A game recommendation method and device, electronic equipment and medium

Abstract

The present disclosure provides a game recommendation method and device, electronic equipment and medium, and relates to the technical field of network games. The method comprises: processing long-term game information of a target user, short-term game information of the target user, target user information and a plurality of candidate game information based on an attention mechanism to obtain an interaction index corresponding to the candidate game; and if the interaction index of the candidate game meets a game recommendation condition, determining the candidate game that meets the game recommendation condition as a recommended game. The game recommendation method provided by the present disclosure is used for recommending games.

Description

Technical Field

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

[0002] To improve user retention on a platform, it's necessary to recommend games to users based on their interests. However, current game recommendation methods simply extract user game data without filtering the extracted user behavior data or considering the characteristics of the game data. This results in inaccurate user interest analysis based on user behavior data, leading to recommended games that are not of interest to users, making it difficult for game platforms to retain users. Summary of the Invention

[0003] According to one aspect of this disclosure, a game recommendation method is provided, the method comprising:

[0004] Based on the attention mechanism, the long-term game information of the target user, the short-term game information of the target user, the target user information, and the information of multiple candidate games are processed to obtain the interaction index of the corresponding candidate games;

[0005] If the interaction metrics of the candidate games meet the game recommendation criteria, the candidate games that meet the game recommendation criteria are determined as recommended games.

[0006] According to another aspect of this disclosure, a game recommendation device is provided, the device comprising:

[0007] The acquisition module is used to process the target user's long-term game information, the target user's short-term game information, the target user information, and multiple candidate game information based on the attention mechanism, and to obtain the interaction index corresponding to the candidate game.

[0008] The determining module, if the interaction metrics of the recommended game meet the game recommendation criteria, determines the candidate game that meets the game recommendation criteria as the recommended game.

[0009] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0010] Processor; and,

[0011] Memory for stored programs;

[0012] The program includes instructions that, when executed by the processor, cause the processor to perform a method provided according to an exemplary embodiment.

[0013] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions for causing the computer to perform a method provided according to an exemplary embodiment of this disclosure.

[0014] In one or more technical solutions provided in this disclosure, an attention mechanism is used to process the target user's long-term game information, short-term game information, target user information, and multiple candidate game information to obtain interaction metrics corresponding to the candidate games. This establishes a correlation between the user's long-term game information, short-term game information, target user information, and multiple candidate game information. This allows for a focus on features related to the target user information and multiple candidate game information when obtaining the interaction metrics for the corresponding candidate games, paying closer attention to the impact of the target user information and multiple candidate game information on the target user's game interest, thereby improving the accuracy of the obtained interaction metrics for the corresponding candidate games. Based on this, if the interaction metrics of a candidate game meet the game recommendation criteria, and a candidate game meeting the criteria is determined as a recommended game, the recommended game can more accurately represent the target user's multifaceted and highly personalized interest characteristics. When the recommended game is recommended to the target user, it matches the target user's game interests, attracting the target user to continue playing on the game platform and improving the platform's retention rate. Attached Figure Description

[0015] Further details, features, and advantages of this disclosure are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0016] Figure 1 A schematic structural diagram of a neural network model according to an exemplary embodiment of the present disclosure is shown;

[0017] Figure 2 A schematic flowchart of a game recommendation method according to an exemplary embodiment of the present disclosure is shown;

[0018] Figure 3 A schematic flowchart illustrating a processing method of a long-term and short-term game information processing unit according to an exemplary embodiment of the present disclosure is shown.

[0019] Figure 4 A schematic flowchart illustrating a method for deduplicating long-term game information of a target user according to an exemplary embodiment of the present disclosure is shown.

[0020] Figure 5 A schematic structural diagram of a game recommendation network included in a neural network model according to an exemplary embodiment of the present disclosure is shown;

[0021] Figure 6A schematic flowchart illustrating a method for obtaining interaction metrics of a corresponding candidate game according to an exemplary embodiment of the present disclosure is shown.

[0022] Figure 7 A schematic flowchart illustrating a method for obtaining short-term gaming interest characteristics of a target user according to an exemplary embodiment of the present disclosure is shown.

[0023] Figure 8 A schematic flowchart illustrating a method for obtaining interaction metrics of a corresponding candidate game according to an exemplary embodiment of the present disclosure is shown.

[0024] Figure 9 A schematic block diagram of the functional modules of a game recommendation device according to an exemplary embodiment of the present disclosure is shown;

[0025] Figure 10 A schematic block diagram of a chip according to an exemplary embodiment of the present disclosure is shown;

[0026] Figure 11 A structural block diagram of an exemplary electronic device that can be used to implement embodiments of the present disclosure is shown. Detailed Implementation

[0027] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0028] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0029] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below. It should be noted that the concepts of "first", "second", etc., used in this disclosure are only used to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0030] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0031] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0032] Before introducing the embodiments of this disclosure, the relevant terms involved in the embodiments of this disclosure are first defined as follows:

[0033] ResNet, also known as a residual network, is constructed from residual building blocks. Each residual block consists of multiple cascaded convolutional layers and a shortcut connection. The outputs of these layers are summed and then passed through a ReLU activation layer to obtain the output of the residual block. Multiple residual blocks can be chained together to create deeper networks.

[0034] Recurrent Neural Networks (RNNs) are a type of recursive neural network that takes sequence data as input, recurses along the direction of sequence evolution, and connects all nodes (recurrent units) in a chain-like manner. RNNs possess memory, parameter sharing, and Turing completeness, thus offering advantages in learning the nonlinear characteristics of sequences.

[0035] Attention mechanisms are deep learning techniques that help models more effectively focus on important information when processing sequential data. They assign a weight to the input data at each time step, representing the importance of that input data in the current task. This weight is obtained by calculating the similarity between the input data and the model's current state; the higher the similarity, the greater the weight. This mechanism can improve model performance and accuracy. Common attention mechanisms include one-way, two-way, and multi-head attention.

[0036] Improving user retention is a major challenge for current gaming platforms. Currently, personalized game recommendations are typically based on users' short-term and long-term interests. This approach usually comprises four modules: a user short-term and long-term behavior processing and extraction module, a user short-term and long-term interest extraction network, a short-term and long-term interest fusion and item recommendation module, and a long-term and short-term interest fusion and item recommendation module. The user short-term and long-term behavior processing and extraction module primarily focuses on user behavior, such as purchasing and listening behaviors. User behaviors are arranged chronologically and then divided into short-term and long-term behavior sequences using specific methods. The user short-term and long-term interest network often employs RNNs and multi-head self-attention mechanisms to extract long-term interest features, while fully connected neural networks or attention mechanisms are used to extract short-term interest features. The short-term and long-term interest fusion and item recommendation module often uses gate structures, weighted averages, or weighted summations to fuse users' short-term and long-term interests.

[0037] Because current short-term and long-term interest networks primarily target immediate user behaviors such as shopping and news browsing, this data contains relatively few repetitive user behaviors, and users' interests shift frequently and have short durations. Game data differs significantly from this. Not only does it contain a large amount of repetitive user behavior, but users' game interests do not shift frequently within a short period, and the behavior exhibited by users showing interest is not necessarily completely continuous. When users play games, they often enjoy playing one or a few games for a period of time, but they don't play those games every day, or they lose interest in some games but become interested in others. For example, a user might play a game at the beginning and end of a week, but not from Tuesday to Friday, which doesn't necessarily indicate a shift in their interest during that period. This makes the user interest analysis results obtained from the methods of personalized game recommendations based on users' short-term and long-term interests inaccurate.

[0038] To overcome the aforementioned problems, this exemplary embodiment provides a game recommendation method. This method comprehensively considers the persistence and time insensitivity of a target user's game interests, designs long-term and short-term game information for the target user, and processes the target user's long-term game information, short-term game information, target user information, and multiple candidate game information based on an attention mechanism. This allows the focus to be placed on features related to multiple candidate game information and target user information when determining the interaction indicators of multiple candidate games, thereby enabling the recommended games to represent the target user's multifaceted and highly personalized interest characteristics. It should be understood that the games involved in this exemplary embodiment, in terms of game classification, can be tower defense games, puzzle games, war games, management games, etc., and are not limited thereto. In terms of the game's state, the game can be a game that is about to be removed from the platform, or a game that may be removed in the future, etc., and are not limited thereto.

[0039] The game recommendation method of the exemplary embodiments of this disclosure is applied to a neural network model. The neural network model can extract the development and long-term game interest features of the target user from the target user's long-term game information, extract the target user's short-term game interest features from the target user's short-term game information, and use candidate game features and target user features for attention. This allows the neural network model to focus on features related to the target user and candidate games, thereby improving the prediction accuracy of the neural network model and making the recommended games more in line with the target user's game interests.

[0040] Figure 1 A schematic structural diagram of a neural network model according to an exemplary embodiment of the present disclosure is shown. Figure 1 As shown, the aforementioned neural network model can include five parts: a long-term and short-term game information processing unit, an input layer, a long-term and short-term game interest extraction network, a long-term and short-term game interest fusion unit, and a game recommendation network. First, the long-term and short-term game information processing unit processes the target user's historical game information, dividing it into long-term and short-term game information. Then, the target user's long-term game information, short-term game information, target user information, and multiple candidate game information are input into the long-term and short-term interest extraction network to obtain the target user's long-term and short-term game feature vectors. Finally, the extracted feature vectors are interacted with multiple candidate game information to obtain the probability that the target user will click on a corresponding candidate game, and game recommendations are made based on the click probabilities of all candidate games.

[0041] Figure 2 A schematic flowchart of a game recommendation method according to an exemplary embodiment of this disclosure is shown. Figure 2 As shown, the game recommendation method of this exemplary embodiment includes:

[0042] Step 201: Process the target user's long-term game information, short-term game information, target user information, and multiple candidate game information based on an attention mechanism to obtain the interaction metrics for the corresponding candidate games. By applying attention to the target user's long-term and short-term game information using the target user information and multiple candidate game information, the focus can be placed on features related to the target user information and multiple candidate game information when obtaining the interaction metrics for the corresponding candidate games. This allows for greater attention to the impact of the target user information and multiple candidate game information on the target user's game interest, thereby improving the accuracy of the obtained interaction metrics for the corresponding candidate games.

[0043] In practical applications, the target user's long-term game information, short-term game information, target user information, and multiple candidate game information can be used as samples in the aforementioned neural network model. It should be noted that this neural network model may include an attention mechanism network. This network processes the target user's long-term and short-term game information, assigning different weights based on the target user information and the candidate game information. Based on these weighted long-term and short-term game information, an interaction index for the corresponding candidate game can be obtained, allowing the interaction index to more accurately express the user's game interest. It should be understood that "corresponding" here refers to the interaction index of the candidate game determined when attention is applied to the target user's long-term and short-term game information based on one candidate game's information and the target user's information.

[0044] Step 202: If the interaction metrics of a candidate game meet the game recommendation criteria, then the candidate game that meets the criteria is designated as the recommended game. This ensures that the recommended games can more accurately represent the multifaceted and highly personalized interests of the target user. When the recommended game is recommended to the target user, it matches the target user's gaming interests, attracting the target user to continue playing on the game platform and improving the platform's retention rate. It should be understood that the recommended game here can be one game or multiple games; there is no limitation on this.

[0045] Considering the persistence and time insensitivity of target users' game interests, this exemplary embodiment proposes a method for fusing user interests over time. Based on this, the target user's long-term game information includes user game information from multiple periods, each period having different information. This ensures that the same game information is reflected only once in the target user's long-term game information, effectively avoiding a large amount of redundant information and thus reducing model overhead. It should be understood that "the same game information" here can refer to the user playing identical games in different periods.

[0046] For example, the target user's long-term game information can be divided into daily segments. The games played by the target user on each day are then sorted according to their playtime to obtain the target user's long-term game information. The target user's games played on day one, a1 = [g...]. 11 ,g 12 The games played on the second day, a2 = [g 21 ,g 22 When considering that the games played by the target user on the first day and the games played on the second day are the same game information, the games played by the target user on the first day or the second day can be deleted to remove a large amount of redundant information in the target user's long-term game information and reduce model consumption. At this time, the game information of the first period S1 included in the target user's long-term game information can be a1=[g 11 ,g 12 ] or a2 = [g 21 ,g 22 ], can be represented as S1=a1=[g 11 ,g 12 Or S1 = a2 = [g 21 ,g 22 ].

[0047] Figure 3 A schematic flowchart illustrating a processing method for a long-term and short-term game information processing unit according to an exemplary embodiment of this disclosure is shown. Figure 3 As shown, the processing method of the aforementioned long-term and short-term game information processing unit may include:

[0048] Step 301: Based on the target user's historical game information, obtain a list of games played by the target user across multiple historical periods. It should be understood that the target user's historical game information can be game information played before the current statistical time, or game information played before playing the candidate games. Based on the continuity of the target user's gaming interests, the target user's historical game information can be divided into long-term game information and short-term game information.

[0049] The target user's historical game information can include basic game information and the date of game activity. Basic game information can include game ID, game type, publisher, and other game details. When this historical game information is input into the neural network model, the target user's historical game information can include the game ID and the date of play. The neural network model then retrieves the game ID, game type, publisher, and other game details based on the game ID.

[0050] In practical applications, the target user's historical game information can be divided into days to obtain a list of games played by the target user across multiple historical periods, A = [a1, a2, a3, ..., a...]. n ], where a n =[g n1 ,g n2 ,...,g nj In the formula, a n G represents the list of games played by the target user on day n of their play history. nj This represents the j-th game played by the target user on day n in their game history. Considering that selecting a target user with a very long game history would increase the training cycle of the neural network model, the selected game history can be the game information played by the target user in the 30 days prior to playing the candidate games, where n = 30. Of course, the time frame for the target user's game history can also be set to other values, which are not limited here.

[0051] Step 302: Obtain the target user's long-term game information from game lists across multiple historical periods based on long-term game information sampling parameters. It should be understood that the long-term game information sampling parameters here include the sampling duration of the long-term game information, which is longer than the duration of the user's interest cycle.

[0052] In practical applications, the sampling duration of long-term game information can be equal to the sampling duration of the target user's historical game information to reduce the training cycle of the neural network model and improve its training efficiency. For example, when the sampling duration of the target user's historical game information is 30 days before the target user played the candidate game, the sampling duration of the long-term game information can also be 30 days before the target user played the candidate game. In this case, the target user's long-term game information A1 = A.

[0053] Considering that target users' interest in games shifts frequently and has a short duration, the method of the exemplary embodiment of this disclosure characterizes the persistence of users' interest in playing games by setting a user interest period T1. For game a1, if the target user does not play game a1 again within an interest period T1 after playing game a1 for a period of time, it is considered that the target user has lost interest in game a1.

[0054] In practical applications, the aforementioned interest period T1 can be set based on a large amount of users' historical game play data. For example, the initial value T0 of the interest period can be set based on the statistical results of a large amount of users' historical game play data. Using this large amount of users' historical game play data as a sample and the initial value T0 of the interest period as a label, the sample is input into the model for validation, and the initial value T0 of the interest period is adjusted according to the loss value to determine the interest period T1.

[0055] The statistical results of the aforementioned large-scale user historical game play data can be obtained by processing the data using methods such as large-scale observation, statistical grouping, comprehensive index methods, or time series analysis. It should be noted that the sampling duration for long-term game information can be an integer multiple of the duration of the interest period T1, in order to facilitate the identification of the target games played by target users in each period included in the long-term game information.

[0056] Step 303: Obtain the target user's short-term game information from game lists across multiple historical periods based on short-term game information sampling parameters. It should be understood that the short-term game information sampling parameters here include the sampling duration of the short-term game information, which is shorter than the sampling duration of the long-term game information. For example, when the sampling duration of the long-term game information is 30 days before the target user played the candidate games, the sampling duration of the short-term game information can be shorter than the sampling duration of the long-term game information. In this case, the sampling duration of the short-term game information could be 1 day or 3 days before the target user played the candidate games, etc., and is not limited to these.

[0057] For example, suppose the sampling duration of the above short-term game information is T2, then the short-term game information can be A2 = [a n-T2+1 ,…,a n-1 ,a n When the sampling duration for short-term game information is one day before the target user plays a candidate game, T2 = 1. In this case, the short-term game information can be A2 = [a n When the sampling period for short-term game information is 3 days before the target user plays the candidate game, T2 = 3. At this time, the short-term game information can be A2 = [a n-2 ,a n-1 ,a n ].

[0058] To reduce redundant data and improve the training efficiency of neural network models, the short-term game information of the target users can be deduplicated. This can be done using methods such as for loop deduplication, indexOf deduplication, filter deduplication, or array union deduplication.

[0059] Considering that short-term game information is mostly in one-dimensional arrays and the data processing volume is relatively small, the data union deduplication method can be used to deduplicate the short-term game information. Specifically, the short-term game information of the target user can be merged first, then the extended operation can be used to decompose the short-term game information of the target user, then the short-term game information of the target user can be deduplicated based on the Set array, and finally, the type of the returned Set array can be converted.

[0060] To remove redundant data from long-term game information sampling parameters and improve the training efficiency of neural network models, a for loop can be used to deduplicatize the long-term game information of the target user. Figure 4 A schematic flowchart illustrating a method for deduplicating long-term game information of a target user according to an exemplary embodiment of this disclosure is shown. Figure 4 As shown, obtaining a target user's long-term game information from game lists across multiple historical periods based on long-term game information sampling parameters can include:

[0061] Step 401: If the target game included in the target user's game list during the first target period is not in the first temporary list, add the target game to the first temporary list.

[0062] For example, a first temporary list of target users can be established. Since the target user's long-term game information is a two-dimensional array, this two-dimensional array can be processed through two loops. The first loop can determine the first target period of the target user's long-term game information. The second loop can determine whether the target games included in the game list of the first target period of the target user's long-term game information are in the first temporary list. Target games included in the game list of the first target period that are not in the first temporary list are added to the first temporary list, while target games included in the game list of the first target period that are in the first temporary list are ignored. This ensures that the game information included in the final determined first temporary game list is different from the game information included in the target user's game list of the first target period.

[0063] Step 402: If the first temporary list differs from the second temporary list, incrementally update the target user's long-term game information based on the first temporary list. It should be understood that, assuming the first target period is t, the first temporary list can be S. t The second target period can be t-1, and the second temporary list can be S. t-1This refers to the game list for a period prior to the first target period. By judging the similarity between the first temporary list and the second temporary list, the long-term game information is incrementally updated based on the first temporary list, which is different from the second temporary list. This removes the first temporary list that is completely identical to the second temporary list, avoiding a large amount of redundant data in the target user's long-term game information. While fully expressing the user's game interests, it can also reduce the consumption of the neural network model and improve the performance of the neural network model.

[0064] For example, when t=1, the first temporary list can be S1, and the second temporary list can be S0. If S1 and S0 are different, the games included in S1 are assigned to the target user's long-term game information to obtain the updated target user's long-term game information. Then, it is determined whether the games included in S1 are in S2 to S... 1+T1 If a game appears in the game list, keep it; otherwise, remove it to obtain an updated second temporary list. When t=2, the first temporary list can be S2, and the second temporary list can be the updated S1. If S2 is different from the updated S1, assign the games included in S2 to the target user's long-term game information to obtain the updated target user's long-term game information. It should be understood that the updated target user's long-term game information at this time includes both game list S1 and game list S2. Then, determine whether the games included in S2 are in S3 to S4. 2+T1 Games that appear in the initial game list are retained, while those that don't are removed, resulting in an updated second temporary list. This process continues until the target user's game list for multiple historical periods has been updated, thus obtaining the target user's long-term game information.

[0065] Step 403: If the target game included in the first temporary list is not in the target user's game list during the second target period, remove the target game from the first temporary list and update the second temporary list based on the first temporary list. The second target period is later than the first target period. It should be understood that the second target period can be determined based on the interest cycle mentioned earlier. If the target game does not appear in the game list within an interest cycle following the first target period corresponding to the first temporary list, it is considered that the target user has lost interest in the target game. Therefore, the target game can be removed from the first temporary list to improve the training accuracy of the neural network model.

[0066] The duration of the second target period is longer than the duration of the first target period, and the duration of the second target period is equal to the duration of the user interest cycle. At this point, we can set the first temporary list S1 = [g 11 ,g 12 ,g 13 The first target period corresponding to the first temporary list is t, and the target game is g.13 Then the second target period can be t+1 to t+T1. If the target game g does not appear in the corresponding game list within the time period of t+1 to t+T1... 13 Then it is assumed that the target user is interested in the target game g. 13 Loss of interest, therefore, the first temporary list S1 = [g 11 ,g 12 The updated second temporary list S2 = [g 11 ,g 12 ].

[0067] In practical applications, the long-term game information of the target user can be first set as Q = [S1, S2, S3, ..., S...]. m [m≤n]. Divide the target user's game list into multiple historical periods on a daily basis to obtain a long-term game sequence A1=[a1,a2,a3,…,a…]. n Among them, a n =[g n1 ,g n2 ,...,g nj ] and short-term game sequence A2 = [a n-T2+1 ,…,a n-1 ,a n At this point, a temporary interest sequence S = [] can be created for the target user, and the first day of the game period included in the long-term game sequence A1 can be looped to create a temporary storage sequence S. t =S, at this time, t=1. Then, iterate through each game in the game list of the first day in the long-term game sequence A1, and determine whether the corresponding game is in the temporary storage sequence S. t If the corresponding game is not in the temporary storage sequence S t In the process, the corresponding game is added to the temporary storage sequence S. t Then, determine the temporary storage sequence S. t Check if the game list included is completely consistent with the game list included in the temporary interest sequence S. If they are inconsistent, temporarily store sequence S. t Add the target user's long-term game information Q. Then determine whether the game list included in the period t+1 to t+T1 contains a temporary storage sequence S. t For each game included, if it is included, keep it; otherwise, remove it. Finally, process the S... t Update to the target user's temporary interest sequence S, that is, the target user's temporary interest sequence S = [S t ].

[0068] For example, let a1 = [g 11 ,g 12 ], then S1=[g 11 ,g 12If game g 11 If a game does not appear in the list of games included in the period 2 to 1+T1, then S = [g 12 ];a2=[g 22 ,g 23 ], then S2=[g 22 ,g 23 If game g 22 If a game does not appear in the list of games included in the period 3 to 2+T1, then S = [g 23 ];a3=[g 22 ,g 23 ], then S3 = [g 22 ,g 23 Since S3 = S2, remove S3; the resulting long-term game information for the target user is Q = [S1, S2] = [[g 11 ,g 12 ] , [g 22 ,g 23 ]).

[0069] Figure 5 A schematic structural diagram of a game recommendation network included in a neural network model according to an exemplary embodiment of the present disclosure is shown. Figure 5 As shown, the game recommendation network comprises five parts: an input layer, an embedding layer, a long-term and short-term game interest extraction network, a long-term and short-term game interest fusion unit, and a game recommendation network. It is used to recommend games to target users. The long-term and short-term game information processing unit within the neural network model described above processes the target user's historical game information, separating the target user's long-term and short-term game information. This information is then input into the game recommendation network to recommend games to the target user.

[0070] like Figure 5 As shown, the aforementioned long-term and short-term game interest extraction network is used to model the long-term game information and short-term game information of the target user respectively, to obtain the long-term interest features and short-term interest features of the target user, and to make game recommendations based on these features.

[0071] As one possible implementation method, Figure 6 A schematic flowchart illustrating a method for obtaining interaction metrics of corresponding candidate games according to an exemplary embodiment of this disclosure is shown. Figure 6 As shown, the above-mentioned processing of long-term game information, short-term game information, target user information, and multiple candidate game information based on attention mechanisms to obtain interaction metrics for corresponding candidate games may include:

[0072] Step 601: Determine the long-term sequence characteristics of the target game based on the target user's long-term game information.

[0073] For example, the long-term game information of the target user can be input into the embedding layer of the game recommendation network. The long-term game information of the target user is a list Q = [[g](qq. ... n1 ,g n2 ,...,g nj ],...,[g m1 ,g m2 ,...,g mj The process involves inputting the target user's long-term game information into the embedding layer to obtain the target user's long-term game vector for each game. This long-term game vector is then input into an average pooling unit, where it is averaged chronologically to obtain the target user's game vector for each period. This averaging of the target user's long-term game vectors further reduces their dimensionality, allowing them to be matched with the target user's short-term game information. Finally, the target user's game vector for each period is input into the temporal feature extraction network within the long-short-term game interest extraction network to obtain the target game's long-term sequence features. It should be understood that the temporal feature extraction network can be an RNN, GRU, LSTM, or Transformer network, etc.

[0074] Step 602: Process the target user's short-term game information and multiple candidate games based on an attention mechanism to obtain the target user's short-term game interest characteristics. It should be understood that the multiple candidate games input here include features such as the candidate game's ID information, game tags, and game release date.

[0075] For example, in a short-term game interest extraction network, attention scores can be assigned to each game in the target user's short-term game information based on an item attention network, along with the game features of the target game, to obtain the feature weights related to the target game. Then, the feature weights related to the target game, i.e. the features, are input into a pooling layer for fusion to obtain the short-term game interest features of the target user in the current period.

[0076] In the aforementioned long-term game interest extraction network, each target game's game vector represents the target user's game interest characteristics within a certain period. The differences and changes between the game vectors of two adjacent target games reflect the changes in the target user's interest in the game. The long-term sequence features of the entire target game can reflect the user's interest development. Using a temporal feature extraction network, we can uncover the changing trends of user interests and the impact of time on user interests. Based on the temporal feature extraction network, we model the long-term game interest development and game features of the target user. The output of the last time step represents the current development direction of the target user's game interest. However, although the temporal feature extraction network can extract the temporal features from the target user's long-term sequence features, it lacks modeling of the target user's information. The short-term game interest extraction network is similar to the long-term interest extraction network, also lacking personalized modeling of the target user's information, resulting in poor recommendation performance of the game recommendation model.

[0077] To overcome the above problems, the game recommendation model of the exemplary embodiment of this disclosure incorporates a user attention network after the temporal feature extraction network. It further personalizes the long-term sequence features of the target game and the short-term game interest features of the target user by utilizing the target user features to obtain the long-term game interest feature vector and the short-term game interest feature vector of the target user, thereby improving the recommendation effect of the game recommendation model.

[0078] Step 603: Based on the attention mechanism, process the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and the information of multiple candidate games to obtain the interaction indicators of the corresponding candidate games. It should be understood that the target user information input here includes target user ID information, target user age information, target user gender information, etc.

[0079] For example, the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, and the target user information can be input into the user attention network included in the short-term and long-term game interest extraction network. The target user characteristics can be used to further personalize the long-term sequence characteristics of the target game and the short-term game interest characteristics of the target user to obtain the interaction index of the corresponding candidate games.

[0080] As one possible implementation method, Figure 7 A schematic flowchart illustrating a method for obtaining short-term gaming interest characteristics of a target user according to an exemplary embodiment of this disclosure is shown. Figure 7 As shown, the above-mentioned attention-based processing of short-term game information and multiple candidate games for the target user yields short-term game interest characteristics for the target user, including:

[0081] Step 701: Based on the target user's short-term game information and multiple candidate game information, determine the attention evaluation information of the corresponding candidate game on the target user's short-term game information. It should be understood that "corresponding candidate game" here refers to determining the attention evaluation information of that candidate game on the target user's short-term game information based on one of the multiple candidate game information. Similarly, attention evaluation information for multiple target users' short-term game information can be obtained, enabling the game recommendation model to personalize the filtering of short-term game information based on each candidate game information when processing it.

[0082] Step 702: Fuse the attention evaluation information of multiple candidate games on the target user's short-term game information to obtain the target user's short-term game interest characteristics.

[0083] In practical applications, a pooling layer is added after the item attention network to fuse the attention evaluation information of multiple candidate games on the target user's short-term game information, so as to obtain the target user's short-term game interest characteristics.

[0084] In some alternative methods, Figure 8 A schematic flowchart illustrating a method for obtaining interaction metrics of corresponding candidate games according to an exemplary embodiment of this disclosure is shown. Figure 8 As shown, the above-mentioned processing of the target user's short-term game interest characteristics, the target game's long-term sequence characteristics, the target user information, and multiple candidate game information based on the attention mechanism to obtain the corresponding candidate game's interaction indicators can include:

[0085] Step 801: Obtain self-attention fusion information based on the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and information of multiple candidate games.

[0086] In practical applications, after obtaining the short-term game interest features of the target user and the long-term sequence features of the target game, the short-term game interest features of the target user, the long-term sequence features of the target game, the target user information, and information from multiple candidate games can be input into the Concat function included in the short- and long-term game interest fusion unit for fusion to obtain Concat fusion information. Then, the Concat fusion information is input into the multi-head self-attention network included in the short- and long-term game interest fusion unit for deep fusion to obtain sub-attention fusion information.

[0087] To better represent user characteristics and improve the accuracy of predictions in the game recommendation model, the short- and long-term game interest fusion unit can include a combination of a multi-head self-attention network and a residual network. Therefore, step 802: residual information can be obtained based on the self-attention fusion information, the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and information from multiple candidate games.

[0088] In practical applications, after obtaining the self-attention fusion information, the self-attention fusion information, the short-term game interest features of the target user, the long-term sequence features of the target game, the target user information, and the information of multiple candidate games can be used as the input vector of the residual network. This input vector is then used to perform deep fusion within the residual network of the long-term and short-term game interest fusion unit to obtain residual information.

[0089] Since high-order feature fusion requires multiple layers of multi-head self-attention networks, each layer of the multi-head self-attention network can be combined with a residual network. The output of each multi-head self-attention network and its input can be used as the input to the residual network, resulting in the residual network output being y. l =MultiHeadSelfAttention(y (l-1) )+y (l-1) In the formula, y l MultiHeadSelfAttention(y) represents the output vector of the l-th residual network. (l-1) ) represents the output vector of the previous layer's multi-head self-attention network, y (l-1) The input vector represents the multi-head self-attention network of the previous layer, and the output of the last residual network is the residual information. The game recommendation model of this exemplary embodiment fuses high-order and low-order features by combining a multi-head self-attention network and a residual network, resulting in more complete fusion of high and low-order features. Simultaneously, it retains low-order features while obtaining high-order features, allowing the game recommendation model to better represent user characteristics and thus improve prediction performance. Furthermore, this game recommendation model can perform feature interaction based on the original features, preventing feature loss that may occur due to transformation operations before feature interaction. Moreover, feature interaction is implemented using a multi-head self-attention mechanism. Compared to gate structures and other attention methods, multiple feature vectors are combined through the concat function, without requiring restrictions on the consistency of feature dimensions.

[0090] Step 803: Determine the interaction metrics for the corresponding candidate games based on the residual information. After obtaining the residual information, it can be input into the DNN neural network included in the game recommendation network for further feature fusion and click-through rate prediction. The DNN neural network outputs the click probability of the target user for the current target game, and recommends games to the target user based on the click probability.

[0091] As one possible implementation, the above game recommendation criteria may include: the interaction index of the candidate game is greater than or equal to the preset interaction index.

[0092] The interaction metrics for the candidate games can be either range values ​​or point values. For example, assuming the preset interaction metric is 0.97, the first candidate game has an interaction metric of 0.99, and the second candidate game has an interaction metric of 0.95. It is clear that the first candidate game's interaction metric is greater than the preset interaction metric, while the second candidate game's interaction metric is less than the preset interaction metric. Therefore, the first candidate game is selected as the recommended game.

[0093] As one possible implementation, the game recommendation method of the exemplary embodiment of this disclosure further includes: sorting multiple candidate games according to the interaction index of the multiple candidate games in descending order, wherein the game recommendation condition includes: the order of the interaction index of the recommended game in the sorting is less than or equal to M, where M represents an integer less than the total number of candidate games.

[0094] For example, multiple candidate games can be sorted in descending order of their interaction metrics, with the recommended game's interaction metric being less than or equal to M. This effectively limits the recommended games to the top M ranked candidate games. For instance, if the interaction metrics of the first candidate game in the top position are sorted in descending order, and the interaction metric of the first candidate game is 1, and the interaction metric of the Mth candidate game in the Mth position is 0.95, then the candidate games with an interaction metric greater than or equal to 0.95 among the top M games are recommended to the target user.

[0095] The foregoing primarily describes the solutions provided by the embodiments of this disclosure from the perspective of the server. It is understood that, in order to implement the above functions, the server includes the corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this disclosure can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0096] This disclosure embodiment can divide the server into functional units according to the above method example. For example, it can divide each function into a separate functional module, or it can integrate two or more functions into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in this disclosure embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0097] By dividing each functional module according to its corresponding function, an exemplary embodiment of this disclosure provides a game recommendation device, which can be a server or a chip applied to a server. Figure 9 A schematic block diagram of the functional modules of a game recommendation device according to an exemplary embodiment of the present disclosure is shown. Figure 9 As shown, the game recommendation device 900 includes:

[0098] The module 901 is used to process the long-term game information of the target user, the short-term game information of the target user, the target user information and multiple candidate game information based on the attention mechanism, and to obtain the interaction index corresponding to the candidate game.

[0099] Module 902 determines candidate games that meet the game recommendation criteria if the interaction metrics of the recommended games meet the game recommendation criteria.

[0100] As one possible implementation, the long-term game information of the target user mentioned above includes user game information from multiple periods, with each period having different user game information.

[0101] As one possible implementation, the game recommendation device 900 further includes an acquisition module 903, which is used to acquire a list of games played by the target user in multiple historical periods based on the target user's historical game information; acquire long-term game information of the target user from the list of games played by the target user in multiple historical periods based on long-term game information sampling parameters; and acquire short-term game information of the target user from the list of games played by the target user in multiple historical periods based on short-term game information sampling parameters.

[0102] In some alternative methods, the aforementioned long-term game information sampling parameters include the sampling duration of the long-term game information, which is longer than the duration of the user's interest cycle.

[0103] The sampling parameters for short-term game information include the sampling duration of short-term game information, which is shorter than the sampling duration of long-term game information.

[0104] In some alternative embodiments, the game recommendation device 900 further includes an adding module 904, which adds the target game to the first temporary list if the target game included in the target user's game list during the first target period is not in the first temporary list.

[0105] The aforementioned game recommendation device 900 also includes an update module 905. If the first temporary list is different from the second temporary list, the update module 905 is used to incrementally update the target user's long-term game information based on the first temporary list.

[0106] The aforementioned game recommendation device 900 also includes a removal module 906, which removes the target game from the first temporary list if the target game in the first temporary list is not in the target user's game list during the second target period.

[0107] The update module 905 is also used to update a second temporary list based on a first temporary list, wherein the second target period is later than the first target period.

[0108] In some alternative approaches, the duration of the second target period is longer than the duration of the first target period, and the duration of the second target period is equal to the duration of the user interest cycle.

[0109] As one possible implementation, the aforementioned determining module 902 is also used to determine the long-term sequence characteristics of the target game based on the target user's long-term game information.

[0110] The acquisition module 901 is also used to process the short-term game information of the target user and multiple candidate games based on the attention mechanism, to obtain the short-term game interest characteristics of the target user, and to process the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information and the information of multiple candidate games based on the attention mechanism to obtain the interaction index of the corresponding candidate game.

[0111] In some alternative methods, the determining module 902 is further configured to determine the attention evaluation information of the corresponding candidate game on the target user's short-term game information based on the target user's short-term game information and multiple candidate game information.

[0112] The module 901 is also used to fuse the attention evaluation information of multiple candidate games on the target user's short-term game information to obtain the target user's short-term game interest characteristics.

[0113] In some alternative methods, the aforementioned obtaining module 901 is further configured to obtain self-attention fusion information based on the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and multiple candidate game information; and to obtain residual information based on the self-attention fusion information, the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and multiple candidate game information.

[0114] The aforementioned determining module 902 is also used to determine the interaction index of the corresponding candidate game based on the residual information.

[0115] As one possible implementation, the above game recommendation criteria include: the interaction index of the candidate game is greater than or equal to the preset interaction index.

[0116] As one possible implementation, the game recommendation device 900 also includes a sorting module 907, which sorts the multiple candidate games according to their interaction metrics from largest to smallest. The game recommendation criteria include: the order of the interaction metrics of the recommended game in the sort is less than or equal to M, where M represents an integer less than the total number of candidate games.

[0117] Figure 10 A schematic block diagram of a chip according to an exemplary embodiment of the present disclosure is shown. Figure 10 As shown, the chip 1000 includes one or more (including two) processors 1001 and a communication interface 1002. The communication interface 1002 can support the server in performing the data transmission and reception steps in the above-described image processing method, and the processor 1001 can support the server in performing the data processing steps in the above-described image processing method.

[0118] Optional, such as Figure 10 As shown, the chip 1000 also includes a memory 1003, which may include read-only memory and random access memory, and provides operation instructions and data to the processor. A portion of the memory may also include non-volatile random access memory (NVRAM).

[0119] In some implementations, such as Figure 10As shown, processor 1001 executes corresponding operations by calling operation instructions stored in memory (which may be stored in the operating system). Processor 1001 controls the processing operations of any terminal device; processor can also be called a central processing unit (CPU). Memory 1003 may include read-only memory and random access memory, and provides instructions and data to processor 1001. A portion of memory 1003 may also include NVRAM. For example, in applications, memory, communication interfaces, and other components are coupled together via a bus system, which may include, in addition to a data bus, a power bus, a control bus, and a status signal bus, etc. However, for clarity, in... Figure 10 The general labeled all buses as Bus System 1004.

[0120] The methods disclosed in the embodiments of this disclosure can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above methods can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an 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 embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0121] Exemplary embodiments of this disclosure also provide an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the electronic device to perform a method according to an embodiment of this disclosure.

[0122] Exemplary embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform a method according to embodiments of this disclosure.

[0123] Exemplary embodiments of this disclosure also provide a computer program product, including a computer program, wherein, when executed by a processor of a computer, the computer program is used to cause the computer to perform a method according to an embodiment of this disclosure.

[0124] refer to Figure 11 The present invention describes a structural block diagram of an electronic device 1100 that can serve as a server or client of the present disclosure, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0125] like Figure 11 As shown, the electronic device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. The RAM 1103 may also store various programs and data required for the operation of the electronic device 1100. The computing unit 1101, ROM 1102, and RAM 1103 are interconnected via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0126] Multiple components in electronic device 1100 are connected to I / O interface 1105, including: input unit 1106, output unit 1107, storage unit 1108, and communication unit 1109. Input unit 1106 can be any type of device capable of inputting information to electronic device 1100. Input unit 1106 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 1107 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 1108 may include, but is not limited to, disk and optical disk. Communication unit 1109 allows electronic device 1100 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0127] The computing unit 1101 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1101 performs the various methods and processes described above. For example, in some embodiments, the methods of the exemplary embodiments of this disclosure can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1108. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 1100 via ROM 1102 and / or communication unit 1109. In some embodiments, the computing unit 1101 can be configured to perform the methods of the exemplary embodiments of this disclosure by any other suitable means (e.g., by means of firmware).

[0128] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0129] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0130] As used in this disclosure, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0131] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0132] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0133] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

[0134] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this disclosure are performed, in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a terminal, a user equipment, or other programmable device. The computer program or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center integrating one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; it can also be an optical medium, such as a digital video disc (DVD); or it can be a semiconductor medium, such as a solid-state drive (SSD).

[0135] Although this disclosure has been described in conjunction with specific features and embodiments, it will be apparent that various modifications and combinations can be made therein without departing from the spirit and scope of this disclosure. Accordingly, this specification and drawings are merely exemplary illustrations of the disclosure as defined by the appended claims and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this disclosure. It is obvious that those skilled in the art can make various alterations and modifications to this disclosure without departing from its spirit and scope. Thus, this disclosure is also intended to include any such modifications and modifications that fall within the scope of the claims of this disclosure and their equivalents.

Claims

1. A game recommendation method, characterized in that, The method includes: Determine the long-term sequence characteristics of the target game based on the long-term game information of the target users; Based on the attention mechanism, the short-term game information of the target user and multiple candidate games are processed to obtain the short-term game interest characteristics of the target user; Based on the attention mechanism, the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and the information of multiple candidate games are processed to obtain the interaction index of the corresponding candidate game. If the interaction metrics of the candidate games meet the game recommendation criteria, the candidate games that meet the game recommendation criteria are determined as recommended games.

2. The game recommendation method according to claim 1, characterized in that, The method further includes: Based on the target user's historical game information, obtain the target user's game list in multiple historical periods; Based on long-term game information sampling parameters, the target user's long-term game information is obtained from game lists of multiple historical periods in the target user's game list. Based on short-term game information sampling parameters, the short-term game information of the target user is obtained from the game lists of the target user in multiple historical periods.

3. The game recommendation method according to claim 2, characterized in that, The long-term game information sampling parameters include the sampling duration of the long-term game information, and the sampling duration of the long-term game information is longer than the duration of the user's interest cycle; The short-term game information sampling parameters include the sampling duration of the short-term game information, which is less than the sampling duration of the long-term game information.

4. The game recommendation method according to claim 2, characterized in that, The step of obtaining the target user's long-term game information from game lists across multiple historical periods based on long-term game information sampling parameters includes: If the target game included in the target user's game list during the first target period is not in the first temporary list, the target game is added to the first temporary list; If the first temporary list is different from the second temporary list, the long-term game information of the target user is updated incrementally based on the first temporary list. If the target game included in the first temporary list is not in the target user's game list during the second target period, the target game is removed from the first temporary list, and the second temporary list is updated based on the first temporary list, where the second target period is later than the first target period.

5. The game recommendation method according to claim 1, characterized in that, The process of processing the target user's short-term game information and multiple candidate games based on an attention mechanism to obtain the target user's short-term game interest characteristics includes: Based on the target user's short-term game information and multiple candidate game information, determine the attention evaluation information of each candidate game on the target user's short-term game information; The attention evaluation information of multiple candidate games on the target user's short-term game information is fused to obtain the target user's short-term game interest characteristics.

6. The game recommendation method according to claim 1, characterized in that, The process of processing the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and information on multiple candidate games based on an attention mechanism to obtain interaction metrics corresponding to the candidate games includes: Based on the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and the information of multiple candidate games, self-attention fusion information is obtained; Based on the self-attention fusion information, the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and the information of multiple candidate games, residual information is obtained; The interaction metrics corresponding to the candidate games are determined based on the residual information.

7. A game recommendation device, characterized in that, The device includes: The determination module identifies long-term sequence characteristics of the target game based on the target user's long-term game information; The acquisition module processes the short-term game information of the target user and multiple candidate games based on an attention mechanism to obtain the short-term game interest characteristics of the target user. The acquisition module is further configured to process the short-term game interest characteristics of the target user, the long-term sequence characteristics of the target game, the target user information, and the information of multiple candidate games based on an attention mechanism to obtain the interaction index corresponding to the candidate game; The determining module is further configured to determine, if the interaction metrics of the candidate games meet the game recommendation criteria, that the candidate games that meet the game recommendation criteria are recommended games.

8. An electronic device, characterized in that, include: processor; as well as, Memory for stored programs; The program includes instructions that, when executed by the processor, cause the processor to perform the method according to any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method according to any one of claims 1-6.

Images