Behavior recognition methods, devices, computer equipment, and computer-readable storage media

Abstract

This application discloses a behavior recognition method, apparatus, computer device, and computer-readable storage medium. The method involves acquiring an object set, including object samples using an auxiliary program and objects to be predicted, along with object feature information and policy behavior information for each object. Implicit behavior information mining is performed on the policy behavior information of each object to obtain frequent behavior information for each object. Based on the object feature information and frequent behavior information, the objects in the object set are clustered to obtain a target clustering result. Target object clusters are determined based on the distribution ratio of object samples using the auxiliary program in each object cluster of the target clustering result. The objects to be predicted contained in the target object clusters are identified as objects using the auxiliary program. This scheme can cluster objects with similar behavioral characteristics and then identify objects to be predicted using the auxiliary program through object samples. It is applicable to different application scenarios and offers high flexibility.

Description

Technical Field

[0001] This application relates to the field of communication technology, specifically to a behavior recognition method, apparatus, computer device, and computer-readable storage medium. Background Technology

[0002] In gaming applications or online competitions, such as chess tournaments, some users use cheat programs during gameplay or competitions to gain an advantage. This use of cheat programs severely impacts the fairness of the game, making the identification of cheat program use crucial for such scenarios. While manually setting rules to determine cheat program use is possible, this requires experts in the relevant field to pre-list and summarize potential cheat program usage behaviors within the scenario. Such rules are only applicable to a specific application scenario, resulting in low flexibility. Furthermore, the variety of cheat programs used by users necessitates continuous rule updates based on the specific programs employed, leading to high application costs. Summary of the Invention

[0003] This application provides a behavior recognition method that clusters objects with similar behavioral characteristics and then identifies the objects to be predicted using an auxiliary program through object samples. It can be applied to different application scenarios and has high flexibility.

[0004] This application provides a behavior recognition method, including:

[0005] Obtain an object set, which includes object samples and objects to be predicted using the auxiliary program, as well as object feature information and policy behavior information corresponding to each object;

[0006] For each object's strategy behavior information, implicit behavior information mining is performed to obtain the frequent behavior information corresponding to the object;

[0007] Based on the object feature information and the frequent behavior information, the objects in the object set are clustered to obtain the target clustering result;

[0008] The target object cluster is determined based on the distribution proportion of the object samples using the auxiliary program in each object cluster of the target clustering result;

[0009] The objects to be predicted contained in the target object cluster are identified as objects for which an auxiliary program is used.

[0010] Accordingly, embodiments of this application also provide a behavior recognition device, comprising:

[0011] The acquisition unit is used to acquire a set of objects, which includes object samples and objects to be predicted using the auxiliary program, as well as object feature information and policy behavior information corresponding to each object;

[0012] The mining unit is used to perform implicit behavior information mining processing on the strategy behavior information of each object to obtain the frequent behavior information corresponding to the object.

[0013] A clustering unit is used to cluster objects in the object set based on the object feature information and the frequent behavior information to obtain the target clustering result;

[0014] The first determining unit is used to determine the target object cluster based on the distribution ratio of the object samples using the auxiliary program in each object cluster of the target clustering result;

[0015] The second determining unit is used to determine the objects to be predicted contained in the target object cluster as objects for use with the auxiliary program.

[0016] Accordingly, this application also provides a computer device including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute any of the behavior recognition methods provided in this application.

[0017] Accordingly, embodiments of this application also provide a computer-readable storage medium for storing a computer program, which is loaded by a processor to execute any of the behavior recognition methods provided in embodiments of this application.

[0018] As described above, this embodiment of the application obtains an object set, which includes object samples using the auxiliary program and objects to be predicted, as well as object feature information and strategy behavior information corresponding to each object; it performs implicit behavior information mining processing on the strategy behavior information of each object to obtain frequent behavior information corresponding to each object; based on the object feature information and frequent behavior information, the objects in the object set are clustered to obtain the target clustering result; the target object cluster is determined according to the distribution ratio of the object samples using the auxiliary program in each object cluster of the target clustering result; and the objects to be predicted contained in the target object cluster are identified as objects using the auxiliary program. This scheme can obtain the frequent behavior information of objects by mining the strategy behavior information of objects. The frequent behavior information can reflect the behavioral characteristics of objects. Clustering based on the frequent behavior information can cluster objects with similar behavioral characteristics, and then identify the objects to be predicted using the auxiliary program through object samples. It can be applied to different application scenarios and has high flexibility. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a scene diagram of the behavior recognition method provided in the embodiments of this application;

[0021] Figure 2 This is a flowchart of the behavior recognition method provided in the embodiments of this application;

[0022] Figure 3 This is a schematic diagram of the behavior recognition method provided in the embodiments of this application;

[0023] Figure 4 This is another flowchart of the behavior recognition method provided in the embodiments of this application;

[0024] Figure 5 This is a schematic diagram of the behavior recognition device provided in an embodiment of this application;

[0025] Figure 6 This is a schematic diagram of the structure of the computer device provided in the embodiments of this application. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] This application provides a behavior recognition method, apparatus, computer device, and computer-readable storage medium. The behavior recognition apparatus can be integrated into a computer device, which may be a server or a terminal, etc.

[0028] The terminal may include mobile phones, wearable smart devices, tablets, laptops, personal computers (PCs), and in-vehicle computers, etc.

[0029] The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0030] For example, such as Figure 1 As shown, the computer device can acquire a set of objects, including object samples using the auxiliary program, objects to be predicted, object feature information corresponding to each object, and strategy behavior information in historical competitions. Implicit behavior information mining is performed on each strategy behavior information to obtain frequent behavior information corresponding to the strategy behavior information. Based on the object feature information and frequent behavior information, the objects in the object set are clustered to obtain multiple clustering results. A target clustering result is determined based on the multiple clustering results. The target object cluster is determined based on the distribution ratio of object samples using the auxiliary program in each object cluster of the target clustering result. The objects to be predicted in the target object cluster are then identified as objects using the auxiliary program. This scheme, by mining the strategy behavior information of objects, can obtain frequent behavior information of objects. Frequent behavior information can reflect the behavioral characteristics of objects. Clustering based on frequent behavior information can group objects with similar behavioral characteristics together. Furthermore, object samples can be used to identify whether the objects to be predicted using the auxiliary program have used it. This scheme is applicable to different application scenarios and has high flexibility.

[0031] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the preferred order of the embodiments.

[0032] This embodiment will be described from the perspective of a behavior recognition device, which can be integrated into a computer device, such as a server or a terminal. Figure 2 As shown, the process of this behavior recognition method is as follows;

[0033] 101. Obtain the object set, which includes the object samples used by the auxiliary program and the objects to be predicted, as well as the object feature information and policy behavior information corresponding to each object.

[0034] The object set can be a collection containing multiple objects and their corresponding information. For example, it can include object samples using auxiliary programs and objects to be predicted. An object can be a virtual object in a game scene or a competition scene. Each object can correspond to an object identifier, such as a game account, as well as some attribute information, such as account ID, avatar, and age.

[0035] Among them, auxiliary programs generally refer to cheating programs or software that modify game or online competition data to gain an advantage for users. That is, they use computer technology to perform non-original operations on one or more software programs, tamper with the original normal settings and rules of game or online competition data, and greatly enhance the user's skills and extraordinary abilities in the game or competition process, thereby easily obtaining the benefits of victory and rewards. They are cheating programs created by changing parts of the software program.

[0036] Among them, the object sample using the auxiliary program can be an object that is known to use the auxiliary program, and the object to be predicted can be an object that needs to be identified and judged to use the auxiliary program.

[0037] Among them, object feature information can be the object's attribute information, which is used to characterize the object's features. For example, the object's age, the object's level in the application scenario (for example, if the object is a player in a game, it can be the game level; if the object corresponds to a contestant in an online competition, it can be the object's rank, etc.), the object's historical score in the application scenario, gender, location, personality, zodiac sign, habitual tactics (for example, in a game scenario, the object's commonly used game moves, etc.), and win rate, etc.

[0038] Among them, strategic behavior information can be behavioral information identified by strategy identifiers. Behavioral information can be a sequence of behaviors that include all or part of the object's behaviors in a game. For example, in chess, opening strategies include the Italian Opening, the Two Knights Defense, the Hungarian Defense, the Spanish Opening, and the Sicilian Defense, while tactical strategies include checkmate, double check, double checkmate, pinning, and luring away, etc. Each opening strategy and each tactical strategy is marked with an identifier, and behaviors that conform to the strategy are marked with the identifier corresponding to that strategy. Specifically, behavioral information can be agjief, and mapping each behavior to the corresponding strategy can yield A1A2A3A1A2A2, where A1, A2, and A3 are identifiers corresponding to different strategies.

[0039] For example, the object collection can be retrieved from a database or memory. Alternatively, if the object collection is stored in a blockchain, it can be retrieved from the blockchain.

[0040] In one embodiment, upon receiving an identification request, for example, when a game user reports a virtual object in a game using an auxiliary program through the game user interface, the object feature information corresponding to the reported object, as well as the strategic behavior information in historical game matches, and the object feature information and strategic behavior information corresponding to the object sample can be obtained to obtain an object set.

[0041] Optionally, one can periodically obtain object feature information corresponding to the reported object, as well as strategic behavior information in historical game competitions, and obtain object feature information and strategic behavior information corresponding to object samples to obtain an object set.

[0042] In one embodiment, the strategy behavior information can be obtained by determining the strategy corresponding to the behavior information through similarity with reference behavior information of known strategies, and then using this strategy to identify the behavior information. Specifically, before the step "obtain the object set", the strategy behavior information can be:

[0043] Obtain an initial object set and a policy behavior information set. The initial object set includes object samples and objects to be predicted using the auxiliary program, as well as initial object information and behavior information corresponding to each object. The policy behavior information set includes reference behavior information of at least one known policy.

[0044] Preprocess the initial object information to obtain object feature information;

[0045] Based on the similarity between the behavioral information and the reference behavioral information of each known strategy in the set of strategic behavioral information, the strategy of the behavioral information is determined to obtain the strategic behavioral information corresponding to the behavioral information.

[0046] Based on the object samples and the objects to be predicted, as well as the object feature information and policy behavior information corresponding to each object, an object set is obtained.

[0047] The initial object information may be information that has been preprocessed, such as filtering object information, and may contain outliers and missing values.

[0048] Among them, behavioral information can be a sequence of behaviors that includes all or part of the object's behaviors in a game.

[0049] Because some users may not have completed their personal information in the initial object set, or may have filled in personal information that is seriously inconsistent with their own information, or some data may have been lost during the process of obtaining the object set, the initial object information in the initial object set may be incomplete and incorrect. If the initial object information is not processed, it will affect the subsequent clustering of similar objects into an object cluster.

[0050] Preprocessing the initial object information can specifically include feature data corresponding to multiple features of each object, such as age, gender, location, and zodiac sign. If the number of missing feature data exceeds a threshold, the feature is discarded. The threshold can be flexibly set according to the specific application scenario.

[0051] Optionally, single-value features in the initial object information can be deleted. These single-value features are those where all objects in the initial object information have the same feature data value. Since all objects have the same feature data, it is impossible to distinguish one object from another. Therefore, these single-value features can be deleted to reduce the amount of data processing and enable the object feature information to more accurately represent the object.

[0052] Optionally, outliers in the initial object information can also be processed. For example, the feature data contained in each feature can be sorted, and the feature data whose values ​​meet the conditions can be filtered. The condition can be that the feature data is ranked in the top 0.01%, that is, the feature data whose values ​​are ranked in the top 0.01% can be discarded.

[0053] Optionally, after discarding features and filtering some feature data, there may still be missing feature data in some features. The missing feature data can be filled in to make the feature data in each feature complete. For example, for features with continuous feature data, the average value of the feature data can be used to fill in the missing feature data, and for features with discrete feature data, a constant can be used to fill in the missing feature data.

[0054] Optionally, features in the initial object information can be transformed or combined to generate new features that are meaningful and more informational. For example, features can be combined and derived through feature transformation, feature squaring, and feature addition and subtraction.

[0055] Optionally, features can also be encoded, such as binning and discretizing continuous features, or one-hot encoding discrete features.

[0056] For behavioral information, the strategy corresponding to the behavioral information can be determined based on its similarity to the reference behavioral information of each known strategy in the set of strategy behavioral information. For example, the strategy of the reference behavioral information with the highest similarity is determined as the strategy corresponding to the behavioral information. The behavioral information is represented by the identifier of the corresponding strategy. For example, specifically, the behavioral information can be agjief, where the behavior ahj conforms to the common feature information of strategy A1, so the behavior ahj can be represented by A1; the behavior ef conforms to the common feature information of strategy A3, so the behavior ahj can be represented by A3; the behavior i does not conform to the common feature information of any strategy in the set of strategy common information, so the behavior i can be represented separately, for example, denoted as 1. The strategy behavioral information corresponding to the behavioral information agjief is A11 A3.

[0057] The similarity between behavioral information and reference behavioral information can be calculated using their corresponding feature information, which can be extracted using a neural network model. Specifically, the step "based on the similarity between the behavioral information and the reference behavioral information of each known strategy in the strategy behavioral information set" can include:

[0058] The behavioral information is obtained by extracting behavioral features from the trained feature extraction model.

[0059] The trained feature extraction model extracts behavioral features from each reference behavioral information in the policy behavior information set to obtain the reference feature information of the reference behavioral information.

[0060] Based on the similarity between the feature information and each reference feature information, the strategy corresponding to the feature information is determined to obtain the strategy feature information of the feature information.

[0061] Feature extraction can be a process of analyzing and transforming behavioral information or reference behavioral information to extract characteristic information of the behavioral information or reference behavioral information. The characteristic information and reference characteristic information can be information obtained through feature extraction.

[0062] In one embodiment, a trained feature extraction model can be obtained by training an initial feature extraction model. Specifically, before the step "extracting behavioral features from behavioral information using the trained feature extraction model to obtain the feature information of the behavioral information," the following can be included:

[0063] Acquire behavioral information samples and a set of strategies;

[0064] Behavioral features are extracted from behavioral information samples using an initial feature extraction model to obtain sample feature information of the behavioral information samples;

[0065] For each policy in the policy set, predict the probability that the behavioral information sample belongs to that policy based on the sample feature information;

[0066] Based on the policy labels and probabilities of behavioral information samples, the initial feature extraction model is trained to obtain the trained feature extraction model.

[0067] Among them, the behavioral information sample can be the training sample of the initial feature extraction model, and the behavioral information sample carries a policy label.

[0068] The strategy set can be a collection of multiple strategies, where each strategy can be equivalent to a category.

[0069] For example, a specific approach could be to extract features from behavioral information samples using an initial feature extraction model to obtain sample feature information, and then use a normalization function (softmax) or an additive margin softmax (AM-Softmax) function to predict the probability that the behavioral information sample belongs to each policy in the policy set based on the sample feature information.

[0070] Using the loss function corresponding to the function, the loss is calculated based on the probability between the behavioral information sample and each policy and its sample label. The calculated loss is then used for backpropagation to train the initial feature extraction model. The model parameters of the initial feature extraction model are adjusted to make the extracted features more accurate and better reflect the behavioral information features, until the preset training completion conditions are met, resulting in the trained feature extraction model.

[0071] In one embodiment, such as Figure 3 As shown, the initial feature extraction model can include a BI-LSTM model and an AM-Softmax classification model. The initial feature model performs embedding or one-hot encoding on the behavioral information samples to obtain embedded feature information x, x = [x0, x1, x2, x3]. This embedded feature information is then used as input to the BI-LSTM model. After the BI-LSTM model encodes the embedded feature information and performs feature fusion, the output y is obtained. Based on the output y, the probability that the behavioral information sample belongs to each policy in the policy set is predicted using the AM-Softmax classification model. , For the corresponding set of strategies, i.e. ,Right now .

[0072] The loss is calculated based on the policy label of the behavioral information sample and the probability of the behavioral information sample for each policy in the policy set:

[0073]

[0074] in, represent and The included angle, 30 is acceptable. A value of 0.35 is acceptable. t represents the policy label of the behavioral information sample.

[0075] The initial feature extraction model is trained by minimizing this loss to obtain the trained feature extraction model.

[0076] Optionally, in application, the trained feature extraction model can include only the trained BI-LSTM model, excluding the classification model. The BI-LSTM model is used to extract behavioral features from the behavioral information and reference behavioral information in the initial object set, resulting in feature information and reference feature information. Based on the similarity between the feature information and each reference feature information, the policy corresponding to the reference feature information with the highest similarity is determined as the policy of the behavioral information. Based on the policy corresponding to the behavioral information, the policy feature information of the object is obtained.

[0077] 102. For each object's strategy behavior information, perform implicit behavior information mining to obtain the object's frequent behavior information.

[0078] Among them, frequent behavior information can be behavior information mined based on strategy behavior information, which represents behavior information that the object executing the strategy behavior information may also execute.

[0079] For example, it could be based on the frequency of each behavior in the strategy behavior information of each object, such as merging behaviors that meet the frequency threshold to obtain a frequent behavior sequence.

[0080] Optionally, frequent behavior information can be gradually mined to improve its accuracy. Specifically, the step "perform implicit behavior information mining processing on the strategy behavior information of each object to obtain the frequent behavior information corresponding to the object" can be:

[0081] The current prefix information of the strategy behavior information is determined based on the preset support, and the current prefix information is the first information in the strategy behavior information.

[0082] Based on the prefix information, the corresponding suffix information is mined. The suffix information is the second information in the strategy behavior information. The second information and the first information are non-overlapping information in the strategy behavior information.

[0083] The sub-behavior information that meets the preset support in the suffix information is merged with the current prefix information to obtain the target prefix information;

[0084] Based on the current prefix information and the target prefix information, determine the frequent behavior information corresponding to each object.

[0085] The preset support can be the number of times the sub-behavior information appears. For example, if an object has a total of four policy behavior information and the preset support is 0.5, then the sub-behavior information that appears in at least two policy behavior information is the sub-behavior information that satisfies the preset support.

[0086] Among them, sub-behavior information can be behaviors in strategy behavior information. For example, if the behavior information is agjief, then a, g, j, i, e, and f are all sub-behavior information of this behavior information.

[0087] For example, let's take a strategy behavior information consisting of four behavior sequences: agjief, gefaei, gbefi, and bjigec, with a preset support of 0.5 as an example. Starting with a prefix information of length 1 (called a prefix), the sub-behavior information that satisfies the preset support is determined as the current prefix information. The sub-behavior information appearing in at least two behavior sequences are: a, b, e, f, g, and i. a, b, e, f, g, and i are determined as the current prefix information. The suffix information corresponding to each current prefix information is shown below:

[0088]

[0089] After mining the current prefix information of length 1, the suffix information corresponding to the prefix information of length 2 is mined. Taking a prefix information g as an example: the sub-behavioral information that satisfies the preset support in the suffix information corresponding to g are e, f, and i; the sub-behavioral information that satisfies the preset support and the current prefix information are merged to obtain the target prefix information ge, gf, and gi. The current prefixes a, b, e, f, g, and i, and the target prefix information ge, gf, and gi are taken as the frequent behavioral information of this object.

[0090] Optionally, the strategy behavior information can be continuously mined until the current prefix information has no corresponding suffix information, i.e., the step "Based on the current prefix information and the target prefix information, determine the frequent behavior information corresponding to each object", which can specifically include:

[0091] The target prefix information is used as the current prefix information. The corresponding suffix information is mined based on the prefix information. The sub-behavior information that meets the preset support in the suffix information is merged with the current prefix information to obtain the target prefix information, until the current prefix information has no corresponding suffix information.

[0092] Based on each current prefix information obtained, determine the frequent behavior information corresponding to each object.

[0093] For example, taking a prefix information g as an example, its corresponding target prefix information is used as the current prefix information. We continue to mine the suffix information of the current prefix information with a length of 2. The suffix information corresponding to each current prefix information is as follows:

[0094]

[0095] For a binary prefix ge, the sub-behavioral information that satisfies the preset support is f and i. F and i are merged with ge respectively to obtain the target prefix information gef and fei.

[0096] Using the target prefix information gef and fei as the current prefix information, we continue to mine the suffix information corresponding to the three prefixes. The suffix information corresponding to each three-term prefix is ​​as follows:

[0097]

[0098] The suffix information corresponding to the three prefixes does not contain sub-behavioral information that meets the preset support. Therefore, the mining of frequent behavioral information ends. Each current prefix information obtained is used as frequent sequence information for the following: one-term prefix: a, b, e, f, g, and i; two-term prefix: ge, gf, and gi; and three-term prefix: gef and fei.

[0099] Optionally, the prefixspan algorithm can be used to mine implicit behavioral information for the policy behavior information of each object to obtain the frequent behavior information of the object.

[0100] 103. Based on object feature information and frequent behavior information, cluster the objects in the object set to obtain the target clustering result.

[0101] Clustering can be a process of dividing a set of objects into different object classes or clusters according to a specific criterion (such as distance), so that the similarity of objects within the same object cluster is as high as possible, while the differences between objects in different object clusters are also as high as possible.

[0102] For example, K-means clustering, DBSCAN clustering, or other clustering algorithms can be used to cluster objects in a set of objects based on object feature information and frequent behavior information to obtain the target clustering result.

[0103] Optionally, multiple clustering algorithms can be used to obtain multiple clustering results, which can then be integrated to obtain the target clustering result. For example, given five objects a, b, c, d, and e, three clustering results can be obtained, as shown below:

[0104]

[0105] Clustering results 1 and 2 group objects a and b into one cluster, and objects c, d, and e into another. Clustering result 3 groups object a into one cluster, object b into another, and objects c, d, and e into yet another. Since more clustering results group objects a and b into one cluster, and objects c, d, and e into another, the target clustering result is that objects a and b are in one cluster, and objects c, d, and e are in another.

[0106] Optionally, clustering can also be performed based on different features. For example, clustering can be performed based on multiple features in object feature information, or clustering can be performed based on multiple features in frequent behavior information, resulting in multiple clustering results. These multiple clustering results are then integrated to obtain the target clustering result. Specifically, the step "clustering objects in the object set based on object feature information and frequent behavior information to obtain the target clustering result" can be as follows:

[0107] Based on object feature information, the objects in the object sample are clustered to obtain the first clustering result;

[0108] The objects in the object sample are clustered based on frequent behavior information to obtain the second clustering result;

[0109] The target clustering result is determined based on the first and second clustering results.

[0110] For example, the first clustering result could include the clustering result of each feature in the object's feature information, and the second clustering result could include the clustering result of each feature in the frequent behavior information. Integrating the first and second clustering results yields the target clustering result. The integration of the first and second clustering results can refer to the process described above for integrating multiple clustering results obtained from multiple clustering algorithms, and will not be elaborated upon here.

[0111] 104. Determine the target object cluster based on the distribution ratio of the object samples using the auxiliary program in each object cluster of the target clustering result.

[0112] For example, each category in the target result can be a cluster of objects. Each cluster of objects may contain objects to be predicted and object samples using auxiliary programs. Based on the proportion of object samples in each cluster, the clusters of objects whose proportions meet certain conditions are determined as the target clusters of objects. These conditions can be the highest proportion or the proportion exceeding a certain threshold.

[0113] 105. Identify the objects to be predicted contained in the target object cluster as objects for which auxiliary programs are used.

[0114] For example, specifically, because objects within the same object cluster have high similarity, the objects to be predicted contained in the target object cluster are identified as objects for which auxiliary programs are used.

[0115] As can be seen from the above, this embodiment of the application obtains an object set, which includes object samples using the auxiliary program and objects to be predicted, as well as object feature information and strategy behavior information corresponding to each object; it performs implicit behavior information mining processing on the strategy behavior information of each object to obtain frequent behavior information corresponding to each object; based on the object feature information and frequent behavior information, it clusters the objects in the object set to obtain target clustering results; it determines the target object cluster based on the distribution ratio of object samples using the auxiliary program in each object cluster of the target clustering results; and it identifies the objects to be predicted contained in the target object cluster as objects using the auxiliary program. This scheme can obtain frequent behavior information of objects by mining the strategy behavior information of objects. Frequent behavior information can reflect the behavioral characteristics of objects. Clustering based on frequent behavior information can cluster objects with similar behavioral characteristics, and then identify the objects to be predicted using the auxiliary program through object samples. It can be applied to different application scenarios and has high flexibility.

[0116] Based on the above embodiments, the following examples will provide further detailed explanations.

[0117] This embodiment will be described from the perspective of a behavior recognition device, which can be integrated into a computer device, such as a server.

[0118] This application provides a behavior recognition method, such as... Figure 4 As shown, the specific process of this behavior recognition method can be as follows:

[0119] 201. The server obtains behavioral information samples and policy sets, trains the initial feature extraction model, and obtains the trained feature extraction model.

[0120] For example, one-hot encoding or embedding can be used to embed features of behavioral information samples to obtain embedding vectors of behavioral information samples. The embedding vectors are used as input features of the initial feature extraction model. The initial feature extraction model extracts features from the feature vectors of behavioral information samples to obtain sample feature information. The additive margin softmax (AM-Softmax) function is used to predict the probability that the behavioral information sample belongs to each policy in the policy set based on the sample feature information.

[0121] Using the loss function corresponding to the AM-Softmax function, the loss is calculated based on the probability between the behavioral information sample and each policy in the policy set, as well as its sample label. The calculated loss is then used for backpropagation to train the initial feature extraction model. The model parameters of the initial feature extraction model are adjusted to make the extracted features more accurate and better reflect the behavioral information, until the preset training completion conditions are met, resulting in the trained feature extraction model.

[0122] 202. The server preprocesses the initial object information in the initial object set to obtain object feature information.

[0123] For example, the initial object information could include multiple features for each object, such as age, gender, location, and zodiac sign. If the number of missing feature data exceeds a threshold, the feature is discarded. The threshold can be flexibly set according to the specific application scenario.

[0124] Optionally, the server can also process outliers in the initial object information. For example, it can sort the feature data contained in each feature and filter the feature data whose values ​​meet the conditions. The condition can be that the feature data is ranked in the top 0.01%, that is, the feature data whose values ​​are ranked in the top 0.01% are discarded.

[0125] Optionally, after the server discards features and filters some feature data, there may still be missing feature data in some features. The missing feature data can be filled in to make the feature data in each feature complete. For example, for features with continuous feature data, the average value of the feature data can be used to fill in the missing feature data, and for features with discrete feature data, a constant can be used to fill in the missing feature data.

[0126] Optionally, the server can also transform or combine the features in the initial object information to generate new features that are meaningful and more informational. For example, this can be done by combining and deriving features through feature transformation, feature squaring, feature addition and subtraction.

[0127] Optionally, features can also be encoded, such as binning and discretizing continuous features, or one-hot encoding discrete features.

[0128] The server obtains object characteristic information by performing the above processing on the initial object information.

[0129] 203. The server determines the policy behavior information corresponding to the behavior information through the trained feature extraction model to obtain the object set.

[0130] For example, the server can use a trained feature extraction model based on a Long Short-Term Memory (LSTM) network, or a neural network composed of a forward LSTM and a backward LSTM (Bi-directional Long Short-Term Memory, Bi-LSTM), or a gated neural network (GateRecurrent Unit, GRU) to extract features from each reference behavior information in the behavioral information and policy behavior information sets, respectively, to obtain the feature information corresponding to the behavior information and the reference feature information of the reference behavior information.

[0131] The server determines the strategy corresponding to the behavioral information based on the similarity between the feature information and each reference feature information. Optionally, the feature information can be in the form of feature vectors, and the reference feature information can be in the form of reference feature vectors. The similarity between the behavioral information and the reference behavioral information is calculated based on the distance between the feature vector and the reference feature vector. The distance between the feature vector and the reference feature vector can be Manhattan distance, Euclidean distance, Chebyshev distance, cosine similarity, or Hamming distance, etc. The closer the distance between the feature vector and the reference feature vector, the greater the similarity between the behavioral information and the reference behavioral information.

[0132] The server determines the policy of the reference behavior information that has the highest similarity to the behavior information as the policy of the behavior information, and represents the behavior information through the corresponding policy identifier to obtain the policy behavior information.

[0133] 204. The server performs implicit behavior information mining on the policy behavior information of each object to obtain the frequent behavior information corresponding to the object.

[0134] For example, the server can use the prefixspan algorithm to progressively mine frequent behavior information. Sub-behaviors that satisfy a preset support level within the policy behavior information are identified as prefixes. For each prefix, the corresponding suffix information is mined. Sub-behaviors satisfying the preset support level in the corresponding suffix information are merged with the prefix to obtain a binary prefix. For each binary prefix, its corresponding suffix information is mined. Sub-behaviors satisfying the preset support level in the corresponding suffix information are merged with the binary prefix to obtain a ternary prefix. This process is iterated until no sub-behaviors satisfying the preset support level are found in the suffix information corresponding to the current prefix. Each obtained prefix is ​​then considered as the frequent behavior information of that object.

[0135] 205. Based on object feature information and frequent behavior information, the server clusters the objects in the object set to obtain the target clustering result.

[0136] For example, the server can use K-means clustering, DBSCAN clustering, or other clustering algorithms to cluster objects in a set of objects based on object feature information and frequent behavior information, and obtain the target clustering result.

[0137] Optionally, multiple clustering algorithms can be used to obtain multiple clustering results, and then the multiple clustering results can be integrated to obtain the target clustering result.

[0138] Optionally, clustering can be performed on multiple features within the object's feature information, or on multiple features within frequent behavior information, resulting in multiple clustering results. These multiple clustering results are then integrated to obtain the target clustering result.

[0139] 206. The server determines the target object cluster based on the distribution ratio of the object samples using the auxiliary program in each object cluster of the target clustering result.

[0140] For example, each category in the target result can be an object cluster. Each object cluster may contain the object to be predicted and the object samples using the auxiliary program. The server determines the object cluster that meets the condition of the object sample ratio in each object cluster as the target object cluster. The condition can be the highest ratio or the ratio exceeding a certain threshold.

[0141] 207. The server identifies the objects to be predicted contained in the target object cluster as objects for which the auxiliary program is used.

[0142] For example, the server might identify the objects to be predicted within the target object cluster as objects for which an auxiliary program is used because the objects within the same object cluster are highly similar.

[0143] As can be seen from the above, in this embodiment, the server obtains behavioral information samples and a policy set, trains an initial feature extraction model to obtain a trained feature extraction model, preprocesses the initial object information in the initial object set to obtain object feature information, determines the policy behavior information corresponding to the behavioral information through the trained feature extraction model to obtain an object set, performs implicit behavioral information mining processing on the policy behavior information of each object to obtain the frequent behavior information corresponding to the object, clusters the objects in the object set based on the object feature information and the frequent behavior information to obtain the target clustering result, determines the target object cluster based on the distribution ratio of the object samples using the auxiliary program in each object cluster of the target clustering result, and identifies the objects to be predicted contained in the target object cluster as objects using the auxiliary program. This scheme can obtain the frequent behavior information of objects by mining the policy behavior information of objects. The frequent behavior information can reflect the behavioral characteristics of objects. Clustering based on the frequent behavior information can cluster objects with similar behavioral characteristics, and then identify the objects to be predicted using the auxiliary program through object samples. It can be applied to different application scenarios and has high flexibility.

[0144] To facilitate better implementation of the behavior recognition method provided in the embodiments of this application, a behavior recognition device is also provided in one embodiment. The meanings of the terms used are the same as in the behavior recognition method described above, and specific implementation details can be found in the description of the method embodiments.

[0145] This behavior recognition device can be integrated into computer equipment, such as... Figure 5 As shown, the behavior recognition device may include: an acquisition unit 301, a mining unit 302, a clustering unit 303, a first determination unit 304, and a second determination unit 305, as detailed below:

[0146] (1) Acquisition unit 301: used to acquire a set of objects, which includes object samples and objects to be predicted using the auxiliary program, as well as object feature information and policy behavior information corresponding to each object.

[0147] For example, the object collection can be retrieved from a database or memory. Alternatively, if the object collection is stored in a blockchain, it can be retrieved from the blockchain.

[0148] In one embodiment, upon receiving an identification request, for example, if a game user reports a game object for using an auxiliary program, the system can obtain the object feature information corresponding to the reported object, as well as the strategic behavior information in historical game matches, and obtain the object feature information and strategic behavior information corresponding to the object sample to obtain an object set.

[0149] Optionally, one can periodically obtain object feature information corresponding to the reported object, as well as strategic behavior information in historical game competitions, and obtain object feature information and strategic behavior information corresponding to object samples to obtain an object set.

[0150] In one embodiment, the behavior recognition device may further include a set acquisition unit, a preprocessing unit, a strategy determination unit, and a set obtaining unit, specifically:

[0151] Set acquisition unit: used to acquire an initial set of objects and a set of policy behavior information. The initial set of objects includes object samples and objects to be predicted using the auxiliary program, as well as initial object information and behavior information corresponding to each object. The set of policy behavior information includes reference behavior information of at least one known policy.

[0152] Preprocessing unit: used to preprocess the initial object information to obtain object feature information;

[0153] Strategy determination unit: used to determine the strategy of the behavior information based on the similarity between the behavior information and the reference behavior information of each known strategy in the strategy behavior information set, so as to obtain the strategy behavior information corresponding to the behavior information;

[0154] The set of units is used to obtain a set of objects based on object samples, objects to be predicted, and object feature information and policy behavior information corresponding to each object.

[0155] For example, the initial object information could include multiple features for each object, such as age, gender, location, and zodiac sign. If the number of missing feature data exceeds a threshold, the feature is discarded. The threshold can be flexibly set according to the specific application scenario.

[0156] Optionally, outliers in the initial object information can also be processed. For example, the feature data contained in each feature can be sorted, and the feature data whose values ​​meet the conditions can be filtered. The condition can be that the feature data is ranked in the top 0.01%, that is, the feature data whose values ​​are ranked in the top 0.01% can be discarded.

[0157] Optionally, after discarding features and filtering some feature data, there may still be missing feature data in some features. The missing feature data can be filled in to make the feature data in each feature complete. For example, for features with continuous feature data, the average value of the feature data can be used to fill in the missing feature data, and for features with discrete feature data, a constant can be used to fill in the missing feature data.

[0158] Optionally, features in the initial object information can be transformed or combined to generate new features that are meaningful and more informational. For example, features can be combined and derived through feature transformation, feature squaring, feature addition and subtraction.

[0159] Optionally, features can also be encoded, such as binning and discretizing continuous features, or one-hot encoding discrete features.

[0160] For behavioral information, the strategy corresponding to the behavioral information can be determined based on its similarity to the reference behavioral information of each known strategy in the set of strategy behavioral information. For example, the strategy of the reference behavioral information with the highest similarity is determined as the strategy corresponding to the behavioral information. The behavioral information is represented by the identifier of the corresponding strategy. For example, specifically, the behavioral information can be agjief, where the behavior ahj conforms to the common feature information of strategy A1, so the behavior ahj can be represented by A1; the behavior ef conforms to the common feature information of strategy A3, so the behavior ahj can be represented by A3; the behavior i does not conform to the common feature information of any strategy in the set of strategy common information, so the behavior i can be represented separately, for example, denoted as 1. The strategy behavioral information corresponding to the behavioral information agjief is A11 A3.

[0161] Optionally, the policy determination unit may include a first feature extraction subunit, a second feature extraction subunit, and a policy determination subunit, specifically:

[0162] The first feature extraction subunit is used to extract behavioral features from behavioral information using the trained feature extraction model, thereby obtaining the feature information of the behavioral information.

[0163] The second feature extraction subunit is used to extract behavioral features from each reference behavioral information in the policy behavior information set through the trained feature extraction model, so as to obtain the reference feature information of the reference behavioral information.

[0164] Strategy determination subunit: used to determine the strategy corresponding to the feature information based on the similarity between the feature information and each reference feature information, so as to obtain the strategy feature information of the feature information.

[0165] In one embodiment, the policy determination unit may include an acquisition subunit, a sample feature extraction subunit, a prediction subunit, and a training subunit, specifically:

[0166] Acquisition Subunit: Used to acquire behavioral information samples and policy sets;

[0167] Sample feature extraction subunit: used to extract behavioral features from behavioral information samples through the initial feature extraction model, and obtain sample feature information of behavioral information samples;

[0168] Prediction subunit: Used to predict the probability that a behavioral information sample is a policy for each policy in the policy set, based on sample feature information;

[0169] Training subunit: Used to train the initial feature extraction model based on the policy labels and probabilities of behavioral information samples, resulting in a trained feature extraction model.

[0170] For example, a specific approach could be to extract features from behavioral information samples using an initial feature extraction model to obtain sample feature information, and then use a normalization function (softmax) or an additive margin softmax (AM-Softmax) function to predict the probability that the behavioral information sample belongs to each policy in the policy set based on the sample feature information.

[0171] Using the loss function corresponding to the function, the loss is calculated based on the probability between the behavioral information sample and each policy and its sample label. The calculated loss is then used for backpropagation to train the initial feature extraction model. The model parameters of the initial feature extraction model are adjusted to make the extracted features more accurate and better reflect the behavioral information features, until the preset training completion conditions are met, resulting in the trained feature extraction model.

[0172] (2) Mining unit 302: used to perform implicit behavior information mining processing on the strategy behavior information of each object to obtain the frequent behavior information corresponding to the object.

[0173] For example, it could be based on the frequency of each behavior in the strategy behavior information of each object, such as merging behaviors that meet the frequency threshold to obtain a frequent behavior sequence.

[0174] Optionally, the mining unit 302 may include an information determination subunit, a mining subunit, a merging subunit, and a frequent information determination subunit, specifically:

[0175] Information Determination Subunit: Used to determine the current prefix information of the strategy behavior information based on a preset support level. The current prefix information is the first information in the strategy behavior information.

[0176] Mining subunit: used to mine corresponding suffix information based on prefix information. The suffix information is the second information in the strategy behavior information. The second information and the first information are non-overlapping information in the strategy behavior information.

[0177] Merge sub-unit: Used to merge the sub-behavior information in the suffix information that meets the preset support with the current prefix information to obtain the target prefix information;

[0178] Frequent information determination subunit: used to determine the frequent behavior information corresponding to each object based on the current prefix information and the target prefix information.

[0179] Optionally, the frequent information determination subunit may include a loop module and a determination module, specifically:

[0180] The loop module is used to take the target prefix information as the current prefix information, return to perform the mining of the corresponding suffix information based on the prefix information, merge the sub-behavior information in the suffix information that meets the preset support with the current prefix information to obtain the target prefix information, until the current prefix information has no corresponding suffix information;

[0181] Determining module: Used to determine the frequent behavior information corresponding to each object based on each current prefix information obtained.

[0182] (3) Clustering unit 303: used to cluster objects in the object set based on object feature information and frequent behavior information to obtain the target clustering result.

[0183] For example, K-means clustering, DBSCAN clustering, or other clustering algorithms can be used to cluster objects in a set of objects based on object feature information and frequent behavior information to obtain the target clustering result.

[0184] Optionally, clustering unit 303 may include a first clustering subunit, a second clustering subunit, and a result determination subunit, specifically:

[0185] The first clustering subunit is used to cluster objects in the object sample based on object feature information to obtain the first clustering result;

[0186] The second clustering subunit is used to cluster objects in the object sample based on frequent behavior information to obtain the second clustering result.

[0187] Result Determination Subunit: Used to determine the target clustering result based on the first clustering result and the second clustering result.

[0188] For example, the first clustering result could include the clustering result of each feature in the object's feature information, and the second clustering result could include the clustering result of each feature in the frequent behavior information. Integrating the first and second clustering results yields the target clustering result. The integration of the first and second clustering results can refer to the process described above for integrating multiple clustering results obtained from multiple clustering algorithms, and will not be elaborated upon here.

[0189] (4) First determining unit 304: used to determine the target object cluster based on the distribution ratio of the object samples using the auxiliary program in each object cluster of the target clustering result.

[0190] For example, each category in the target result can be a cluster of objects. Each cluster of objects may contain objects to be predicted and object samples using auxiliary programs. Based on the proportion of object samples in each cluster, the clusters of objects whose proportions meet certain conditions are determined as the target clusters of objects. These conditions can be the highest proportion or the proportion exceeding a certain threshold.

[0191] (5) Second determining unit 305: used to determine the objects to be predicted contained in the target object cluster as objects to be used with the auxiliary program.

[0192] For example, specifically, because objects within the same object cluster have high similarity, the objects to be predicted contained in the target object cluster are identified as objects for which auxiliary programs are used.

[0193] As can be seen from the above, in this embodiment, the acquisition unit 301 acquires an object set, which includes object samples using the auxiliary program and objects to be predicted, as well as object feature information and strategy behavior information corresponding to each object; the mining unit 302 performs implicit behavior information mining processing on the strategy behavior information of each object to obtain frequent behavior information corresponding to each object; the clustering unit 303 clusters the objects in the object set based on the object feature information and frequent behavior information to obtain the target clustering result; the first determining unit 304 determines the target object cluster based on the distribution ratio of the object samples using the auxiliary program in each object cluster of the target clustering result; and the second determining unit 305 determines the objects to be predicted contained in the target object cluster as objects using the auxiliary program. This scheme can obtain the frequent behavior information of the objects by mining the strategy behavior information of the objects. The frequent behavior information can reflect the behavior characteristics of the objects. Clustering based on the frequent behavior information can cluster objects with similar behavior characteristics, and then identify the objects to be predicted using the auxiliary program through object samples. It can be applied to different application scenarios and has high flexibility.

[0194] This application also provides a computer device, which can be a terminal or a server, such as... Figure 6 As shown, it illustrates a structural schematic diagram of the computer device involved in the embodiments of this application, specifically:

[0195] The computer device may include components such as a processor 1001 with one or more processing cores, a memory 1002 with one or more computer-readable storage media (also referred to as storage media), a power supply 1003, and an input unit 1004. Those skilled in the art will understand that... Figure 6The computer device structure shown does not constitute a limitation on the computer device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0196] The processor 1001 is the control center of the computer device. It connects various parts of the computer device via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 1002, and by calling data stored in the memory 1002, it performs various functions of the computer device and processes data, thereby performing overall detection of the computer device. Optionally, the processor 1001 may include one or more processing cores; preferably, the processor 1001 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and computer programs, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 1001.

[0197] The memory 1002 can be used to store software programs and modules. The processor 1001 executes various functional applications and data processing by running the software programs and modules stored in the memory 1002. The memory 1002 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, computer programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 1002 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 1002 may also include a memory controller to provide the processor 1001 with access to the memory 1002.

[0198] The computer equipment also includes a power supply 1003 that supplies power to the various components. Preferably, the power supply 1003 can be logically connected to the processor 1001 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 1003 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0199] The computer device may also include an input unit 1004, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0200] Although not shown, the computer device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 1001 in the computer device loads the executable files corresponding to the processes of one or more computer programs into the memory 1002 according to the following instructions, and the processor 1001 runs the computer programs stored in the memory 1002 to realize various functions, as follows:

[0201] Obtain the object set, which includes object samples using the auxiliary program and the object to be predicted, as well as object feature information and policy behavior information corresponding to each object;

[0202] For each object's strategy behavior information, implicit behavior information mining is performed to obtain the object's frequent behavior information;

[0203] Based on object feature information and frequent behavior information, the objects in the object set are clustered to obtain the target clustering result;

[0204] The target object cluster is determined based on the distribution proportion of the object samples using the auxiliary program in each object cluster of the target clustering result;

[0205] The objects to be predicted contained in the target object cluster are identified as objects for which auxiliary programs are used.

[0206] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0207] As can be seen from the above, this embodiment of the application obtains an object set, which includes object samples using the auxiliary program and objects to be predicted, as well as object feature information and strategy behavior information corresponding to each object; it performs implicit behavior information mining processing on the strategy behavior information of each object to obtain frequent behavior information corresponding to each object; based on the object feature information and frequent behavior information, it clusters the objects in the object set to obtain target clustering results; it determines the target object cluster based on the distribution ratio of object samples using the auxiliary program in each object cluster of the target clustering results; and it identifies the objects to be predicted contained in the target object cluster as objects using the auxiliary program. This scheme can obtain frequent behavior information of objects by mining the strategy behavior information of objects. Frequent behavior information can reflect the behavioral characteristics of objects. Clustering based on frequent behavior information can cluster objects with similar behavioral characteristics, and then identify the objects to be predicted using the auxiliary program through object samples. It can be applied to different application scenarios and has high flexibility.

[0208] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations of the above embodiments.

[0209] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by a computer program, or by a computer program controlling related hardware. The computer program can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0210] Therefore, embodiments of this application provide a computer-readable storage medium storing a computer program that can be loaded by a processor to execute any of the behavior recognition methods provided in embodiments of this application.

[0211] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0212] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0213] Since the computer program stored in the computer-readable storage medium can execute any of the behavior recognition methods provided in the embodiments of this application, it can achieve the beneficial effects that any of the behavior recognition methods provided in the embodiments of this application can achieve, as detailed in the preceding embodiments, and will not be repeated here.

[0214] The above provides a detailed description of a behavior recognition method, apparatus, computer device, and computer-readable storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A behavior recognition method, characterized by, The method comprises the following steps: obtaining an object set, the object set comprising an object sample using an auxiliary program and an object to be predicted, and object feature information and strategy behavior information corresponding to each object, the strategy behavior information being behavior information identified by a strategy, the behavior information comprising a behavior sequence of all or part of behaviors of the object in a game; performing implicit behavior information mining processing on the strategy behavior information of each object to obtain frequent behavior information corresponding to the object; clustering the objects in the object set based on the object feature information and the frequent behavior information to obtain a target clustering result, wherein the objects in the object set are clustered based on the object feature information to obtain a first clustering result, the objects in the object set are clustered based on the frequent behavior information to obtain a second clustering result, and the target clustering result is determined according to the first clustering result and the second clustering result; determining a target object cluster according to a distribution proportion of the object sample using the auxiliary program in each object cluster of the target clustering result; determining the object to be predicted included in the target object cluster as an object using the auxiliary program.

2. The method of claim 1, wherein, Before the step of obtaining the object set, the method further comprises the following steps: obtaining an initial object set and a strategy behavior information set, the initial object set comprising an object sample using an auxiliary program and an object to be predicted, and initial object information and behavior information corresponding to each object, and the strategy behavior information set comprising reference behavior information of at least one known strategy; preprocessing the initial object information to obtain the object feature information; determining a strategy of the behavior information according to a similarity between the behavior information and reference behavior information of each known strategy in the strategy behavior information set, to obtain the strategy behavior information corresponding to the behavior information; obtaining the object set based on the object sample and the object to be predicted, and the object feature information and the strategy behavior information corresponding to each object.

3. The method of claim 2, wherein, The step of determining a strategy of the behavior information according to a similarity between the behavior information and reference behavior information of each known strategy in the strategy behavior information set comprises the following steps: performing behavior feature extraction on the behavior information by using a trained feature extraction model to obtain feature information of the behavior information; performing behavior feature extraction on each reference behavior information in the strategy behavior information set by using the trained feature extraction model to obtain reference feature information of the reference behavior information; determining a strategy corresponding to the behavior information according to a similarity between the feature information and each reference feature information.

4. The method of claim 3, wherein, Before the step of performing behavior feature extraction on the behavior information by using a trained feature extraction model to obtain feature information of the behavior information, the method further comprises the following steps: obtaining a behavior information sample and a strategy set; performing behavior feature extraction on the behavior information sample by using an initial feature extraction model to obtain sample feature information of the behavior information sample; predicting a probability of the behavior information sample being the strategy for each strategy in the strategy set based on the sample feature information; Based on the policy label of the behavior information sample and the probability, the initial feature extraction model is trained to obtain a trained feature extraction model.

5. The method of claim 1, wherein, The strategy behavior information of each object is subjected to implicit behavior information mining processing to obtain corresponding frequent behavior information of each object, including: Current prefix information of the strategy behavior information is determined based on a preset support degree, and the current prefix information is first information in the strategy behavior information. Corresponding suffix information is mined based on the prefix information, and the suffix information is second information in the strategy behavior information, and the second information and the first information are non-overlapping information in the strategy behavior information. Sub-behavior information in the suffix information satisfying the preset support degree is combined with the current prefix information to obtain target prefix information. Based on the current prefix information and the target prefix information, the frequent behavior information corresponding to each object is determined.

6. The method of claim 5, wherein, The frequent behavior information corresponding to each object is determined based on the current prefix information and the target prefix information, including: The target prefix information is taken as the current prefix information, and the corresponding suffix information is mined based on the prefix information. The sub-behavior information in the suffix information satisfying the preset support degree is combined with the current prefix information to obtain target prefix information, until there is no corresponding suffix information for the current prefix information. According to the obtained current prefix information, the frequent behavior information corresponding to each object is determined.

7. A behavior recognition apparatus characterized by comprising: It includes: An acquisition unit is configured to acquire an object set, the object set including object samples using an auxiliary program and to-be-predicted objects, and object feature information and strategy behavior information corresponding to each object, the strategy behavior information being behavior information identified by a strategy, and the behavior information containing behavior sequences of all or part of behaviors of the object in a game; An excavation unit is configured to perform implicit behavior information mining processing on the strategy behavior information of each object to obtain frequent behavior information corresponding to the object; A clustering unit is configured to cluster objects in the object set based on the object feature information and the frequent behavior information to obtain a target clustering result, wherein the objects in the object set are clustered based on the object feature information to obtain a first clustering result, the objects in the object set are clustered based on the frequent behavior information to obtain a second clustering result, and the target clustering result is determined according to the first clustering result and the second clustering result; A first determination unit is configured to determine a target object cluster according to a distribution proportion of the object samples using the auxiliary program in each object cluster of the target clustering result; A second determination unit is configured to determine the to-be-predicted objects included in the target object cluster as objects using the auxiliary program.

8. A computer device, comprising: It includes a memory and a processor; the memory stores a computer program, and the processor is configured to run the computer program in the memory to execute the behavior recognition method in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium is configured to store a computer program, and the computer program is loaded by a processor to execute the behavior recognition method in any one of claims 1 to 6.

10. A computer program product, characterised in that, The computer program product comprises computer instructions stored in a computer readable storage medium, and a processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the behavior recognition method in any one of claims 1 to 6.

Images