical variables or parameters used to quantify the intensity of a level's challenge. As an input factor or weight for game logic operations, the difficulty coefficient acts on the attributes of virtual entities, environmental mechanisms, or logical judgment algorithms within the game through a preset mapping relationship, thereby dynamically adjusting the frequency of interactions, reaction speed, strategy complexity, or resource consumption required for players to achieve the level completion conditions.

[0023] 4) Extreme Gradient Boosting (XGBoost): This is a dedicated classification model within the extreme gradient boosting framework. It falls under the category of ensemble learning (boosting strategies) in machine learning and is essentially an improved and optimized version of the traditional gradient boosting decision tree (GBDT). XGBoost classifies classes by sequentially ensembled multiple decision trees (base learners), gradually correcting the prediction errors of preceding models and ultimately outputting discrete class labels, achieving accurate classification of samples.

[0024] 5) Lightweight Gradient Boosting Machine (LightGBM): This is an efficient, lightweight variant of Gradient Boosting Decision Tree (GBDT), and also belongs to the Boosting ensemble learning model category. Through targeted engineering optimizations and algorithmic improvements, the LightGBM model significantly reduces memory usage and increases training speed while maintaining or even surpassing the prediction accuracy of XGBoost. It is particularly suitable for machine learning tasks such as classification and regression on large-scale datasets (millions / hundreds of millions of samples).

[0025] In related technologies, player profile features are generated using fixed logical judgments to produce static player data tags. This method results in low tag updates and makes it difficult to reflect real-time changes in player status. Limited by this static profile processing method, existing game level difficulty control typically employs a pre-configured, fixed parameter scheme. That is, during the level design phase, various difficulty coefficients are pre-set through configuration files. During level execution, it's impossible to dynamically adjust the difficulty coefficients based on the player's current real-time skill level or competitive state. This static difficulty mechanism easily leads to a mismatch between level difficulty and player ability. When the difficulty coefficient is too high or too low relative to the player's skill level, it can easily lead to user churn or reduced game engagement.

[0026] Furthermore, in the advertising delivery process within games, related technologies typically employ a single delivery strategy, randomly displaying ads based solely on coarse-grained static tags, or showing the same ad content to players with different characteristics. Because this approach fails to consider players' specific level completion results, game progress, and real-time behavioral patterns for targeted matching, the displayed ad content has a low relevance to players' current potential needs, thus limiting the efficiency of ad resource utilization and conversion rates.

[0027] This application provides a method for recommending information in a game program, a device for recommending information in a game program, an electronic device, a computer-readable storage medium, and a computer program product, which can improve the efficiency and accuracy of information recommendation in game programs.

[0028] The following describes exemplary applications of the electronic devices provided in the embodiments of this application. These electronic devices can be implemented as terminal devices, such as laptops, tablets, desktop computers, set-top boxes, smart TVs, in-vehicle terminals, virtual reality (VR) devices, augmented reality (AR) devices, and other various types of terminals. They can also be implemented as servers. The following will describe exemplary applications when the electronic device is implemented as a terminal device or a server.

[0029] refer to Figure 1 , Figure 1 This is a schematic diagram illustrating the application mode of the information recommendation method for game programs provided in this application embodiment; for example, Figure 1 The system involves server 200, network 300, terminal device 400, and database 500. Terminal device 400 is connected to server 200 through network 300, which can be a wide area network (WAN), a local area network (LAN), or a combination of both.

[0030] For example, terminal device 400 is used to install game applications, server 200 is the game application platform, and database 500 is used to store game data and player-related data.

[0031] In some embodiments, a user logs into a game application via a terminal device 400. The terminal device 400 sends user data to a server 200 via a network 300. The server 200 calls the game program information recommendation method provided in this application embodiment to determine a first player profile and a first difficulty coefficient, and sends game data to the terminal device 400 for running a game level with the first difficulty coefficient. The terminal device 400 runs the game level with the first difficulty coefficient and collects user behavior data in the level, and sends the user behavior data to the server 200. The server 200 calls the game program information recommendation method provided in this application embodiment to update the player profile and difficulty coefficient based on the user behavior data to obtain a second player profile and a second difficulty coefficient, and sends game data to the terminal device 400 for running a game level with the second difficulty coefficient. When the terminal device 400 finishes running the game level, the server 200 sends a target advertisement to the terminal device 400 based on the level running result, progress, and the second player profile, and the terminal device 400 displays the target advertisement.

[0032] In some embodiments, server 200 may be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Terminals and servers can be connected directly or indirectly via wired or wireless communication, which is not limited in this embodiment.

[0033] See Figure 2 , Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may be... Figure 1 Server 200, Figure 2 The server 200 shown includes at least one processor 410, memory 450, and at least one network interface 420. Various components in the terminal device 400 are coupled together via a bus system 440. It is understood that the bus system 440 is used to implement communication between these components. In addition to a data bus, the bus system 440 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2 The general labeled all buses as Bus System 440.

[0034] Processor 410 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. Among them, the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0035] The memory 450 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 450 may optionally include one or more storage devices physically located away from the processor 410.

[0036] The memory 450 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 450 described in this application embodiment is intended to include any suitable type of memory.

[0037] In some embodiments, memory 450 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0038] Operating system 451 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks; The network communication module 452 is used to reach other electronic devices via one or more (wired or wireless) network interfaces 420, exemplary network interfaces 420 including: Bluetooth, WiFi, and Universal Serial Bus (USB), etc. In some embodiments, the apparatus provided in this application can be implemented in software. Figure 2 An information recommendation device 455 for a game program stored in memory 450 is shown. It can be software in the form of programs and plug-ins, and includes the following software modules: data acquisition module 4551, difficulty mapping module 4552, and advertisement recommendation module 4553. These modules are logically related and can therefore be arbitrarily combined or further divided according to the functions they implement. The functions of each module will be described below.

[0039] The method for recommending game programs provided in this application will be described in conjunction with exemplary applications and implementations of the terminal devices provided in the embodiments of this application.

[0040] The following describes the game program information recommendation method provided in the embodiments of this application. As mentioned above, the electronic device implementing the game program information recommendation method of the embodiments of this application can be a terminal device or a server, or a combination of both. Therefore, the executing entity of each step will not be described again below.

[0041] It should be noted that the information recommendation processing example below is based on the example of virtual item advertisements within a game application. Based on the understanding of the following, those skilled in the art can apply the information recommendation method for game programs provided in this application to the processing of recommendations for other types of advertisements, such as advertisements for goods outside of game applications.

[0042] See Figure 3A , Figure 3A This is a flowchart illustrating the information recommendation method for game programs provided in this application embodiment, which will be combined with... Figure 3A The steps shown are explained. Figure 3A The entity responsible for executing the steps is Figure 1 Server 200 in the middle.

[0043] In step 301, the first difficulty coefficient of the game level is determined based on the first player profile features and the first mapping rule.

[0044] Here, the first player profile features are obtained before the game level runs in the game program.

[0045] For example, player profile features are derived from static player data, which refers to pre-existing data that does not update with player behavior before the game level runs. Difficulty coefficients are parameters or a set of parameters used to adjust the difficulty of game levels.

[0046] In some embodiments, the static data includes: channel data, device attributes, and installation time period. Before step 301, the channel data, device attributes, and installation time period are acquired, wherein the device attributes are the device attributes of the terminal device that installs the game program, and the channel data are the characteristics of the installation channel of the game program; the channel data, device attributes, and installation time period are cleaned to obtain cleaned static data; the static data is encoded to obtain the first player profile features.

[0047] For example, channel data refers to the characteristics of the installation channels of a game application. For instance, if a game application is downloaded from a social media platform via a terminal device, the user information might include the characteristics of the advertising materials encountered on that social media platform during the game installation. Device attributes are the device attributes of the terminal device that installed the game application, and the installation period is the time period during which the game application was installed on the terminal device. Device information includes hardware attributes and software environment. Hardware attributes include, for example, the device model, terminal device brand, memory (e.g., 8GB), and graphics processing unit (GPU) performance level (high-end / mid-range / low-end). Software environment includes, for example, the operating system, network type (5G or Wi-Fi), and screen resolution (e.g., 2796×1290). Device information is obtained in real-time through system application programming interfaces (APIs), such as Build.MODEL for Android and UIDevice.model for iOS.

[0048] Data cleaning includes noise reduction and missing value imputation. Encoding involves vectorizing static data; this can be done through numerical encoding or one-hot encoding.

[0049] In some embodiments, step 301 is implemented in the following way: classifying the features of the first player profile to obtain the first player type, wherein the first mapping rule includes the mapping relationship between the first player type and the difficulty coefficient; querying the first mapping rule based on the first player type, and using the queried difficulty coefficient as the first difficulty coefficient of the game level.

[0050] For example, classification processing can be implemented using a decision tree model. The first mapping rule is pre-set and can be obtained through A / B testing. A / B testing involves creating two solutions (such as two pages) for the same goal, having some users use solution A and others use solution B, recording user behavior, and seeing which solution better matches the design. In this embodiment, players are clustered based on player profile characteristics to obtain multiple player types. For players of the same type, game levels with different difficulty coefficients are issued, and the difficulty coefficient corresponding to the game level with the highest success rate is selected as the initial difficulty coefficient for that player type.

[0051] In step 302, in response to the trigger operation for the game level, the game level is run based on the first difficulty coefficient, and player behavior data during the game level operation is obtained.

[0052] For example, running a level based on the first difficulty level means setting the game parameters corresponding to the first difficulty level as the parameters for the game level, and running the game level based on the set data. Player behavior data during the game level's execution is obtained through the terminal device, which then sends the obtained player behavior data to the server.

[0053] In some embodiments, obtaining player behavior data during the game level operation in step 302 is achieved by: obtaining candidate data of at least one of the following types during the game level operation: login data, operation data within the game level, level, consumption data, and virtual resource consumption data; performing at least one of the following desensitization processes on the candidate data to obtain player behavior data: noise addition based on Gaussian noise; deletion processing; encryption processing.

[0054] For example, player actions within a game application include: logging in, game level operations, game spending, and game resource consumption. Examples include: Login patterns: login time (e.g., peak hours), login frequency (daily / weekly active users), and duration of each online session; In-level operation data: Operation density: number of operations per unit time (e.g., clicks, skill releases), error rate, and response time for critical actions; Difficulty adaptability: number of level attempts, distribution of failure points, and deviation between completion time and expected time; Spending capacity: recharge amount, average single transaction amount, total cumulative spending, and spending frequency (e.g., 3-5 times per month); Resource consumption: usage and consumption of various virtual items, virtual currency consumption, and other data.

[0055] For example, noise reduction refers to perturbing the parameters of the original candidate data with Gaussian noise to obtain the noisy parameters (e.g., if the original login timestamp is 1000, and the introduced noise value is 0.2, then the uploaded value will be between 800 and 1200). When performing statistical calculations on the server side, noise reduction can be used to obtain an overall login time distribution that is unaffected by noise; deletion refers to using the data minimization principle for candidate data, collecting only the necessary fields and deleting other non-essential fields; encryption refers to encrypting the private data in the candidate data.

[0056] In step 303, the first player profile features are updated based on player behavior data to obtain the second player profile features.

[0057] For example, the update process can be implemented as follows: extract features from player behavior data, and then weight and fuse the extracted features with the features of the first player profile to obtain the features of the second player profile.

[0058] In some embodiments, step 303 is implemented in the following manner: feature extraction is performed on player behavior data to obtain incremental behavior features; based on the attention mechanism, the first weight value corresponding to each sub-feature of each dimension in the incremental behavior features and the first player profile features is determined; the first weight value of each sub-feature is normalized to obtain the second weight value of each sub-feature; and the second weight value of each sub-feature and its corresponding second weight value are weighted and summed to obtain the second player profile features.

[0059] For example, an attention weight dynamic allocation mechanism is adopted to automatically calculate the contribution weights of various sub-features in static data and player behavior features. During the iteration process, feature types can be continuously improved and added to obtain the first weight value. The temporal correlation between sub-features is extracted by a recurrent neural network (GRU), and a weight distribution is generated by combining it with a normalization (Softmax) function to obtain the second weight value. The weighted sum of each sub-feature is then used to obtain the updated player profile features.

[0060] In step 304, the first difficulty coefficient is updated based on the second player profile features to obtain the second difficulty coefficient, and the game level is run with the second difficulty coefficient.

[0061] For example, based on the characteristics of the second player profile and the preset mapping rules, an adjustment value for the difficulty coefficient is determined. This adjustment value is then added to the first difficulty coefficient to obtain the second difficulty coefficient. The adjustment value can be a positive or negative number, such as -3 or +1.

[0062] In some embodiments, step 304 is implemented by: classifying the features of the second player profile to obtain the second player type; obtaining the second mapping rule, wherein the second mapping rule includes the mapping relationship between the second player type and the difficulty coefficient adjustment value; querying the second mapping rule based on the second player type, and adding the queried difficulty coefficient adjustment value to the first difficulty coefficient to obtain the second difficulty coefficient.

[0063] For example, classification processing can be implemented using the decision tree model described above, which will not be repeated here. The second preset mapping rule can be obtained through A / B testing. For instance, the second preset mapping rule can be determined through long-term A / B group testing, such as reducing the difficulty after 3 failures for group A and 5 failures for group B. By comprehensively comparing player retention data and advertising revenue data from groups A and B, the second preset mapping rule for this type of player can be determined.

[0064] In step 305, in response to the end of a game level, a target advertisement is predicted based on the result of the game level, the progress of the game level, and the characteristics of the second player profile.

[0065] For example, the targeted ad is displayed at the end of a game level. The targeted ad is displayed on the terminal device and is sent to the terminal device via a server. Based on the game level result, game level progress, second player profile characteristics, and candidate ad characteristics, the corresponding metrics for the candidate ads are predicted, and the target ad is selected based on these metrics.

[0066] In some embodiments, reference Figure 3B , Figure 3B This is a second flowchart illustrating the information recommendation method for game programs provided in this application embodiment. Step 305 involves... Figure 3B Steps 3051 to 3054 are implemented, and the details are explained below.

[0067] In step 3051, the results and progress of the game level are encoded to obtain encoded vector features.

[0068] For example, the outcome of a game level can be success (level completed, corresponding to 100% game level progress) or failure (level not completed, corresponding to less than 100% game level progress), with game level progress expressed as a percentage. Encoding is the process of converting the numerical values ​​of the game level outcome and game level progress into a vector of a uniform dimension; this encoding process can be one-hot encoding.

[0069] In step 3052, the advertising features and encoding vector features of the candidate advertisement are concatenated with the second player profile features to obtain the concatenated features.

[0070] For example, the advertising features of candidate ads are pre-acquired and stored on a server or in a database. The advertising features of candidate ads, the encoded vector features, and the second player profile features are concatenated together to form a concatenated feature.

[0071] In step 3053, the prediction balance coefficient corresponding to each candidate advertisement is predicted based on the splicing features.

[0072] Here, the predicted balance coefficient is the value of a balance function that measures advertising revenue and user retention rate.

[0073] For example, the balance function between advertising revenue (R_ad) and user retention rate (R_retention) is expressed as the following formula (1): Maximize(α×R_ad+(1 (α)×R_retention)(1) The weighting coefficient α (0 ≤ α ≤ 1) is used to adjust the frequency of ads inserted in game levels, and Maximize is the maximization function. The formula means obtaining the maximum value of the weighted sum of ad revenue and user retention rate. When user retention rate decreases, the value of α is lowered to reduce ad interference. Lowering the ad weight means reducing the number and frequency of ads in the game, which can reduce interruptions to the player's game flow and optimize the player experience, thus reducing player churn and improving retention rate. Predictive processing is implemented through a decision model.

[0074] In step 3054, the candidate advertisement corresponding to the largest prediction balance coefficient is selected as the target advertisement.

[0075] For example, the candidate ad with the largest prediction balance coefficient indicates that it achieves a better balance between user retention and ad revenue compared to other candidate ads, and this candidate ad is selected as the target ad.

[0076] In this embodiment, by collecting player behavior data in real time during level execution and adjusting the difficulty from a first to a second difficulty coefficient, the shortcomings of traditional static difficulty configuration in handling player state fluctuations are overcome. The game difficulty can be adjusted instantly based on the player's current behavior data, making the difficulty curve dynamically match the player's actual skill level and play needs, reducing the risk of player churn due to excessive difficulty or player boredom due to excessively low difficulty. The prediction of targeted advertisements not only relies on historical profiles but also incorporates real-time contextual information at level completion (level results, progress) and the real-time adjusted second player profile. Multi-dimensional and highly timely feature inputs enable the advertising recommendation logic to more accurately obtain the player's psychological state and potential needs at the moment the game ends, thereby improving the accuracy of advertising recommendations.

[0077] The following will describe an exemplary application of the information recommendation method for a game program according to the embodiments of this application in a practical application scenario.

[0078] Player profile feature fusion schemes in related technologies generate player data tags through manually preset static rules and directly collect player data. The server then statistically analyzes player data by periodically batch processing, cleaning, and transforming user data. However, these data fusion schemes have the following drawbacks: insufficient data breadth, relying solely on device / log data while ignoring channel data, making it impossible to personalize game content in the early stages of the game; and the offline batch processing for statistical analysis of player data results in significant latency, hindering real-time dynamic adjustments to player profile features; furthermore, the player data tags generated through preset static rules are fixed and cannot respond to changes in player behavior in real time. In related technologies, advertisements in game applications appear at set times or when the number of operations reaches a certain threshold. This mechanical insertion of advertisements based solely on fixed rules (such as time intervals or number of operations) leads to a separation between game application levels and the timing of advertisement insertion, completely disconnecting the specific advertisement triggering from level progress and player status. The proposed solution of separating technical levels from advertisements has the following drawbacks: Firstly, it can interrupt the player's gaming experience. Ads forcibly pop up when players are immersed in the game, disrupting the game's rhythm and exacerbating negative emotions, leading to some players leaving. Secondly, it can cause mismatches between the ad scene and the game scene. For example, the ad type may be out of sync with the level stage, pushing high-value item ads during the new player period, or showing low-priced incentive videos after a player fails in the game, which affects the player's gaming experience and reduces the efficiency of ad recommendations.

[0079] This application provides a method for recommending information in a game program. It adds installation source and channel data to the device information to more comprehensively construct the initial player profile features, providing more personalized game content in the early game levels. Additionally, it collects player data locally on the client side, extracts corresponding behavioral features, and sends them to the server in real time. The server processes this data in real time, updates the player profile features, and then returns the updated profile data to the client, enabling the client to respond to changes in player behavior and adjust the game content accordingly.

[0080] See Figure 4 , Figure 4 This is a flowchart illustrating the information recommendation method for game programs provided in this application embodiment, which will be combined with... Figure 4 The steps shown are explained. Figure 4 The entity responsible for executing the steps is Figure 1 Server 200 in the middle.

[0081] In step 401, initial player profile features are generated based on the game application's channel data, device attributes, and installation time period.

[0082] For example, channel data refers to the characteristics of the installation channels of a game application. For instance, if a game application is downloaded from a social media platform via a terminal device, the user information might include the characteristics of the advertising materials encountered on that social media platform during the game installation. Device attributes are the device attributes of the terminal device that installed the game application, and installation time period is the time period during which the game application was installed on the terminal device.

[0083] Ad creative characteristics include ad creative type (e.g., video / image / text), ad theme (e.g., fantasy / realism), and the game type recommended by the ad (e.g., competitive, simulation, story-driven). The backend uses a Software Development Kit (SDK) or attribution API (such as AppsFlyer, Adjust) to match the mapping relationship between users and channels to obtain user information.

[0084] Device information includes hardware attributes and software environment. Hardware attributes include, for example, the model of the data acquisition device, the brand of the terminal device, memory (e.g., 8GB), and the performance level of the graphics processing unit (GPU) (high-end / mid-range / low-end). Software environment includes, for example, the operating system, network type (5G or Wi-Fi), and screen resolution (e.g., 2796×1290). Device information is obtained in real time through system application programming interfaces (APIs), such as Build.MODEL for Android and UIDevice.model for iOS.

[0085] For example, channel data, device attributes, and installation time periods are considered static data. After collecting this static data, it undergoes data cleaning. Data cleaning includes noise reduction and missing value imputation. Noise reduction is achieved by removing abnormal devices (such as emulators and developer devices) and fake installations (clickjacking). Missing value imputation is achieved by filling missing channel tags with the mean or cluster centers of similar ad creatives.

[0086] Obtaining player profile features based on static data can be achieved by vectorizing the static data. Vectorization can be done using numerical encoding or one-hot encoding. For example, there are three types of advertising creatives: video, image, and text. Video can be encoded as [1, 0, 0], and image can be encoded as [0, 1, 0]. Device information can be categorized into device levels (high-end = 3, mid-range = 2, low-end = 1) based on the image processor / memory / CPU, serving as a hardware capability label.

[0087] In step 402, in response to the game level being triggered in the game application, an initial difficulty coefficient is generated based on the initial player profile characteristics, and the game level is loaded based on the initial difficulty coefficient.

[0088] For example, the initial difficulty coefficient is for the currently triggered game level. Based on a preset matching method, player profile features are mapped to a specific initial difficulty coefficient. This mapping process can be implemented as follows: classify the player profile features to obtain the player types corresponding to those features; query the first preset mapping rule between the player type and the initial difficulty coefficient to obtain the initial difficulty coefficient. The first preset mapping rule can be obtained through A / B testing. A / B testing involves creating two solutions (e.g., two pages) for the same goal, having some users use solution A and others use solution B, recording user behavior, and seeing which solution better suits the design. In this embodiment, players are clustered based on their profile features to obtain multiple player types. For players of the same type, game levels with different difficulty coefficients are issued, and the difficulty coefficient corresponding to the game level with the highest success rate is selected as the initial difficulty coefficient for that player type.

[0089] Clustering can be achieved using a clustering model (K-Means) to classify player types. K-Means is an unsupervised learning algorithm that iteratively optimizes the data to divide it into K clusters (e.g., player groups, where each group represents a player type), aiming to minimize intra-cluster differences and maximize inter-cluster differences. The process includes: randomly selecting K initial centroids (e.g., the center points of player profile features). Assigning each player profile feature to the cluster with the nearest Euclidean distance centroid. Recalculating the mean of all points within the cluster as the new centroid. Repeating this process until the centroids stabilize or the maximum number of iterations is reached.

[0090] Player profile features can be classified using the XGBoost classification model. XGBoost is a supervised learning algorithm based on Gradient Boosting Decision Tree (GBDT), which integrates multiple weak classification trees (such as decision trees) to achieve a strong classifier.

[0091] In step 403, during the operation of the game level, the player's behavior data in the game level is acquired.

[0092] For example, behavioral data is acquired locally by the player's terminal device and sent to the server. The behavioral data is anonymized before being sent to the server.

[0093] For example, desensitization can be achieved by adding noise to behavioral data. Gaussian noise conforming to GDPR standards is injected into the data to ensure that individual player data is untraceable while group statistical features remain usable. This is achieved by adding noise to user behavioral data, using a preset noise value to add noise to parameters in the data. During server-side statistical calculations, the original data can be obtained by removing noise using the preset noise value. For example, when uploading player login times, noise is added to the login times of individual users. For instance, if the original login timestamp is 1000, and the introduced noise value is 0.2, the uploaded value will be between 800 and 1200, avoiding uploading the original data. During server-side statistical calculations, the overall login time distribution can be obtained through noise removal, unaffected by the noise. Another example is using differential privacy technology. Laplace noise is added before feature upload (e.g., adding noise to "level stay time 120 seconds" converts it to "level stay time 118 to 122 seconds, random duration") to ensure that individual players are unidentifiable.

[0094] For example, data anonymization can involve deletion. For instance, before uploading data to the server, fields that directly identify an individual (such as device ID and GPS coordinates) are deleted, while behavioral patterns (such as "average number of operations per session = 25") are retained. Another example is adopting a data minimization principle before uploading data to the server, collecting only necessary fields (such as not collecting player chat content, only counting social interaction frequency). Terminal devices convert sensitive data (such as physiological information and operation trajectories) into anonymized feature vectors, uploading only non-privacy-sensitive feature values ​​(such as "level frustration index 0.8" instead of specific operation logs), while other data that could expose privacy is deleted, mitigating privacy risks at the source.

[0095] For example, data anonymization can be achieved through encryption. The terminal device encrypts the calculated model gradient parameters and some additional player features before transmitting them to the server to prevent data leakage. The server integrates parameters from multiple terminals to update the global model and returns the results to the client. The original data does not leave the user's device, ensuring user privacy. Security verification can also be performed through penetration testing to verify that the features in the local data of the terminal device cannot reconstruct the original behavior (such as the operation sequence).

[0096] In this embodiment, initial player profile feature tags are generated based on channel data such as advertising material characteristics from the user's installation source and device information. The player profile features are dynamically expanded and improved based on player behavior, and in-game behavioral data (e.g., level dwell time, level operation data, item reuse rate, social collaboration data, and some physiological information) are integrated in real time to gradually build a dynamic profile of the user account throughout the game's lifecycle. The player profile features are anonymized before being uploaded to avoid privacy risks associated with directly uploading source data.

[0097] For example, player actions within a game application include: logging in, game level operations, game spending, and game resource consumption. Examples include: Login patterns: login time (e.g., peak hours), login frequency (daily / weekly active users), and duration of each online session; In-level operation data: Operation density: number of operations per unit time (e.g., clicks, skill releases), error rate, and response time for critical actions; Difficulty adaptability: number of level attempts, distribution of failure points, and deviation between completion time and expected time; Spending capacity: recharge amount, average single transaction amount, total cumulative spending, and spending frequency (e.g., 3-5 times per month); Resource consumption: usage and consumption of various virtual items, virtual currency consumption, and other data.

[0098] In step 404, the player profile features are adjusted based on the behavioral data to obtain updated player profile features, and the difficulty coefficient of the game level is adjusted based on the updated player profile features.

[0099] For example, feature extraction is performed on behavioral data to obtain behavioral features. These behavioral features and player profile features are then weighted and summed. An attention weight dynamic allocation mechanism is employed to automatically calculate the contribution weights of various sub-features within the static data and player behavioral features. This allows for continuous improvement and the addition of new feature types during the iteration process. A recurrent neural network (GRU) is used to extract the temporal correlation between sub-features, and a weight distribution is generated using a normalization (Softmax) function. Finally, the weighted summation of each sub-feature yields the updated player profile features.

[0100] For example, the updated player profile features are categorized to obtain updated player types. Based on a second pre-defined mapping rule between the updated player types and the difficulty adjustment value, a difficulty adjustment value is obtained. This adjusted value is then added to the current difficulty coefficient to obtain the adjusted difficulty coefficient. The second pre-defined mapping rule can be obtained through A / B testing. For instance, the second pre-defined mapping rule can be determined through long-term A / B group testing, such as lowering the difficulty after 3 failures for group A and 5 failures for group B. By comprehensively comparing player retention data and advertising revenue data from groups A and B, a second pre-defined mapping rule specific to that player type can be determined.

[0101] In step 405, in response to the end of the game level, the result of the game level is determined.

[0102] For example, the end of a game level can be caused by success or failure, and the result includes a success state and a failure state.

[0103] In step 406, when the result is failure, the decision model is invoked to determine the target advertisement based on the failure result, level progress, and player profile characteristics. The target advertisement is then displayed in the game application, and the click status and conversion information of the target advertisement are recorded.

[0104] For example, different types of targeted ads can pop up based on level progress and player status.

[0105] In some embodiments, if the game level completion reaches a completion threshold and the result is failure, a first target advertisement is displayed; if the game level completion is less than the completion threshold and the result is failure, a second target advertisement is displayed. The second target advertisement recommends more virtual game items than the first target advertisement.

[0106] In some embodiments, based on failure results, level progress, and player profile characteristics, a decision model is invoked to classify the player's interests and preferences to obtain interest types. A third advertisement corresponding to the interest type is then displayed. The third advertisement is used to recommend other products outside the game application. If the user watches the third advertisement for a preset duration or triggers the third advertisement, a preset number of virtual game items are awarded.

[0107] In some embodiments, based on failure results, level progress, and player profile characteristics, a decision model is invoked to classify the player's emotional type, resulting in an emotional type and player ability level. A fourth advertisement is then matched based on the emotional type and player ability level. This fourth advertisement is an advertisement with a preset mapping rule between the emotional type and the player ability level. For example: after three consecutive failures, the player's emotional type is high anxiety; if the completion time is less than the average completion time, the player's ability is marked as high-ability. Another example: in a high anxiety state and with level completion below a threshold, a fourth advertisement containing incentive items (revival items / coin rewards) is displayed; in a relaxed state and with level completion above a threshold, a lightweight graphic advertisement (without interrupting the flow) is displayed.

[0108] For example, the decision model is used to determine the target advertisement based on the failure result, level progress, and player profile characteristics. Therefore, the optimization objective of the decision model can be to balance advertising revenue and user retention rate. The advertising revenue (R_ad) is calculated as: advertising display revenue + revenue generated based on ad clicks × estimated click-through rate (CTR). User retention rate (R_retention) is a core operational indicator for measuring the continued use of Internet application users. It is defined as the proportion of new users who remain active after a specified period of time. The balance function between advertising revenue and user retention rate (the balance coefficient mentioned above) is expressed as the following formula (1): Maximize(α×R_ad+(1 (α)×R_retention)(1) The weighting coefficient α (0 ≤ α ≤ 1) is used to adjust the frequency of ads inserted in game levels, and Maximize is the maximization function. The formula means obtaining the maximum value of the weighted sum of ad revenue and user retention rate. When user retention rate decreases, the value of α is lowered to reduce ad interference. Lowering the ad weight means reducing the number and frequency of ads in the game, which can reduce interruptions to the player's game flow and optimize the player experience, thus reducing player churn and improving retention rate.

[0109] The training data for the decision model consists of sample results, level progress, player profile features, and sample ads. The labels are ad revenue and user retention rate function values. Based on the sample results, level progress, and player profile features, the decision model predicts the balance function values ​​of ad revenue and user retention rate for multiple sample ads. The sample ad with the largest function value is selected as the target ad. The loss function can be cross-entropy loss, which measures the difference between the predicted balance function value and the actual balance function value.

[0110] Step 408 is executed after step 406.

[0111] In step 407, when the result is success, the success result, the level progress, and the operation data in this game level are recorded.

[0112] For example, the operation data includes information such as level progress, pass rate, number of virtual items of players, and other player operation information, which is used to extract player behavior characteristics locally and then upload them to the server for streaming updates of player profile characteristics.

[0113] Step 408 is also executed after step 407.

[0114] In step 408, the decision model is updated based on the collected data.

[0115] For example, the decision model update is performed on the server side. Each time the client uploads new player action data, it's uploaded as a stream. This uploaded data doesn't contain all player information. The server analyzes and makes decisions based on the client's uploads, updating the player profile features. The update method can be referenced from step 404, where the updated player profile features are obtained. The updated profile content is then returned to the client. The decision model update method is based on Kafka streaming data processing, incrementally updating model parameters hourly to ensure high-value scenario acquisition rates and avoid revenue loss due to strategy lag. The decision model can be a Lightweight Gradient Boosting Machine (LightGBM), an efficient machine learning framework based on Gradient Boosting Decision Trees (GBDT), primarily used for tasks such as classification, regression, and ranking. The LightGBM inference model extracts features from player behavior data, fusing player behavior data with player profile features to form updated player profile features. The LightGBM inference model employs a dynamic pruning strategy, setting different pruning policies for players of different values. For example, the pruning rate is 5% for highly active players, 30% for moderately active players, and 60% for inactive players. The level of activity is determined by a preset activity range, and activity is calculated based on the frequency of player logins, actions, and other behaviors in the game. The pruning rate refers to the incremental proportion (compared to the previous change) in the player behavior data uploaded by the client. A pruning rate of 5% means that the incremental change does not exceed 5%, and the LightGBM inference model is not used to recalculate, thus reducing server costs.

[0116] In this embodiment, a dynamic level adjustment mechanism and an advertising delivery mechanism are integrated. Initial player profile feature data is mapped to an initial difficulty coefficient using preset mapping rules. Player profile features are dynamically adjusted based on real-time player behavior, and the difficulty coefficient is adjusted synchronously with the profile data. Furthermore, by combining level progress (e.g., tutorial levels, later game levels) with the current difficulty status (e.g., level difficulty measured by failure rate, average failure rate 50%, average failure rate 80%), different types and values ​​of advertisements are triggered in different scenarios. Dynamic difficulty calibration improves the completion rate and reduces early player churn. Static data accurately matches the initial difficulty, providing novice players with a better transition level, while experienced players can more easily navigate tutorial and simple levels, quickly adapting to their skill level and providing a better experience from the beginning. Ads are triggered based on level status and displayed according to the needs of actual application scenarios, reducing interference, improving ad conversion rates, and reducing user churn. Considering the balance between advertising revenue and retention rate, both advertising revenue and user retention are improved. Privacy compliance risk control reduces the risk of user data leakage and solves cross-platform data issues. When the client extracts user data locally and uploads the user data to the server, it avoids directly uploading the source data and only extracts the data features.

[0117] The following description continues to illustrate the exemplary structure of the game program information recommendation device 455 provided in the embodiments of this application as a software module. In some embodiments, such as... Figure 2 As shown, the software modules in the game program information recommendation device 455 stored in the memory 450 may include: a difficulty mapping module 4552, used to determine a first difficulty coefficient of the game level based on a first player profile feature and a first mapping rule, wherein the first player profile feature is obtained before the game level runs in the game program; a data acquisition module 4551, used to run the game level based on the first difficulty coefficient in response to a trigger operation for the game level, and acquire player behavior data during the running of the game level; the data acquisition module 4551, used to update the first player profile feature based on the player behavior data to obtain a second player profile feature; the difficulty mapping module 4552, used to update the first difficulty coefficient based on the second player profile feature to obtain a second difficulty coefficient, and run the game level with the second difficulty coefficient; and an advertisement recommendation module 4553, used to predict a target advertisement based on the result of the game level, the game level progress, and the second player profile feature in response to the end of the game level, wherein the target advertisement is used to be displayed when the game level ends.

[0118] In some embodiments, the data acquisition module 4551 is configured to acquire channel data, device attributes, and installation time period before determining the first difficulty coefficient of the game level based on the first player profile features and the first mapping rule, wherein the device attributes are the device attributes of the terminal device that installed the game program, and the channel data are the features of the installation channel of the game program; perform data cleaning on the channel data, device attributes, and installation time period to obtain cleaned static data; and encode the static data to obtain the first player profile features.

[0119] In some embodiments, the difficulty mapping module 4552 is used to classify the first player profile features to obtain a first player type, wherein the first mapping rule includes a mapping relationship between the first player type and the difficulty coefficient; based on the first player type, the first mapping rule is queried, and the queried difficulty coefficient is used as the first difficulty coefficient of the game level.

[0120] In some embodiments, the data acquisition module 4551 is used to acquire at least one type of candidate data during the operation of the game level: login data, operation data within the game level, level, consumption data, and virtual resource consumption data; and to perform at least one of the following desensitization processes on the candidate data to obtain player behavior data: noise addition based on Gaussian noise; deletion processing; and encryption processing.

[0121] In some embodiments, the data acquisition module 4551 is used to extract features from the player behavior data to obtain incremental behavior features; based on an attention mechanism, determine the first weight value corresponding to each sub-feature of each dimension in the incremental behavior features and the first player profile features; normalize the first weight value of each sub-feature to obtain a second weight value of each sub-feature; and perform a weighted summation of each sub-feature and its corresponding second weight value to obtain the second player profile features.

[0122] In some embodiments, the difficulty mapping module 4552 is used to classify the second player profile features to obtain a second player type; obtain a second mapping rule, wherein the second mapping rule includes a mapping relationship between the second player type and the difficulty coefficient adjustment value; query the second mapping rule based on the second player type, and add the queried difficulty coefficient adjustment value to the first difficulty coefficient to obtain the second difficulty coefficient.

[0123] In some embodiments, the ad recommendation module 4553 is used to encode the results and progress of the game level to obtain encoded vector features; concatenate the ad features of the candidate ads, the encoded vector features, and the second player profile features to obtain concatenated features; predict the prediction balance coefficient corresponding to each candidate ad based on the concatenated features, wherein the prediction balance coefficient is a balance function value that measures ad revenue and user retention rate; and select the candidate ad corresponding to the largest prediction balance coefficient as the target ad.

[0124] This application provides a computer program product, which includes a computer program or computer executable instructions. A processor executes the computer program or computer executable instructions to implement the game program information recommendation method described in this application.

[0125] This application provides a computer-readable storage medium storing computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the information recommendation method for the game program provided in this application. For example, ... Figure 3A The information recommendation method for the game program is shown.

[0126] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0127] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, 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 stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0128] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, 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 co-located files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0129] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0130] In summary, this application's embodiments, by collecting player behavior data in real time during level execution and adjusting the difficulty from a first to a second difficulty coefficient, overcome the shortcomings of traditional static difficulty configurations that cannot cope with fluctuations in player states. It can instantly adjust the game difficulty based on the player's current behavior data, making the difficulty curve dynamically match the player's actual skill level and play needs, reducing the risk of player churn due to excessive difficulty or player boredom due to excessively low difficulty. The prediction of targeted advertisements not only relies on historical profiles but also incorporates real-time contextual information at level settlement (level results, progress) and the real-time adjusted second player profile. Multi-dimensional and highly timely feature inputs enable the advertising recommendation logic to more accurately obtain the player's psychological state and potential needs at the moment the game ends, thereby improving the accuracy of advertising recommendations.

[0131] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. An information recommendation method for a game program, characterized by, The method includes: Based on the first player profile features and the first mapping rule, the first difficulty coefficient of the game level is determined, wherein the first player profile features are obtained before the game level is run in the game program; In response to a trigger operation on a game level, the game level is run based on the first difficulty coefficient, and player behavior data is acquired during the running of the game level. The process of updating the first player profile feature based on the player behavior data to obtain the second player profile feature includes: extracting features from the player behavior data to obtain incremental behavior features; determining first weight values ​​corresponding to sub-features of each dimension in the incremental behavior features and the first player profile features based on an attention mechanism; normalizing the first weight value of each sub-feature to obtain a second weight value of each sub-feature; and performing a weighted summation of each sub-feature and its corresponding second weight value to obtain the second player profile feature. The first difficulty coefficient is updated based on the second player profile features to obtain a second difficulty coefficient, and the game level is run with the second difficulty coefficient. The process of updating the first difficulty coefficient based on the second player profile features to obtain a second player type includes: classifying the second player profile features to obtain a second player type; obtaining a second mapping rule, wherein the second mapping rule includes a mapping relationship between the second player type and a difficulty coefficient adjustment value; querying the second mapping rule based on the second player type, and adding the queried difficulty coefficient adjustment value to the first difficulty coefficient to obtain the second difficulty coefficient. In response to the end of the game level, the result and progress of the game level are encoded to obtain encoded vector features; The ad features of the candidate ads, the encoded vector features, and the second player profile features are concatenated to obtain the concatenated features; Based on the splicing features, a prediction balance coefficient is predicted for each candidate advertisement, wherein the prediction balance coefficient is a balance function value used to measure advertising revenue and user retention rate. The candidate ad corresponding to the largest predicted balance coefficient is selected as the target ad, wherein the target ad is used to be displayed at the end of the game level.

2. The method of claim 1, wherein, Before determining the first difficulty coefficient of the game level based on the first player profile features and the first mapping rule, the method further includes: Acquire channel data, device attributes, and installation time period, wherein the device attributes are the device attributes of the terminal device on which the game program is installed, and the channel data are the characteristics of the installation channel of the game program; The channel data, the device attributes, and the installation period are cleaned to obtain cleaned static data. The static data is encoded to obtain the first player profile features.

3. The method of claim 1, wherein, The determination of the first difficulty coefficient of the game level based on the first player profile features and the first mapping rule includes: The first player profile features are classified to obtain a first player type, wherein the first mapping rule includes the mapping relationship between the first player type and the difficulty coefficient; Based on the first player type, query the first mapping rule and use the difficulty coefficient obtained from the query as the first difficulty coefficient of the game level.

4. The method of claim 1, wherein, The acquisition of player behavior data during the game level's execution includes: Acquire candidate data of at least one of the following types during the operation of the game level: login data, operation data within the game level, level data, consumption data, and virtual resource consumption data; The candidate data is subjected to at least one of the following desensitization processes to obtain player behavior data: Gaussian noise-based noise addition; deletion; and encryption.

5. An information recommendation device for a game program, characterized by comprising: The device includes: The difficulty mapping module is used to determine a first difficulty coefficient of a game level based on a first player profile feature and a first mapping rule, wherein the first player profile feature is obtained before the game level runs in the game program; the first difficulty coefficient is updated based on a second player profile feature to obtain a second difficulty coefficient, and the game level is run with the second difficulty coefficient. The step of updating the first difficulty coefficient based on the second player profile feature to obtain the second difficulty coefficient is implemented in the following way: classifying the second player profile feature to obtain a second player type; obtaining a second mapping rule, wherein the second mapping rule includes a mapping relationship between the second player type and the difficulty coefficient adjustment value; querying the second mapping rule based on the second player type, and adding the queried difficulty coefficient adjustment value to the first difficulty coefficient to obtain the second difficulty coefficient; The data acquisition module is used to respond to a trigger operation for a game level, run the game level based on the first difficulty coefficient, and acquire player behavior data during the game level's operation; update the first player profile feature based on the player behavior data to obtain a second player profile feature. This updating of the first player profile feature based on the player behavior data to obtain the second player profile feature is achieved through the following methods: feature extraction from the player behavior data to obtain incremental behavior features; determining the first weight value corresponding to each dimension's sub-feature in the incremental behavior features and the first player profile feature based on an attention mechanism; normalizing the first weight value of each sub-feature to obtain a second weight value for each sub-feature; and performing a weighted summation of each sub-feature and its corresponding second weight value to obtain the second player profile feature. An ad recommendation module is used to encode the result and progress of the game level in response to the end of the game level, obtaining an encoded vector feature; concatenate the ad features of the candidate ads, the encoded vector feature, and the second player profile feature to obtain a concatenated feature; predict the prediction balance coefficient corresponding to each candidate ad based on the concatenated feature, wherein the prediction balance coefficient is a balance function value used to measure ad revenue and user retention rate; select the candidate ad corresponding to the largest prediction balance coefficient as the target ad, wherein the target ad is used to be displayed when the game level ends.

6. An electronic device, comprising: The electronic device includes: Memory is used to store executable instructions or computer programs. A processor, when executing computer-executable instructions or computer programs stored in the memory, implements the information recommendation method for the game program as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer program are executed by a processor, they implement the information recommendation method for the game program as described in any one of claims 1 to 4.

Images