Game processing method, device and electronic equipment

Abstract

The application provides a game processing method and device and electronic equipment, according to the operation behavior data and user attribute characteristics of the current time, the user side characteristic vector of the current time is determined; according to the achievement semantic coding characteristics and achievement attribute characteristics of each virtual achievement object, the achievement side feature matrix of each virtual achievement object at the current time is determined; the evaluation result of each virtual achievement object is determined in combination with the user side characteristic vector, and then the target virtual achievement object is recommended to the user. According to the user side characteristic vector and the achievement side feature matrix, the target virtual achievement object is recommended to the user, since the user side characteristic vector is determined according to the operation behavior data and user attribute characteristics of the user, and the achievement side feature matrix is determined according to the achievement semantic coding characteristics and achievement attribute characteristics of each virtual achievement object, more abundant characteristic information of the user side and the achievement side is integrated, so that personalized virtual achievement object recommendation can be quickly and accurately made to the user.

Description

Technical Field

[0001] This invention relates to the field of game technology, and in particular to a game processing method, apparatus, and electronic device. Background Technology

[0002] Achievement systems play a crucial role in MMORPGs. Virtual achievements should provide users with challenging goals. If the difficulty is appropriate, users gain a greater sense of accomplishment upon completion. If the difficulty is too high, users gradually lose confidence; if the difficulty is too low, it lacks challenge and eventually leads to user churn. Current technologies typically recommend virtual achievements by constructing user-side or item-side features. However, in achievement systems, the conditions for achieving virtual achievements and their prerequisites are diverse. Given the complexity of obtaining virtual achievements, current technologies struggle to quickly and accurately provide personalized virtual achievement recommendations to users. Summary of the Invention

[0003] The purpose of this invention is to provide a game processing method, apparatus, and electronic device to quickly and accurately recommend personalized virtual achievement objects to users.

[0004] This invention provides a game processing method. The game includes multiple virtual achievement objects, each corresponding to an achievement semantic encoding feature. The achievement semantic encoding feature characterizes the descriptive information and acquisition conditions of the achievement corresponding to the virtual achievement object. The method includes: determining the user-side feature vector at the current moment based on the user's operational behavior data and user attribute features during the game; determining the achievement-side feature matrix corresponding to each virtual achievement object at the current moment based on the achievement semantic encoding feature and achievement attribute features of each virtual achievement object; wherein the achievement attribute features include the inherent attribute information and completion statistics of the achievement corresponding to the virtual achievement object; determining the evaluation result of each virtual achievement object for the user based on the achievement-side feature matrix corresponding to each virtual achievement object at the current moment and the user-side feature vector at the current moment; and recommending target virtual achievement objects to the user based on the evaluation results of each virtual achievement object.

[0005] This invention provides a game processing device. The game includes multiple virtual achievement objects, each corresponding to an achievement semantic encoding feature. The achievement semantic encoding feature characterizes the descriptive information and acquisition conditions of the achievement corresponding to the virtual achievement object. The device includes: a first determining module, used to determine the user-side feature vector at the current moment based on the user's operational behavior data and user attribute features during the game; a second determining module, used to determine the achievement-side feature matrix corresponding to each virtual achievement object at the current moment based on the achievement semantic encoding feature and achievement attribute features of each virtual achievement object; wherein the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object; a third determining module, used to determine the evaluation result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix corresponding to each virtual achievement object at the current moment and the user-side feature vector at the current moment; and a recommendation module, used to recommend target virtual achievement objects to the user based on the evaluation results of each virtual achievement object.

[0006] The present invention provides an electronic device, including a processor and a memory, wherein the memory stores machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement any of the above-mentioned game processing methods.

[0007] The present invention provides a machine-readable storage medium storing machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions cause the processor to implement any of the above-mentioned game processing methods.

[0008] The game processing method, apparatus, and electronic device provided by this invention determine the user-side feature vector at the current moment based on the user's operational behavior data and user attribute characteristics during the game; determine the achievement-side feature matrix corresponding to each virtual achievement object at the current moment based on the achievement semantic encoding features and achievement attribute features of each virtual achievement object; wherein, the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object; determine the evaluation result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix corresponding to each virtual achievement object at the current moment and the user-side feature vector at the current moment; and recommend target virtual achievement objects to the user based on the evaluation results of each virtual achievement object. This method recommends target virtual achievement objects to the user based on the user-side feature vector and achievement-side feature matrix. Since the user-side feature vector is determined based on the user's operational behavior data and user attribute characteristics, and the achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and achievement attribute characteristics of each virtual achievement object, it integrates richer feature information from both the user and achievement sides, thereby enabling rapid and accurate personalized virtual achievement object recommendations to the user. Attached Figure Description

[0009] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0010] Figure 1 A schematic diagram of a virtual achievement object in a game provided by an embodiment of the present invention;

[0011] Figure 2 A flowchart of a game processing method provided in an embodiment of the present invention;

[0012] Figure 3 A schematic diagram illustrating the generation process of semantic coding features provided in an embodiment of the present invention;

[0013] Figure 4 A memory consumption comparison diagram provided for an embodiment of the present invention;

[0014] Figure 5 This is a schematic diagram of an online real-time monitoring and update model provided in an embodiment of the present invention;

[0015] Figure 6 This is a schematic diagram of the structure of a game processing device provided in an embodiment of the present invention;

[0016] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Achievement systems play a crucial role in MMORPGs (Massive Multiplayer Online Role-Playing Games). For example, they can enhance the game's appeal. As game content becomes more time-sensitive with version updates, achievement systems can record past moments, allowing users to develop an emotional connection to the game.

[0019] Achievements should provide users with challenging goals. If the difficulty of achievements is appropriate, users will gain a greater sense of accomplishment from completing the tasks. If the difficulty of achievements is too high, users will gradually lose confidence, and some users may even be afraid to try. On the other hand, if the difficulty is too low, it will obviously not be challenging, and over time, it will gradually lead to user churn. How to recommend virtual achievement objects with moderate difficulty and challenge to users in a personalized way is an urgent problem to be solved.

[0020] In related technologies, two methods can be used to recommend virtual achievement objects to users. One method is user-based collaborative filtering, which requires characterizing user profile features, based on the assumption that similar users may like the same items. However, in achievement systems, the conditions and prerequisites for achieving virtual achievements are diverse, making the construction of achievement-side features crucial. If only user-side features are constructed, the recommendations will be inaccurate, affecting the user experience. For example, if only basic user attribute features (such as occupation, gender, power score, etc.) are available, the collaborative filtering model lacks input of achievement features during training, preventing the model from learning the patterns corresponding to these features and resulting in insufficient recommendation accuracy.

[0021] Another approach is item-based collaborative filtering, which requires characterizing item-side features, based on the assumption that similar items may be liked by the same user. However, achievement acquisition conditions are numerous. Building features by finding corresponding logs for each condition is extremely complex and tedious. Some achievements have complex conditions requiring the combination of multiple tracking logs to obtain the feature. However, for logs spanning multiple days or those that are missing, it becomes impossible to accurately construct the feature, thus affecting the accuracy of the recommendation results. For example, the conditions for achieving a certain achievement might require associating four logs, which need to be linked through specific time, person, and event. If logs are lost or there is a long delay during log transmission, the feature construction will be inaccurate, thus reducing the accuracy of achievement recommendations.

[0022] The two collaborative filtering models mentioned above have not been used in game achievement systems. They also cannot handle situations where achievement acquisition conditions are complex, such as how to construct achievement features and provide personalized achievement recommendations to users. Furthermore, these two solutions are not suitable for scenarios with a large number of users or items, because once the scale becomes large, there will be (n*n) dimension and (n*n) dimension dot product calculations, which are very resource-intensive and require further optimization, where n represents the number of users.

[0023] To facilitate understanding of this embodiment, a game processing method disclosed in this embodiment of the invention will first be introduced. The game includes multiple virtual achievement objects, which may take the form of achievement cards, virtual badges, virtual pendants, level indicators, etc. Figure 1The image shown is a schematic diagram of a virtual achievement object in a game. Figure 1 The virtual achievement objects in this example are illustrated using achievement cards. The achievement card list includes multiple achievement cards. The types of virtual achievement objects typically differ across games, and the number of virtual achievement objects can be set according to actual game requirements. Each virtual achievement object corresponds to an achievement semantic encoding feature. This feature represents the description information and acquisition conditions of the achievement corresponding to the virtual achievement object. This semantic encoding feature can be a vector composed of multiple specific numbers. The description information can be used to describe the type of achievement, such as leisure, plot, or imperial examination. The acquisition conditions indicate the conditions or prerequisites that need to be met to acquire the achievement corresponding to the virtual achievement object. The conditions or prerequisites for each type of achievement are usually different. Figure 2 As shown, the method includes the following steps:

[0024] Step S202: Determine the user-side feature vector at the current moment based on the user's operation behavior data and user attribute characteristics during the game.

[0025] The aforementioned operational behavior data can include data corresponding to user clicks, favorites, or comments on virtual achievement objects; it mainly constructs user behavior interaction patterns based on user ID and behavior ID according to the timestamp size, where the timestamp represents the time when a specific user behavior occurs; user ID and behavior ID represent the behavior ID that the user ID is participating in at the current time. Experiments show that users with different tendencies exhibit different behaviors. For example, users who enjoy adventure and challenges are more willing to improve their competitiveness, such as upgrading skills and equipment; users who are inclined towards social interaction are more willing to spend time chatting in the game to expand their social relationships, etc. It is mainly used to describe users' short-term and dynamic preferences.

[0026] The aforementioned user attribute features may include basic user information, social preferences, device information, etc. These features can reflect the user's inherent long-term preferences, which are usually static and change slowly over time. The aforementioned "current moment" can be the current point in time or the current time period. For example, if the user-side feature vector is determined in real time based on user operation behavior data and user attribute features, then the current moment can be the current point in time. If the user-side feature vector is determined according to a certain period based on user operation behavior data and user attribute features, for example, if the period is 3 hours, then the current moment refers to the 3-hour time period corresponding to the current period.

[0027] In actual implementation, during gameplay, users may perform actions such as clicking to view, favorite, or comment on virtual achievements. These actions can be recorded as action data. Based on the action data at the current moment and pre-acquired user attribute features, a user-side feature vector can be constructed for the current moment. For example, each specific data value or attribute value can be normalized to determine each vector value of the user-side feature vector.

[0028] Step S204: Based on the achievement semantic encoding features and achievement attribute features of each virtual achievement object, determine the achievement-side feature matrix corresponding to each virtual achievement object at the current moment; wherein, the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object.

[0029] The aforementioned achievement attribute characteristics can include the inherent attribute information and completion count statistics of the virtual achievement object. The inherent attribute information can include basic attributes such as the achievement acquisition level, type (e.g., leisure, plot, imperial examination), whether there is an achievement reward, reward type, and whether there are sub-achievements. These characteristics are usually fixed. If an achievement system is designed in a game, corresponding virtual achievement objects will inevitably be designed, and the designed virtual achievement objects will inevitably have basic inherent achievement attributes. Of course, the inherent attribute information can also be designed according to the characteristics of the achievement itself. The aforementioned completion count statistics can include the number of achievements completed by the user at the current moment.

[0030] In actual implementation, the achievement semantic encoding features and achievement attribute features of each virtual achievement object can be concatenated according to a preset input format to construct the achievement-side feature matrix corresponding to each virtual achievement object at the current moment. The above-mentioned operation behavior data, user attribute features, as well as achievement description information, acquisition conditions and achievement attribute features can all be obtained from the corresponding logs. For example, user attribute features such as the user's occupation, gender, level, etc. can be obtained from the user's login logs.

[0031] Step S206: Determine the evaluation result of each virtual achievement object corresponding to the user at the current time based on the achievement-side feature matrix of each virtual achievement object at the current time and the user-side feature vector of the user at the current time.

[0032] After determining the user-side feature vector and achievement-side feature matrix at the current moment, a pre-optimized collaborative filtering algorithm can be used to calculate the evaluation result for each virtual achievement object corresponding to the user. The evaluation result for each virtual achievement object is usually different. The evaluation result is generally in numerical form. Generally, the higher the value of the evaluation result, the higher the matching degree between the virtual achievement object and the user. The lower the value of the evaluation result, the lower the matching degree between the virtual achievement object and the user.

[0033] Step S208: Recommend target virtual achievement objects to the user based on the evaluation results of each virtual achievement object.

[0034] The number of the aforementioned target virtual achievement objects can be one or more. In actual implementation, after obtaining the evaluation result corresponding to each virtual achievement object, since the evaluation result corresponding to each virtual achievement object is usually different, one or more target virtual achievement objects with a high degree of matching with the user can be recommended to the user based on each evaluation result.

[0035] The aforementioned game processing method determines the user's current-moment user-side feature vector based on the user's operational behavior data and user attribute characteristics during the game; it then determines the achievement-side feature matrix for each virtual achievement object based on its achievement semantic encoding features and achievement attribute features. The achievement attribute features include the inherent attribute information and completion statistics of the virtual achievement object's corresponding achievement. Finally, it determines the evaluation result for each virtual achievement object based on the achievement-side feature matrix and the user's current-moment user-side feature vector; and recommends target virtual achievement objects to the user based on the evaluation results of each virtual achievement object. This method recommends target virtual achievement objects to the user based on the user-side feature vector and achievement-side feature matrix. Since the user-side feature vector is determined based on the user's operational behavior data and user attribute characteristics, and the achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and achievement attribute characteristics of each virtual achievement object, it integrates richer feature information from both the user and achievement sides, thus enabling rapid and accurate personalized virtual achievement object recommendations to the user.

[0036] In an optional implementation, the method is based on the method of the above embodiments. In this method, the game includes multiple virtual achievement objects, each corresponding to an achievement semantic encoding feature. The achievement semantic encoding feature characterizes the descriptive information and acquisition conditions of the achievement corresponding to the virtual achievement object. The achievement semantic encoding feature is obtained in the following manner:

[0037] The description information and acquisition conditions of the achievement corresponding to each virtual achievement object are input into a pre-built feature extractor to obtain the achievement semantic encoding features corresponding to each virtual achievement object output by the feature extractor.

[0038] NLP (Natural Language Processing) uses AI (Artificial Intelligence) to process and analyze text or speech data in order to understand and interpret content, classify content, or gain insights from content. Pre-trained NLP models (such as BERT) can be used to fine-tune feature extractors from game-related corpora (such as text descriptions of plot, tasks, gameplay, and skills) and corpora related to the target task scenario (such as achievement names, achievement descriptions, and achievement acquisition conditions). These feature extractors can map a piece of text input to them into a specific dimension of achievement semantic encoding vector corresponding to that text.

[0039] See Figure 3 The diagram illustrates the generation process of achievement semantic coding features. In actual implementation, unstructured data such as achievement name, achievement description, and achievement acquisition conditions corresponding to each virtual achievement object can be concatenated according to the input format required by the feature extractor, and then input into the constructed BERT feature extractor. The BERT feature extractor can map the unstructured data into the corresponding structured data, that is, output the achievement semantic coding vector corresponding to each virtual achievement object (corresponding to the achievement semantic coding features mentioned above). The specific mapping method can be mapped according to the semantic information of each word in the input unstructured data. For example, "I am very happy" is a positive statement and can be mapped to 1, "Today is really unlucky" is a negative statement and can be mapped to -1, "I just finished eating" is a neutral statement and can be mapped to 0, etc. Of course, the mapping relationship can also be set according to actual needs, which is not limited here.

[0040] For example, as shown in Table 1, the description information and acquisition conditions of an achievement are as follows: Achievement 1 is named AA, described as BB, and acquired under the condition CC. The input to the feature extractor could then be: AA BB CC; The output of the feature extractor is [0.111, 0.225, 0.882, 0.992, ..., 0.447], which is a b-dimensional vector, where b can be any number, specifically related to the attribute dimensions of the achievement description information and acquisition conditions. Each vector value in the output achievement semantic encoding features is also associated with the achievement description information and acquisition conditions.

[0041] Table 1

[0042] Achievement name describe Acquisition conditions 1 AA BB CC

[0043] In game achievement systems, the conditions for obtaining achievements are complex and numerous. Focusing solely on the basic attributes of achievements would miss features of unstructured data, such as the acquisition conditions and descriptions. Developing corresponding features for each achievement's acquisition conditions would be extremely time-consuming and labor-intensive. This embodiment utilizes an NLP pre-trained model to extract features from achievement descriptions and acquisition conditions, obtaining semantic encoding vectors for achievements. This enriches the feature descriptions of achievements and avoids the problem of inaccurate achievement recommendations due to missing important features. For example, an achievement called "Difficult Road Travel" grants an extra coin reward. By incorporating features related to this achievement (such as whether an extra reward is given), a reward feature can be constructed and input into a collaborative filtering model. This model can better learn which users are more likely to want such extra rewards, thus recommending achievements with extra rewards to its target users.

[0044] The method includes the following steps:

[0045] Step 1: Determine the user-side feature vector at the current moment based on the user's operational behavior data and user attribute characteristics during the game.

[0046] Step 2: Based on the achievement semantic encoding features and achievement attribute features of each virtual achievement object, determine the achievement-side feature matrix corresponding to each virtual achievement object at the current moment; wherein, the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object.

[0047] Step 3: Obtain the historical achievement-side feature matrix and the historical user-side feature vector of each virtual achievement object at the current time corresponding to the previous virtual achievement object recommendation time. The historical achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and historical achievement attribute features of the virtual achievement object. The historical user-side feature vector is determined based on the historical operation behavior data and user attribute features of the user at the previous virtual achievement object recommendation time. The historical achievement attribute features include the inherent attribute information of the achievement corresponding to the virtual achievement object and the statistical information of the number of historical completions.

[0048] To recommend target virtual achievements to users more comprehensively and accurately, historical features can be incorporated. For example, user-side feature vectors can be determined based on user behavior data and user attribute characteristics at regular intervals, with each interval being daily. If a user opened the achievement system on January 2nd and clicked, viewed, liked, or commented on virtual achievements, these behavioral data from the 2nd can be statistically analyzed. Combined with user attribute characteristics, the user-side feature vector for the 2nd can be determined. Based on the achievement semantic encoding features and achievement attribute features of the virtual achievements, the achievement-side feature matrix for the 2nd can be determined. Based on the user-side feature vector and achievement-side feature matrix for the 2nd, the evaluation result for each virtual achievement corresponding to the 2nd can be determined. These real-time features allow for faster perception of changes in user behavior and interests, improving the real-time nature of recommendations. Then, the historical achievement-side feature matrix and historical user-side feature vector from January 1st can be combined to recommend virtual achievements. The historical achievement-side feature matrix and historical user-side feature vector can be determined using the same method as the current-moment user-side feature vector and achievement-side feature matrix, which will not be elaborated further here.

[0049] For example, an a-dimensional vector can be constructed based on the achievement's level, type (e.g., leisure, storyline, imperial examination), whether there is an achievement reward, the type of reward, whether there are sub-achievements, and the number of times a certain achievement has been completed within 7 or 14 days. This a-dimensional vector can then be concatenated with achievement semantic encoding features (such as the b-dimensional vector mentioned above) to obtain I. m*(a+b) The achievement feature matrix, where m represents the number of achievements, is the historical achievement feature matrix corresponding to the user.

[0050] Step 4: Based on the historical achievement-side feature matrix, the historical user-side feature vector, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment, determine the evaluation result for each virtual achievement object corresponding to the user.

[0051] Step four can be achieved through the following steps 30 to 33:

[0052] Step 30: Based on the historical completion statistics, construct a first sparse vector and a second sparse vector; wherein, the first sparse vector is used to represent the completion information of the user for each virtual achievement object in the first time period before the current time; the second sparse vector is used to represent the completion information of the user for each virtual achievement object in the second time period before the current time; the first time period is shorter than the second time period.

[0053] For example, a matrix can be constructed using the data of m achievements a user has acquired in the past 30 days. For each user, based on their historical achievement statistics, a first sparse vector and a second sparse vector are constructed. The two sparse vectors correspond to different time periods; for example, the first sparse vector might correspond to a 14-day time period, and the second sparse vector to a 30-day time period, etc. The specific time period can be chosen according to actual needs. Assuming there are n users, the first sparse vectors corresponding to each of the n users constitute the first sparse matrix, and the second sparse vectors corresponding to each of the n users constitute the second sparse matrix. That is, in this embodiment, two sparse matrices are constructed: one is a short-term sparse matrix (e.g., the achievement data from the past 14 days) A1. n*m The other is the long-term sparse matrix A2. n*m (For example, achievements acquired in the last 30 days). The values ​​of these two sparse matrices are 0 or 1, where 1 indicates that the achievement has been acquired and 0 indicates that the achievement has not been acquired.

[0054] Step 31: Construct a historical operation behavior vector based on the user's operation behavior data in the second time period before the current moment.

[0055] Since users may click to view achievements or collect achievements in the game achievement system, taking a second time period of 30 days as an example, a historical operation behavior vector can be constructed based on the user's operation behavior data in the past 30 days. Assuming there are n users, the historical operation behavior vectors of each of the n users constitute a historical operation behavior matrix. This historical operation behavior matrix is ​​usually in the form of a sparse matrix, where the value of the sparse matrix is ​​0 or 1, where 1 indicates that the operation has occurred and 0 indicates that the operation has not occurred.

[0056] Step 32: Determine the historical results of each virtual achievement object corresponding to the user based on the historical achievement-side feature matrix, historical user-side feature vector, first sparse vector, second sparse vector, and historical operation behavior vector.

[0057] Step 33: Based on the historical results, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment, determine the evaluation result for each virtual achievement object corresponding to the user.

[0058] This step 33 can be achieved through the following steps A to D:

[0059] Step A: Construct a third sparse vector based on the current completion statistics; whereby the third sparse vector is used to represent the completion information of the user for each virtual achievement object at the current moment.

[0060] Specifically, a matrix can be constructed using the current status of a user acquiring m achievements. For each user, a third sparse vector is constructed based on the completion status of each virtual achievement object at the current time. Assuming there are n users, the third sparse vectors of the n users constitute a third sparse matrix. The value of the third sparse matrix is ​​either 0 or 1, where 1 indicates that the achievement has been acquired at the current time, and 0 indicates that the achievement has not been acquired at the current time.

[0061] Step B: Construct the current operation behavior vector based on the user's operation behavior data at the current moment.

[0062] Since users may have actions such as clicking to view achievements or adding items to their favorites at the current moment, a current action vector can be constructed based on the user's action data at the current moment. Assuming there are n users, the current action vectors of each of the n users constitute a current action matrix. This current action matrix is ​​usually in the form of a sparse matrix, where the value of the sparse matrix is ​​0 or 1, where 1 indicates that the action has occurred and 0 indicates that the action has not occurred.

[0063] Step C: Based on the achievement-side feature matrix, the user-side feature vector, the third sparse vector, and the current operation behavior vector at the current moment, determine the current result of each virtual achievement object corresponding to the user.

[0064] Step D: Based on historical and current results, determine the evaluation result for each virtual achievement object corresponding to the user.

[0065] Step D may specifically include: performing calibration calculations on historical and current results according to a preset calibration method to obtain the evaluation results for each virtual achievement object corresponding to the user.

[0066] For example, if the score corresponding to the current result is s1 and the score corresponding to the historical result is s2, the historical result and the current result can be calculated according to the calibration formula ctr=αlog(s1+1)+βlog(s2+1). The calculation result is used as the evaluation result for each virtual achievement object corresponding to the user, where α and β are coefficients and α+β=1. By calibrating the historical result and the current result, a more accurate evaluation result can be obtained, thereby improving the accuracy of the target virtual achievement object recommendation.

[0067] The historical operation behavior vector includes: historical touch behavior vector and historical favorite behavior vector; the current operation behavior vector includes: current touch behavior vector and current favorite behavior vector.

[0068] The formula for the traditional collaborative filtering model is CF = U n*p*U T p*n *A n*m This process will first calculate U n*p *U T p*n And U n*p *U T p*n The calculation result is a large-scale dense matrix of n*n dimensions. In the game scenario, n represents the number of users, which is enormous. This means that U n*p *U T p*n This would create a high-dimensional matrix, which is very memory-intensive and computationally slow; this embodiment optimizes this by cleverly utilizing the matrix associative law: First calculate This result creates a p*m dimensional matrix, where p represents the dimension of user attribute features and m represents the number of virtual achievement objects included in the game. The magnitudes of p and m are much smaller than n. This optimizes the collaborative filtering algorithm by introducing sparse matrices and matrix associativity, avoiding the construction of large-scale, high-dimensional dense matrices, thus reducing memory consumption and significantly improving runtime efficiency. See also Figure 4 The diagram shown illustrates a comparison of memory consumption. It can be seen that the traditional collaborative filtering algorithm requires 54GB of memory in a certain game. After optimization in this embodiment, it only requires 27GB of memory, saving 50% of memory and greatly avoiding resource waste.

[0069] In a game achievement system, it is not only necessary to consider user attributes, but more importantly, achievement-side attributes. Therefore, this invention optimizes the collaborative filtering algorithm by combining user-side features and achievement-side features. Furthermore, based on the characteristics of users' long-term and short-term behaviors (long-term behavior reflects user behavior trends, and short-term behavior reflects user recent behavior tendencies) and other behaviors in the game, such as click behavior and collection behavior, different sparse matrices (long-term behavior A1 matrix, short-term behavior A2 matrix, click behavior B and collection behavior C) are designed to obtain a high-order linear collaborative filtering algorithm.

[0070] Specifically, both the historical and current results mentioned above are calculated using the following formula:

[0071] lin erCF=αU cf +βI cf ;

[0072] Ucf=x1U n*p (U T p*n *A1 n*m )+x2U n*p (U Tp*n *A2 n*m )+x3U n*p (U T p*n *B n*m )+x4U n*p (U T p*n *C n*m );

[0073] I cf =γ1(A1 n*m I m*(a+b) )I T (a+b)*m +γ2(A2 n*m I m*(a+b) )I T (a+b)*m ;

[0074] lin erCF represents the matrix consisting of the historical results of each virtual achievement object corresponding to n users, U n*p A1 represents a matrix composed of the historical user-side feature vectors of n users; n*m A2 represents a sparse matrix composed of the first sparse vectors corresponding to each of the n users; n*m B represents the sparse matrix formed by the second sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the historical touch behavior vectors of n users on m virtual achievement objects; n*m I represents a matrix consisting of vectors representing the historical collection behaviors of n users on m virtual achievement objects; m*(a+b) Represents the feature matrix of historical achievements; or,

[0075] lin erCF represents the matrix consisting of the current results of each virtual achievement object corresponding to n users, U n*p A1 represents a matrix consisting of the user-side feature vectors of n users at the current time. n*m and A2 n*m B is a sparse matrix composed of the third sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the current touch behavior vectors of n users on m virtual achievement objects; n*m I represents a matrix consisting of vectors representing the current collection actions of n users on m virtual achievement objects; m*(a+b) This represents the achievement-side feature matrix corresponding to the current moment;

[0076] Where p represents the dimension of user attribute features; m represents the number of virtual achievement objects included in the game; a represents the dimension of achievement semantic encoding features for each virtual achievement object; b represents the dimension of achievement attribute features for each virtual achievement object; α, β, x1-x4, γ1 and γ2 are all coefficients.

[0077] In actual implementation, the coefficients α, β, x1-x4, γ1, and γ2 can be optimized according to the actual application scenario, and α+β=1, x1+x2+x3+x4=1; γ1+γ2=1; preferably, in the game scenario, the optimal coefficient selection is α=β=1 / 2; x1=x2=1 / 3; x3=x4=1 / 6; γ1=γ2=1 / 2.

[0078] For ease of understanding, see Figure 5 The diagram illustrates an online real-time monitoring and update model. First, let's explain the relevant terms: Kafka is a high-throughput distributed publish-subscribe messaging system that can handle all user action data on a website. This data is typically handled through log processing and log aggregation due to throughput requirements. Redis (Remote Dictionary Server) is an open-source, network-enabled, in-memory or persistent log-structured key-value database written in ANSI C, providing APIs (Application Programming Interfaces) in multiple languages. Hive is a data warehouse tool based on Hadoop, used for data extraction, transformation, and loading. It's a mechanism for storing, querying, and analyzing large-scale data stored in Hadoop. Flink is an open-source stream processing framework that executes arbitrary data stream programs in a data-parallel and pipelined manner. Flink's pipeline runtime system can execute both batch and stream processing programs.

[0079] Compared to the real-time performance of features, the real-time performance of recommendation models often considers the problem from a more holistic perspective. The real-time performance of features aims to more accurately describe a user, thereby enabling the recommendation system to provide recommendations that are more suitable for that user. The specific operation is as follows:

[0080] Real-time data is fed into Kafka and processed using Flink. This real-time data includes user achievement acquisition behavior, click behavior, and favorites behavior within the past three hours. These behaviors are developed into real-time features and written to Redis according to a specified format. The model is then incrementally updated. The score generated by the real-time updated model is s1 (corresponding to the current result). Using historical offline features from the Hive distributed database, such as user attribute features and achievement attribute features, the previously updated offline score is obtained as s2 (corresponding to the historical result). The final score of the model is obtained using the calibration formula CTR = αlog(s1+1) + βlog(s2+1), where α + β = 1. The latest recommendation results are synchronized to the Redis database, improving the accuracy of achievement recommendations and providing users with more timely achievements. By introducing real-time features into the model, changes in user behavior and interests can be detected more quickly, allowing for real-time model updates and improved recommendation timeliness. It should be noted that the model mentioned above is the collaborative filtering model implemented using the optimized collaborative filtering algorithm described above.

[0081] Step 5: Recommend target virtual achievement objects to users based on the evaluation results of each virtual achievement object.

[0082] This fifth step can be achieved through the following steps 40 and 41:

[0083] Step 40: Sort each virtual achievement object according to the order of each evaluation result from high to low, and obtain the sorted virtual achievement objects.

[0084] Step 41: Based on the sorted virtual achievement objects, recommend target virtual achievement objects to the user.

[0085] Each virtual achievement object typically has a different score. These virtual achievements can be sorted from highest to lowest score. A higher score indicates a higher match between the virtual achievement object and the user, while a lower score indicates a lower match. Therefore, a number of target virtual achievements can be recommended to the user in descending order of score. For example, if a game contains 30 virtual achievements and you need to recommend 10 cards to the user, you can sort the 30 virtual achievements from highest to lowest score and select the top 10 with the highest scores as the target virtual achievements to recommend to the user.

[0086] The aforementioned game processing method is applicable to all online game achievement systems. Even in systems where achievement acquisition conditions are complex and diverse, it can quickly and effectively construct achievement features. It also solves the problem of high-dimensional dense matrices leading to huge memory consumption in existing collaborative filtering models, resulting in excellent performance and practical application value in both engineering development efficiency and algorithm prediction. Furthermore, the feature construction in this approach is multi-source, including user attribute features, operational behavior data, achievement attribute features, and achievement semantic encoding features. This enriches the feature construction, allowing the model to learn and train better. Through scientifically designed achievement features, it can personalizedly recommend virtual achievement objects to users, increasing their self-efficacy and enhancing their interest in the game as they achieve achievements.

[0087] This disclosure also provides a game processing device. The game includes multiple virtual achievement objects, each virtual achievement object corresponding to an achievement semantic encoding feature. The achievement semantic encoding feature characterizes the descriptive information and acquisition conditions of the achievement corresponding to the virtual achievement object. Figure 6 As shown, the device includes: a first determining module 60, used to determine the user-side feature vector of the user at the current moment based on the user's operation behavior data and user attribute characteristics during the game; a second determining module 61, used to determine the achievement-side feature matrix of each virtual achievement object at the current moment based on the achievement semantic encoding features and achievement attribute features of each virtual achievement object; wherein, the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object; a third determining module 62, used to determine the evaluation result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix of each virtual achievement object at the current moment and the user-side feature vector of the user at the current moment; and a recommendation module 63, used to recommend target virtual achievement objects to the user based on the evaluation results of each virtual achievement object.

[0088] The aforementioned game processing device determines the user's current-moment user-side feature vector based on the user's operational behavior data and user attribute characteristics during the game; it then determines the achievement-side feature matrix for each virtual achievement object based on its achievement semantic encoding features and achievement attribute features. The achievement attribute features include the inherent attribute information and completion statistics of the virtual achievement object. Finally, based on the achievement-side feature matrix and the user's current-moment user-side feature vector, the device determines the evaluation result for each virtual achievement object for the user; and based on the evaluation results, it recommends target virtual achievement objects to the user. This device recommends target virtual achievement objects to the user based on the user-side feature vector and achievement-side feature matrix. Since the user-side feature vector is determined based on the user's operational behavior data and user attribute characteristics, and the achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and achievement attribute characteristics of each virtual achievement object, it integrates richer feature information from both the user and achievement sides, thus enabling rapid and accurate personalized virtual achievement object recommendations to the user.

[0089] In an optional embodiment, the device further includes a semantic coding feature acquisition module, through which achievement semantic coding features are obtained: the description information of the achievement corresponding to each virtual achievement object and the acquisition conditions are input to a pre-built feature extractor to obtain the achievement semantic coding features corresponding to each virtual achievement object output by the feature extractor.

[0090] In an optional implementation, the third determining module 62 is further configured to: obtain the historical achievement-side feature matrix corresponding to the previous virtual achievement object recommendation time at the current time for each virtual achievement object, and the historical user-side feature vector corresponding to the previous virtual achievement object recommendation time; wherein, the historical achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and historical achievement attribute features of the virtual achievement object; the historical user-side feature vector is determined based on the historical operation behavior data and user attribute features corresponding to the previous virtual achievement object recommendation time; the historical achievement attribute features include the inherent attribute information of the achievement corresponding to the virtual achievement object and the historical completion count statistics; and determine the evaluation result of each virtual achievement object corresponding to the user based on the historical achievement-side feature matrix, the historical user-side feature vector, the achievement-side feature matrix corresponding to the current time, and the user-side feature vector at the current time.

[0091] In an optional implementation, the third determining module 62 is further configured to: construct a first sparse vector and a second sparse vector based on historical completion statistics; wherein the first sparse vector is used to represent the completion information of the user for each virtual achievement object within a first time period before the current moment; the second sparse vector is used to represent the completion information of the user for each virtual achievement object within a second time period before the current moment; the first time period is shorter than the second time period; construct a historical operation behavior vector based on the user's operation behavior data within the second time period before the current moment; determine the historical result of each virtual achievement object corresponding to the user based on the historical achievement-side feature matrix, the historical user-side feature vector, the first sparse vector, the second sparse vector, and the historical operation behavior vector; and determine the evaluation result of each virtual achievement object corresponding to the user based on the historical result, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment.

[0092] In an optional implementation, the third determining module 62 is further configured to: construct a third sparse vector based on the current completion count statistics; wherein the third sparse vector is used to represent the completion information of the user for each virtual achievement object at the current time; construct a current operation behavior vector based on the user's operation behavior data at the current time; determine the current result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix corresponding to the current time, the user-side feature vector at the current time, the third sparse vector, and the current operation behavior vector; and determine the evaluation result of each virtual achievement object corresponding to the user based on the historical results and the current results.

[0093] In an optional implementation, the historical operation behavior vector includes: historical touch behavior vector and historical favorite behavior vector; the current operation behavior vector includes: current touch behavior vector and current favorite behavior vector;

[0094] Both historical and current results are calculated using the following formula:

[0095] lin erCF=αU cf +βI cf ;

[0096] Ucf=x1U n*p (U T p*n *A1 n*m )+x2U n*p (U T p*n *A2 n*m )+x3U n*p (U T p*n *B n*m )+x4Un*p (U T p*n *C n*m );

[0097] I cf =γ1(A1 n*m I m*(a+b) )I T (a+b)*m +γ2(A2 n*m I m*(a+b) )I T (a+b)*m ;

[0098] lin erCF represents the matrix consisting of the historical results of each virtual achievement object corresponding to n users, U n*p A1 represents a matrix composed of the historical user-side feature vectors of n users; n*m A2 represents a sparse matrix composed of the first sparse vectors corresponding to each of the n users; n*m B represents the sparse matrix formed by the second sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the historical touch behavior vectors of n users on m virtual achievement objects; n*m I represents a matrix consisting of vectors representing the historical collection behaviors of n users on m virtual achievement objects; m*(a+b) Represents the feature matrix of historical achievements; or,

[0099] lin erCF represents the matrix consisting of the current results of each virtual achievement object corresponding to n users, U n*p A1 represents a matrix consisting of the user-side feature vectors of n users at the current time. n*m and A2 n*m B is a sparse matrix composed of the third sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the current touch behavior vectors of n users on m virtual achievement objects; n*m I represents a matrix consisting of vectors representing the current collection actions of n users on m virtual achievement objects; m*(a+b) This represents the achievement-side feature matrix corresponding to the current moment;

[0100] Where p represents the dimension of user attribute features; m represents the number of virtual achievement objects included in the game; a represents the dimension of achievement semantic encoding features for each virtual achievement object; b represents the dimension of achievement attribute features for each virtual achievement object; and x1-x4, γ1, and γ2 are all coefficients.

[0101] In an optional implementation, the third determining module 62 is further configured to: perform calibration calculations on the historical results and the current results according to a preset calibration method to obtain the evaluation results for each virtual achievement object corresponding to the user.

[0102] In an optional implementation, the recommendation module 63 is further configured to: sort each virtual achievement object according to the order of each evaluation result from high to low, to obtain sorted virtual achievement objects; and recommend target virtual achievement objects to the user based on the sorted virtual achievement objects.

[0103] The game processing device provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned game processing method embodiment. For the sake of brevity, any parts not mentioned in the game processing device embodiment can be referred to the corresponding content in the aforementioned game processing method embodiment.

[0104] This disclosure also provides an electronic device, see [link to relevant documentation] Figure 7 As shown, the electronic device includes a processor 160 and a memory 161. The memory 161 stores machine-executable instructions that can be executed by the processor 130. The processor 160 executes the machine-executable instructions to implement the above-described game processing method.

[0105] Furthermore, Figure 7 The electronic device shown also includes a bus 162 and a communication interface 163, with the processor 160, the communication interface 163 and the memory 161 connected via the bus 162.

[0106] The memory 161 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 163 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network. The bus 162 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0107] Processor 160 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 160 or by instructions in software form. Processor 160 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 161. Processor 160 reads information from memory 161 and, in conjunction with its hardware, completes the steps of the method described in the preceding embodiment. The game includes multiple virtual achievement objects, each corresponding to an achievement semantic encoding feature; wherein, the achievement semantic encoding feature characterizes the description information and acquisition conditions of the achievement corresponding to the virtual achievement object; specifically, the following steps are executed:

[0108] Based on the user's current action behavior data and user attribute characteristics during the game, determine the user's current user-side feature vector; based on the achievement semantic encoding features and achievement attribute features of each virtual achievement object, determine the achievement-side feature matrix corresponding to each virtual achievement object at the current moment; wherein, the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object; based on the achievement-side feature matrix corresponding to each virtual achievement object at the current moment and the user's current user-side feature vector, determine the evaluation result of each virtual achievement object for the user; based on the evaluation result of each virtual achievement object, recommend target virtual achievement objects to the user.

[0109] The specific implementation methods and working processes can be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0110] In an optional implementation, when the processor performs the acquisition of achievement semantic coding features, it is specifically used to: input the description information of the achievement corresponding to each virtual achievement object and the acquisition conditions into a pre-built feature extractor, and obtain the achievement semantic coding features corresponding to each virtual achievement object output by the feature extractor.

[0111] In an optional implementation, when the processor determines the evaluation result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix corresponding to each virtual achievement object at the current time and the user-side feature vector at the current time, it specifically performs the following steps: obtaining the historical achievement-side feature matrix corresponding to the previous virtual achievement object recommendation time for each virtual achievement object at the current time, and the historical user-side feature vector corresponding to the previous virtual achievement object recommendation time; wherein, the historical achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and historical achievement attribute features of the virtual achievement object; the historical user-side feature vector is determined based on the historical operation behavior data corresponding to the previous virtual achievement object recommendation time and the user attribute features of the user; the historical achievement attribute features include the inherent attribute information of the achievement corresponding to the virtual achievement object and the historical completion count statistics; and determining the evaluation result of each virtual achievement object corresponding to the user based on the historical achievement-side feature matrix, the historical user-side feature vector, the achievement-side feature matrix corresponding to the current time, and the user-side feature vector at the current time.

[0112] In an optional implementation, when the processor determines the evaluation result of each virtual achievement object corresponding to a user based on the historical achievement-side feature matrix, historical user-side feature vector, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment, the specific steps are as follows: Constructing a first sparse vector and a second sparse vector based on historical completion statistics; wherein the first sparse vector represents the completion information of the user for each virtual achievement object within a first time period before the current moment; the second sparse vector represents the completion information of the user for each virtual achievement object within a second time period before the current moment; the first time period is shorter than the second time period; constructing a historical operation behavior vector based on the user's operation behavior data within the second time period before the current moment; determining the historical result of each virtual achievement object corresponding to the user based on the historical achievement-side feature matrix, historical user-side feature vector, first sparse vector, second sparse vector, and historical operation behavior vector; and determining the evaluation result of each virtual achievement object corresponding to the user based on the historical result, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment.

[0113] In an optional implementation, when the processor determines the evaluation result of each virtual achievement object corresponding to the user based on historical results, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment, it specifically performs the following steps: constructing a third sparse vector based on the completion quantity statistics at the current moment; wherein the third sparse vector is used to represent the completion information of the user for each virtual achievement object at the current moment; constructing a current operation behavior vector based on the user's operation behavior data at the current moment; determining the current result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix corresponding to the current moment, the user-side feature vector at the current moment, the third sparse vector, and the current operation behavior vector; and determining the evaluation result of each virtual achievement object corresponding to the user based on historical results and current results.

[0114] In an optional implementation, the historical operation behavior vector includes: historical touch behavior vector and historical favorite behavior vector; the current operation behavior vector includes: current touch behavior vector and current favorite behavior vector;

[0115] Both historical and current results are calculated using the following formula:

[0116] lin erCF=αU cf +βI cf ;

[0117] Ucf=x1U n*p (UT p*n *A1 n*m )+x2U n*p (UT p*n *A2 n*m )+x3U n*p (U T p*n *B n*m )+x4U n*p (U T p*n *C n*m );

[0118] I cf =γ1(A1 n*m I m*(a+b) )I T (a+b)*m +γ2(A2 n*m I m*(a+b) )I T (a+b)*m ;

[0119] lin erCF represents the matrix consisting of the historical results of each virtual achievement object corresponding to n users, U n*p A1 represents a matrix composed of the historical user-side feature vectors of n users;n*m A2 represents the sparse matrix formed by the first sparse vectors corresponding to each of the n users. n*m B represents the sparse matrix formed by the second sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the historical touch behavior vectors of n users on m virtual achievement objects; n*m I represents a matrix consisting of vectors representing the historical collection behaviors of n users on m virtual achievement objects; m*(a+b) Represents the feature matrix of historical achievements; or,

[0120] lin erCF represents the matrix consisting of the current results of each virtual achievement object corresponding to n users, U n*p A1 represents a matrix consisting of the user-side feature vectors of n users at the current time. n*m and A2 n*m B is a sparse matrix composed of the third sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the current touch behavior vectors of n users on m virtual achievement objects; n*m I represents a matrix consisting of vectors representing the current collection actions of n users on m virtual achievement objects; m*(a+b) This represents the achievement-side feature matrix corresponding to the current moment;

[0121] Where p represents the dimension of user attribute features; m represents the number of virtual achievement objects included in the game; a represents the dimension of achievement semantic encoding features for each virtual achievement object; b represents the dimension of achievement attribute features for each virtual achievement object; α, β, x1-x4, γ1 and γ2 are all coefficients.

[0122] In an optional implementation, when the processor determines the evaluation result of each virtual achievement object corresponding to the user based on historical results and current results, it specifically performs calibration calculations on the historical results and current results according to a preset calibration method to obtain the evaluation result of each virtual achievement object corresponding to the user.

[0123] In an optional implementation, when the processor recommends a target virtual achievement object to the user based on the evaluation results of each virtual achievement object, it specifically performs the following steps: sorting each virtual achievement object in descending order of each evaluation result to obtain sorted virtual achievement objects; and recommending a target virtual achievement object to the user based on the sorted virtual achievement objects.

[0124] The above method determines the user's current-moment user-side feature vector based on the user's operational behavior data and user attribute characteristics during gameplay. It then determines the achievement-side feature matrix for each virtual achievement object based on its achievement semantic encoding and achievement attribute characteristics. The achievement attribute characteristics include the inherent attribute information and completion statistics of the virtual achievement object. Finally, it determines the evaluation result for each virtual achievement object based on the achievement-side feature matrix and the user's current-moment user-side feature vector. Based on the evaluation results of each virtual achievement object, it recommends target virtual achievement objects to the user. This method recommends target virtual achievement objects to the user based on the user-side feature vector and achievement-side feature matrix. Since the user-side feature vector is determined based on the user's operational behavior data and user attribute characteristics, and the achievement-side feature matrix is ​​determined based on the achievement semantic encoding and achievement attribute characteristics of each virtual achievement object, it integrates richer feature information from both the user and achievement sides, enabling rapid and accurate personalized virtual achievement object recommendations.

[0125] This disclosure also provides a machine-readable storage medium storing machine-executable instructions that, when invoked and executed by a processor, cause the processor to perform the following steps:

[0126] Based on the user's current action behavior data and user attribute characteristics during the game, determine the user's current user-side feature vector; based on the achievement semantic encoding features and achievement attribute features of each virtual achievement object, determine the achievement-side feature matrix corresponding to each virtual achievement object at the current moment; wherein, the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object; based on the achievement-side feature matrix corresponding to each virtual achievement object at the current moment and the user's current user-side feature vector, determine the evaluation result of each virtual achievement object for the user; based on the evaluation result of each virtual achievement object, recommend target virtual achievement objects to the user.

[0127] The specific implementation methods and working processes can be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0128] In an optional implementation, when the processor performs the acquisition of semantic coding features, it specifically performs the following: inputting the description information of the achievement corresponding to each virtual achievement object and the acquisition conditions into a pre-built feature extractor to obtain the achievement semantic coding features corresponding to each virtual achievement object output by the feature extractor.

[0129] In an optional implementation, when the processor determines the evaluation result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix corresponding to each virtual achievement object at the current time and the user-side feature vector at the current time, it specifically performs the following steps: obtaining the historical achievement-side feature matrix corresponding to the previous virtual achievement object recommendation time for each virtual achievement object at the current time, and the historical user-side feature vector corresponding to the previous virtual achievement object recommendation time; wherein, the historical achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and historical achievement attribute features of the virtual achievement object; the historical user-side feature vector is determined based on the historical operation behavior data corresponding to the previous virtual achievement object recommendation time and the user attribute features of the user; the historical achievement attribute features include the inherent attribute information of the achievement corresponding to the virtual achievement object and the historical completion count statistics; and determining the evaluation result of each virtual achievement object corresponding to the user based on the historical achievement-side feature matrix, the historical user-side feature vector, the achievement-side feature matrix corresponding to the current time, and the user-side feature vector at the current time.

[0130] In an optional implementation, when the processor determines the evaluation result of each virtual achievement object corresponding to a user based on the historical achievement-side feature matrix, historical user-side feature vector, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment, the specific steps are as follows: Constructing a first sparse vector and a second sparse vector based on historical completion statistics; wherein the first sparse vector represents the completion information of the user for each virtual achievement object within a first time period before the current moment; the second sparse vector represents the completion information of the user for each virtual achievement object within a second time period before the current moment; the first time period is shorter than the second time period; constructing a historical operation behavior vector based on the user's operation behavior data within the second time period before the current moment; determining the historical result of each virtual achievement object corresponding to the user based on the historical achievement-side feature matrix, historical user-side feature vector, first sparse vector, second sparse vector, and historical operation behavior vector; and determining the evaluation result of each virtual achievement object corresponding to the user based on the historical result, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment.

[0131] In an optional implementation, when the processor determines the evaluation result of each virtual achievement object corresponding to the user based on historical results, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment, it specifically performs the following steps: constructing a third sparse vector based on the completion quantity statistics at the current moment; wherein the third sparse vector is used to represent the completion information of the user for each virtual achievement object at the current moment; constructing a current operation behavior vector based on the user's operation behavior data at the current moment; determining the current result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix corresponding to the current moment, the user-side feature vector at the current moment, the third sparse vector, and the current operation behavior vector; and determining the evaluation result of each virtual achievement object corresponding to the user based on historical results and current results.

[0132] In an optional implementation, the historical operation behavior vector includes: historical touch behavior vector and historical favorite behavior vector; the current operation behavior vector includes: current touch behavior vector and current favorite behavior vector;

[0133] Both historical and current results are calculated using the following formula:

[0134] lin erCF=αU cf +βI cf ;

[0135] Ucf=x1U n*p (U T p*n *A1 n*m )+x2U n*p (U T p*n *A2 n*m )+x3U n*p (U T p*n *B n*m )+x4U n*p (U T p*n *C n*m );

[0136] I cf =γ1(A1 n*m I m*(a+b) )I T (a+b)*m +γ2(A2 n*m I m*(a+b) )I T (a+b)*m ;

[0137] lin erCF represents the matrix consisting of the historical results of each virtual achievement object corresponding to n users, U n*pA1 represents a matrix composed of the historical user-side feature vectors of n users; n*m A2 represents a sparse matrix composed of the first sparse vectors corresponding to each of the n users; n*m B represents the sparse matrix formed by the second sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the historical touch behavior vectors of n users on m virtual achievement objects; n*m I represents a matrix consisting of vectors representing the historical collection behaviors of n users on m virtual achievement objects; m*(a+b) Represents the feature matrix of historical achievements; or,

[0138] lin erCF represents the matrix consisting of the current results of each virtual achievement object corresponding to n users, U n*p A1 represents a matrix consisting of the user-side feature vectors of n users at the current time. n*m and A2 n*m B is a sparse matrix composed of the third sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the current touch behavior vectors of n users on m virtual achievement objects; n*m I represents a matrix consisting of vectors representing the current collection actions of n users on m virtual achievement objects; m*(a+b) This represents the achievement-side feature matrix corresponding to the current moment;

[0139] Where p represents the dimension of user attribute features; m represents the number of virtual achievement objects included in the game; a represents the dimension of achievement semantic encoding features for each virtual achievement object; b represents the dimension of achievement attribute features for each virtual achievement object; α, β, x1-x4, γ1 and γ2 are all coefficients.

[0140] In an optional implementation, when the processor determines the evaluation result of each virtual achievement object corresponding to the user based on historical results and current results, it specifically performs calibration calculations on the historical results and current results according to a preset calibration method to obtain the evaluation result of each virtual achievement object corresponding to the user.

[0141] In an optional implementation, when the processor recommends a target virtual achievement object to the user based on the evaluation results of each virtual achievement object, it specifically performs the following steps: sorting each virtual achievement object in descending order of each evaluation result to obtain sorted virtual achievement objects; and recommending a target virtual achievement object to the user based on the sorted virtual achievement objects.

[0142] The above method determines the user's current-moment user-side feature vector based on the user's operational behavior data and user attribute characteristics during gameplay. It then determines the achievement-side feature matrix for each virtual achievement object based on its achievement semantic encoding and achievement attribute characteristics. The achievement attribute characteristics include the inherent attribute information and completion statistics of the virtual achievement object. Finally, it determines the evaluation result for each virtual achievement object based on the achievement-side feature matrix and the user's current-moment user-side feature vector. Based on the evaluation results of each virtual achievement object, it recommends target virtual achievement objects to the user. This method recommends target virtual achievement objects to the user based on the user-side feature vector and achievement-side feature matrix. Since the user-side feature vector is determined based on the user's operational behavior data and user attribute characteristics, and the achievement-side feature matrix is ​​determined based on the achievement semantic encoding and achievement attribute characteristics of each virtual achievement object, it integrates richer feature information from both the user and achievement sides, enabling rapid and accurate personalized virtual achievement object recommendations.

[0143] The computer program products of the game processing method, apparatus and electronic device provided in the embodiments of the present invention include a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0144] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A game processing method, characterized in that, The game includes multiple virtual achievement objects, each of which corresponds to an achievement semantic encoding feature; wherein, the achievement semantic encoding feature characterizes the descriptive information and acquisition conditions of the achievement corresponding to the virtual achievement object; the method includes: Based on the user's operational behavior data at the current moment during the game and the user's user attribute characteristics, determine the user-side feature vector at the current moment; Based on the achievement semantic encoding features and achievement attribute features of each virtual achievement object, an achievement-side feature matrix corresponding to each virtual achievement object at the current moment is determined; wherein, the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object; The evaluation result for each virtual achievement object corresponding to the user is determined based on the achievement-side feature matrix of each virtual achievement object at the current moment and the user-side feature vector of the user at the current moment. Recommend target virtual achievement objects to the user based on the evaluation results of each virtual achievement object; The step of determining the evaluation result of each virtual achievement object corresponding to the user at the current moment, based on the achievement-side feature matrix corresponding to each virtual achievement object at the current moment and the user-side feature vector of the user at the current moment, includes: Obtain the historical achievement-side feature matrix and the historical user-side feature vector corresponding to the previous virtual achievement object recommendation time for each virtual achievement object at the current time; wherein, the historical achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and historical achievement attribute features of the virtual achievement object; the historical user-side feature vector is determined based on the historical operation behavior data and user attribute features of the user corresponding to the previous virtual achievement object recommendation time; the historical achievement attribute features include the inherent attribute information of the achievement corresponding to the virtual achievement object and the historical completion count statistics. Based on the historical completion statistics, a first sparse vector and a second sparse vector are constructed; wherein, the first sparse vector is used to represent the completion information of the user for each virtual achievement object within a first time period before the current time; the second sparse vector is used to represent the completion information of the user for each virtual achievement object within a second time period before the current time; the first time period is shorter than the second time period; Based on the user's operational behavior data during the second time period prior to the current moment, construct a historical operational behavior vector; Based on the historical achievement-side feature matrix, the historical user-side feature vector, the first sparse vector, the second sparse vector, and the historical operation behavior vector, the historical results of each virtual achievement object corresponding to the user are determined; Based on the historical results, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment, the evaluation result of each virtual achievement object corresponding to the user is determined.

2. The method according to claim 1, characterized in that, The achievement semantic encoding features are obtained in the following manner: The description information and acquisition conditions of the achievement corresponding to each virtual achievement object are input into a pre-built feature extractor to obtain the achievement semantic encoding features corresponding to each virtual achievement object output by the feature extractor.

3. The method according to claim 1, characterized in that, The steps for determining the evaluation result of each virtual achievement object corresponding to the user based on the historical results, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment include: Based on the completion statistics at the current moment, a third sparse vector is constructed; wherein, the third sparse vector is used to represent the completion information of the user for each virtual achievement object at the current moment; Construct a current operation behavior vector based on the user's operation behavior data at the current moment; Based on the achievement-side feature matrix corresponding to the current moment, the user-side feature vector at the current moment, the third sparse vector, and the current operation behavior vector, determine the current result of each virtual achievement object corresponding to the user; Based on the historical results and the current results, the evaluation result for each virtual achievement object corresponding to the user is determined.

4. The method according to claim 3, characterized in that, The historical operation behavior vector includes: historical touch behavior vector and historical favorite behavior vector; the current operation behavior vector includes: current touch behavior vector and current favorite behavior vector; Both the historical results and the current results are calculated using the following formula: ； Ucf=x1U n*p (U T p*n *A1 n*m )+ x2U n*p (U T p*n *A2 n*m )+ x3U n*p (U T p*n *B n*m )+ x4U n*p (U T p*n * C n*m ) ； I cf =γ1( A1 n*m I m*(a+b) )I T （a+b）*m + c2( A2 n*m I m*(a+b) )I T （a+b）*m ; This represents a matrix consisting of the historical results for each of the virtual achievement objects corresponding to n users. U n*p This represents a matrix composed of the historical user-side feature vectors of n users; A1 n*m This represents a sparse matrix composed of the first sparse vectors corresponding to each of the n users. A2 n*m B represents the sparse matrix formed by the second sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the historical touch behavior vectors of n users on m virtual achievement objects; n*m This represents a matrix consisting of vectors representing the historical collection behaviors of n users on m virtual achievement objects. I m*(a+b) Represents the feature matrix of historical achievements; or, This represents a matrix consisting of the current results of each of the virtual achievement objects corresponding to the n users. U n*p This represents a matrix composed of the user-side feature vectors of n users at the current moment; A1 n*m and A2 n*m B is a sparse matrix composed of the third sparse vectors corresponding to each of the n users; n*m C represents a matrix consisting of the current touch behavior vectors of n users on m virtual achievement objects; n*m This represents a matrix consisting of vectors representing the current collection actions of n users on m virtual achievement objects. I m*(a+b) This represents the achievement-side feature matrix corresponding to the current moment; Where p represents the dimension of user attribute features; m represents the number of virtual achievement objects included in the game; a represents the dimension of achievement semantic encoding features for each virtual achievement object; b represents the dimension of achievement attribute features for each virtual achievement object; α, β, x1—x4、 γ1 and γ2 are both coefficients.

5. The method according to claim 3, characterized in that, The step of determining the evaluation result for each virtual achievement object corresponding to the user based on the historical results and the current results includes: The historical results and the current results are calibrated and calculated according to a preset calibration method to obtain the evaluation result of each virtual achievement object corresponding to the user.

6. The method according to claim 1, characterized in that, The step of recommending a target virtual achievement to the user based on the evaluation results of each virtual achievement includes: Each virtual achievement object is sorted according to the order of each evaluation result from high to low, resulting in sorted virtual achievement objects; Based on the sorted virtual achievement objects, target virtual achievement objects are recommended to the user.

7. A game processing device, characterized in that, The game includes multiple virtual achievement objects, each of which corresponds to an achievement semantic encoding feature; wherein, the achievement semantic encoding feature characterizes the descriptive information and acquisition conditions of the achievement corresponding to the virtual achievement object; the device includes: The first determining module is used to determine the user-side feature vector at the current moment of the game based on the user's operation behavior data and the user attribute characteristics of the user. The second determining module is used to determine the achievement-side feature matrix corresponding to each virtual achievement object at the current time based on the achievement semantic encoding features and achievement attribute features of each virtual achievement object; wherein, the achievement attribute features include the inherent attribute information and completion count statistics of the achievement corresponding to the virtual achievement object; The third determining module is used to determine the evaluation result of each virtual achievement object corresponding to the user based on the achievement-side feature matrix corresponding to each virtual achievement object at the current time and the user-side feature vector of the user at the current time. The recommendation module is used to recommend target virtual achievement objects to the user based on the evaluation results of each virtual achievement object; The third determining module is also used for: Obtain the historical achievement-side feature matrix and the historical user-side feature vector corresponding to the previous virtual achievement object recommendation time for each virtual achievement object at the current time; wherein, the historical achievement-side feature matrix is ​​determined based on the achievement semantic encoding features and historical achievement attribute features of the virtual achievement object; the historical user-side feature vector is determined based on the historical operation behavior data and user attribute features of the user corresponding to the previous virtual achievement object recommendation time; the historical achievement attribute features include the inherent attribute information of the achievement corresponding to the virtual achievement object and the historical completion count statistics. Based on the historical completion statistics, a first sparse vector and a second sparse vector are constructed; wherein, the first sparse vector is used to represent the completion information of the user for each virtual achievement object within a first time period before the current time; the second sparse vector is used to represent the completion information of the user for each virtual achievement object within a second time period before the current time; the first time period is shorter than the second time period; Based on the user's operational behavior data during the second time period prior to the current moment, construct a historical operational behavior vector; Based on the historical achievement-side feature matrix, the historical user-side feature vector, the first sparse vector, the second sparse vector, and the historical operation behavior vector, the historical results of each virtual achievement object corresponding to the user are determined; Based on the historical results, the achievement-side feature matrix corresponding to the current moment, and the user-side feature vector at the current moment, the evaluation result of each virtual achievement object corresponding to the user is determined.

8. An electronic device, characterized in that, The device includes a processor and a memory, the memory storing machine-executable instructions that can be executed by the processor, the processor executing the machine-executable instructions to implement the game processing method according to any one of claims 1 to 6.

9. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores machine-executable instructions, which, when invoked and executed by a processor, cause the processor to implement the game processing method according to any one of claims 1 to 6.

Images