Object processing method, apparatus, device, and storage medium

Abstract

Embodiments of the present application disclose an object processing method and device, equipment and a storage medium, which can be applied to the fields of artificial intelligence and computer technology. The method comprises: obtaining a first data set, the first data set comprising interaction data of all objects in a first time period; determining a first weight matrix according to the first data set, a matrix element of an i-th row and a j-th column in the first weight matrix representing a number of objects having an interaction relationship with an object i in a j-th time slice in the first time period; performing singular value decomposition on the first weight matrix, and determining an interaction behavior feature of each object in the first time period according to a decomposition result; and performing object processing according to the interaction behavior features of all objects in the first time period. According to the embodiments of the present application, the interaction behavior features of the objects can be accurately represented, and the object processing effect is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an object processing method, apparatus, device, and storage medium. Background Technology

[0002] With the continuous development of computer technology, communication through interactive networks (such as social networks and game systems) has gradually become an indispensable interaction method in people's daily lives. However, as the number of objects in interactive networks continues to increase and the interactions between objects become more frequent, how to effectively represent the interactive behavior characteristics of each object in the interactive network and effectively process the objects based on these characteristics has become an urgent problem to be solved. Summary of the Invention

[0003] This application provides an object processing method, apparatus, device, and storage medium that can accurately represent the interactive behavior characteristics of objects, improve object processing performance, and has high applicability.

[0004] On one hand, embodiments of this application provide an object processing method, the method comprising:

[0005] Obtain the first dataset, which includes the interaction data of all objects in the object collection during the first time period;

[0006] The first weight matrix is ​​determined based on the first dataset mentioned above. The matrix element in the i-th row and j-th column of the first weight matrix represents the number of objects that interact with object i in the j-th time slice of the first time period. i is a positive integer less than or equal to N, j is a positive integer less than or equal to M, N is the number of objects in the object set mentioned above, and M is the number of time slices in the first time period mentioned above.

[0007] Singular value decomposition is performed on the first weight matrix, and the interaction behavior characteristics of each of the above objects in the first time period are determined based on the decomposition results.

[0008] Object processing is performed based on the interaction behavior characteristics of all the aforementioned objects during the first time period.

[0009] On the other hand, embodiments of this application provide an object processing apparatus, the apparatus comprising:

[0010] The data acquisition module is used to acquire a first dataset, which includes the interaction data of all objects in the object collection during a first time period.

[0011] The feature processing module is used to determine a first weight matrix based on the first dataset mentioned above. The matrix element in the i-th row and j-th column of the first weight matrix represents the number of objects that interact with object i in the j-th time slice of the first time period. i is a positive integer less than or equal to N, j is a positive integer less than or equal to M, N is the number of objects in the object set mentioned above, and M is the number of time slices in the first time period mentioned above.

[0012] The aforementioned feature processing module is used to perform singular value decomposition on the aforementioned first weight matrix and determine the interaction behavior features of each of the aforementioned objects in the aforementioned first time period based on the decomposition results.

[0013] The object processing module is used to process objects based on the interaction behavior characteristics of all the above objects during the first time period.

[0014] On the other hand, embodiments of this application provide an electronic device, including a processor and a memory, which are interconnected;

[0015] The aforementioned memory is used to store computer programs;

[0016] The processor described above is used to execute the object processing method provided in the embodiments of this application when the computer program described above is invoked.

[0017] On the other hand, embodiments of this application provide a computer-readable storage medium storing a computer program that is executed by a processor to implement the object processing method provided in embodiments of this application.

[0018] On the other hand, embodiments of this application provide a computer program product, which includes a computer program that, when executed by a processor, implements the object processing method provided in embodiments of this application.

[0019] In this embodiment, by acquiring all interaction data of the object set within the first time period, it is ensured that the analysis covers all potential interaction scenarios, avoiding biases caused by data omissions, thereby improving the accuracy of the expression of interaction behavior features. Simultaneously, by constructing a first weight matrix, the interaction frequency within the time slice is quantified into matrix elements, which helps to transform complex object interaction behaviors into structured mathematical expressions. This enables the processing of large-scale sparse data in interaction networks with a large number of objects and interactions, improving data processing efficiency. Singular value decomposition can transform high-dimensional data into a low-dimensional feature space, significantly reducing the computational and representational complexity of interaction behavior features while improving object processing performance. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0021] Figure 1 This is a diagram illustrating the application environment of the object processing method in the embodiments of this application;

[0022] Figure 2 This is a flowchart illustrating the object processing method provided in an embodiment of this application;

[0023] Figure 3 This is a schematic diagram of a scenario using the timing dynamic diagram provided in the embodiments of this application;

[0024] Figure 4 This is a schematic diagram of a construction timing bipartite diagram provided in an embodiment of this application;

[0025] Figure 5 This is a schematic diagram illustrating a scenario of recommending friends, as demonstrated in an embodiment of this application.

[0026] Figure 6 This is a flowchart illustrating the target feature matrix update method provided in an embodiment of this application;

[0027] Figure 7 This is a schematic diagram of the structure of the object processing device provided in the embodiments of this application;

[0028] Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0029] 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.

[0030] This application provides an object processing method applicable to the field of computer technology. Specifically, it can determine the interactive behavior characteristics of different objects within a certain time period based on artificial intelligence technology, and then accurately process different objects according to the interactive behavior characteristics.

[0031] For example, based on the object processing method provided in the embodiments of this application, the interactive behavior characteristics of different social accounts in a social network within a certain period of time can be accurately determined, thereby enabling targeted management of different accounts based on the interactive behavior characteristics, such as predicting account activity and detecting account anomalies.

[0032] For example, based on the object processing method provided in the embodiments of this application, the interaction behavior characteristics of different game accounts (virtual objects) in a game application within a certain period of time can be determined, thereby enabling friend recommendations for game accounts (virtual objects) based on the interaction behavior characteristics.

[0033] In addition to the aforementioned social media accounts and game accounts, the objects involved in this application embodiment can also be employee accounts, blockchain addresses, shopping accounts, interactive devices, etc., without any restrictions.

[0034] In this context, any of the aforementioned accounts can also be considered virtual objects in the relevant field. For example, a game account can be considered a virtual character or virtual object in a game, and no restrictions are imposed here.

[0035] See Figure 1 , Figure 1 This is an application environment diagram of the object processing method in the embodiments of this application, such as... Figure 1 As shown, the object processing method provided in this application embodiment can be applied to object management networks, which may be employee management networks, game networks, e-commerce networks, or social networks, etc.

[0036] Different users can control objects on the network based on their terminal devices, such as logging into a game to control game objects and interact with each other, or logging into a social account to interact with other social accounts.

[0037] In the object management network, the server can obtain the interaction data of all objects in the object collection within a time period, and then determine a weight matrix based on the interaction data.

[0038] The element in the i-th row and j-th column of the weight matrix represents the number of objects that interact with object i within the j-th time slice of the time period. i is a positive integer less than or equal to N, j is a positive integer less than or equal to M, N is the number of objects in the object set, and M is the number of time slices within the time period.

[0039] Based on this, the server can perform singular value decomposition on the weight matrix, determine the interaction behavior characteristics of each object in the object network within the time period based on the decomposition results, and then process the objects based on the interaction behavior characteristics of each object within the first time period, such as predicting the activity status of social accounts, recommending friends for game accounts, etc.

[0040] The server can be an independent physical server, such as a game server, social network server, or employee management server; it can also be a server cluster or distributed system composed of multiple physical servers; or it can be 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.

[0041] The terminal device can be a smartphone, tablet, laptop, desktop computer, smartwatch, vehicle terminal, aircraft, smart home appliance (such as smart TV) or wearable device, etc.

[0042] It should be noted that the object processing method provided in this application embodiment can be executed by a server or jointly by a server and a terminal device, and no limitation is imposed here. For example, the object processing method provided in this application embodiment can have the server obtain the interaction data of all objects in the object set within a certain time period from the background data, determine the interaction behavior characteristics of each object within that time period based on the interaction data, and then perform object processing based on the interaction behavior characteristics of each object within that time period. Alternatively, different terminal devices can provide the server with the interaction data of related objects (such as those logged in through the terminal device) within a certain time period, and the server can determine the interaction behavior characteristics of each object within that time period based on the interaction data, and then perform object processing based on the interaction behavior characteristics of each object within that time period.

[0043] The following uses server execution as an example to further illustrate the object processing method provided in the embodiments of this application.

[0044] See Figure 2 , Figure 2 This is a flowchart illustrating the object processing method provided in an embodiment of this application. For example... Figure 2 As shown, the object processing method provided in this application embodiment may include the following steps:

[0045] S21, Obtain the first dataset.

[0046] In some feasible implementations, all objects in the object set are also all objects in the object management network, such as all game accounts or virtual objects in a game network, all social accounts in a social network, all employee accounts in an employee management network, all users in an e-commerce platform, all blockchain addresses in a blockchain network, etc. The specific object set can be determined based on the type of object association network actually applied, and there are no restrictions here.

[0047] In some feasible implementations, the first dataset includes interaction data of all objects in the object set during a first time period.

[0048] The first time period can be any historical time period, such as all past times up to the present, or the past year, past month, or past week, etc. The specific time period can be determined based on the actual application scenario requirements, and there are no restrictions here.

[0049] The interaction data for each object refers to the bidirectional or multidirectional interaction records generated by the object in the object management network, covering relevant data in dimensions such as behavioral events, state changes, relationship chains, and resource flow information.

[0050] For example, when the object is a game account, the interaction data of each game account can be the team-up records, battle records, transaction records, chat interaction records, friend relationship change records, etc. of that game account with other game accounts. The interaction data of each game account can be exported from the backend database or log system, or reported in real time by the application or terminal device where the game account is located, without any restrictions.

[0051] For example, when the object is a social media account, the interaction data for each social media account can be the contact operation records of that social media account (such as adding friends, following accounts, unfollowing accounts), etc., or it can be the social behavior records of that social media account, such as likes, comments, reposts, etc. The interaction data for each social media account can be obtained from the backend database, or it can be reported in real time by the application or terminal device where the social media account is located; there are no restrictions here.

[0052] For example, when the object is an e-commerce account, the interaction data for each e-commerce account can include its transaction records, product sharing records, comment interaction records, etc. The interaction data for each e-commerce account can be obtained from the order management system or the backend data of the backend server; there are no restrictions on this.

[0053] For example, when the object is a blockchain address in a blockchain network, the interaction data of the blockchain address can be transaction records and signature records between other blockchain addresses. The interaction data of each blockchain address can be actively reported by the users of each blockchain address, or it can be determined by the transaction snapshot generated when each blockchain address conducts a transaction. There are no restrictions here.

[0054] By acquiring the first dataset, we can comprehensively cover all interaction data (such as transactions and social behaviors) of the object set within the first time period, ensuring that the analysis foundation is complete and avoiding biases caused by missing data, thereby improving the accuracy of subsequent expression of interaction behavior features.

[0055] S22, determine the first weight matrix based on the first dataset.

[0056] In some feasible implementations, the matrix element in the i-th row and j-th column of the first weight matrix represents the number of objects that interact with object i in the j-th time slice of the first time period, where i is a positive integer less than or equal to N, j is a positive integer less than or equal to M, N is the number of objects in the object set, and M is the number of time slices in the first time period.

[0057] In this first time period, the time slices do not overlap, and all the time slices together constitute the first time period. Furthermore, the number of time slices within the first time period and the duration of each time slice can be determined based on the specific application scenario requirements and are not limited here. For example, the first time period can be allocated to multiple consecutive time slices based on a fixed duration, or the first time period can be divided into a fixed number of time slices of equal duration.

[0058] Optionally, when determining the time slice of the first time period, the time slice can be divided based on the number of active objects in the object association network. Specifically, the number of active objects in each unit time period (such as daily, weekly, hourly, etc.) can be determined, and two consecutive unit time periods with the number of active objects exceeding a threshold, as well as the unit time periods between them, can be defined as a single time slice.

[0059] In some feasible implementations, after determining the first weight matrix, for each object, the number of objects that interact with the object in each time slice can be determined based on the object's interaction data.

[0060] Interaction relationships include, but are not limited to, two-way or multi-way interaction relationships such as transactions, chats, comments, likes, and sharing. For example, the object that interacts with each object can be an object that is transacting with that object, an object that is chatting with that object, an object that is sharing information with that object, an object that is sharing information with that object, etc. There are no restrictions here.

[0061] Furthermore, the object can be used as one dimension, the time slice as another dimension, and the number of objects that interact with each object within each time slice can be used as matrix elements to construct the first weight matrix.

[0062] In some feasible implementations, the construction of the temporal dynamic graph can be determined before determining the first weight matrix. For example... Figure 3 As shown, Figure 3 This is a schematic diagram of a timing dynamic diagram provided in an embodiment of this application. In the timing dynamic diagram, if object v and object v′ are at time t x When an interactive relationship is established, it can be represented as follows: <v,v‘,t x In other words, the edge between object node v and object node v′ represents the interaction relationship, and the weight of the edge represents the time when the interaction relationship is generated.

[0063] Furthermore, the first time period can be divided into different time slices. For example, the first time period can be divided from the initial time t0, with a granularity (i.e., the duration of each time slice) of Δ. t For a certain time t within the first time interval... x This moment can be considered to belong to the time slice [t0+x·Δ]. t ,t0+(x+1)·Δ t That is to say, t x ≥t0+x·Δ t And t x <t0+(x+1)·Δ t If the interaction time t between object v and object v′ x In time slice T y So, we say <v,v‘,t x >Belongs to time slice T y , represented as <v,v‘,t x >∈T y .

[0064] Furthermore, for each time slice T y , can be built Figure 4 The timing bipartite diagram is shown. Figure 4 This is a schematic diagram of a construction timing bipartite diagram provided in an embodiment of this application. For example... Figure 4 As shown, if each object is considered an object node and each time slice is considered a time node, then the edge between each object node and each time node can be considered as... <v,b x ,w(v,T y )>.

[0065] The weight of the edge is w(v,T).y )=|{v| <v,v‘,t x >∈T y or <v‘,v,t x >∈T y}|, that is, in time slice T y The number of objects that interact with object v.

[0066] If Figure 4 If the time interval between t1 and t2 is defined as a time slice T1, and the time interval between t3 and t4 is defined as a time slice T2, then the following can be formed: Figure 4 The diagram shows a temporal bipartite graph with 7 edges. w(v2,T1) = 2 because object node v2 has 2 edges in time slice T1, meaning it interacts with two other objects in time slice T1.

[0067] If we represent the set of nodes corresponding to all time slices in the bipartite time series graph as U = {T1, T2, ..., T...} M}, where M is the number of time slices within the first time period, then the first weight matrix can be represented as a matrix. Where N is the number of objects, and M is the number of time slices within the first time period.

[0068] In other words, when determining the first weight matrix, we can first determine the number M of time slices within the first time period, i.e. Where t max t0 is the largest timestamp in the first time period, and t0 is the smallest timestamp. It takes the upper bound (for example, if x = 3.1, then...). Based on this, an empty matrix can be generated. In this matrix, each element has a value of 0. Further, a temporal bipartite graph can be constructed. Based on the weights of the edges between each object node and the time slice node in the temporal bipartite graph, the value of each element in the empty matrix H is updated to obtain the first weight matrix W.

[0069] By quantifying the frequency of interactions within time slices into matrix elements, complex behaviors are transformed into computable high-dimensional mathematical objects, which facilitates the analysis and extraction of interaction behavior features. Simultaneously, using a time-series bipartite graph reduces data processing volume, improving the efficiency of determining interaction behavior features in large-scale object networks.

[0070] S23, perform singular value decomposition on the first weight matrix, and determine the interaction behavior characteristics of each object in the first time period based on the decomposition results.

[0071] In some feasible implementations, the first weight matrix can be decomposed into three matrices by singular value decomposition: the first left singular matrix, the first singular value vector matrix, and the first right singular matrix.

[0072] The first left singular matrix is ​​used to characterize the participation intensity of each object in different object interaction modes within the first time period. The singular values ​​on the diagonal of the first singular value matrix represent the importance of different interaction modes within the first time period. The first right singular matrix represents the participation degree of each time slice within the first time period in different object interaction modes.

[0073] For example, the first weight matrix W can be obtained using the SVD matrix decomposition algorithm:

[0074] W=U∑V T

[0075] Among them, matrix and Both are orthogonal matrices, meaning UU T =I and VV T =I, where I is the unit matrix, and It is a diagonal matrix.

[0076] Further, determine the first d largest singular values ​​in the first singular value matrix, and the first singular value vector corresponding to the first d largest singular values ​​in the first left singular matrix and the first singular value matrix; where d is a positive integer and is less than or equal to the minimum value of M and N.

[0077] Based on this, the target feature matrix can be determined according to the first d largest singular values ​​in the first singular value matrix and the first singular value vector.

[0078] In this context, the feature vector formed by the matrix elements of each row in the target feature matrix represents the interactive behavior features of an object in the first time period.

[0079] In other words, according to the theory of Truncated SVD, only the top d largest singular values ​​and their corresponding singular value vectors are retained in the decomposition result, thus obtaining a low-rank approximate matrix based on the top d largest singular values ​​and their corresponding singular vectors. If Σ d U represents the first d largest singular values ​​of the first singular value matrix Σ. d Let d be the corresponding left singular vectors of the first left singular matrix U. Then the target feature matrix can be determined as Z = U. d Σ d .

[0080] By using singular value decomposition and retaining the top d largest singular values, low-frequency noisy interaction behaviors (such as accidental account logins) can be filtered out, focusing on mainstream object interaction patterns, which is beneficial to the accuracy of interaction behavior feature representation. Furthermore, the left singular vector quantifies the intensity of an object's participation in the object interaction pattern, while the right singular vector reveals the behavioral distribution patterns of time slices. This improves the computational efficiency of interaction behavior features while ensuring that the features effectively represent the object's interaction behavior within a time period, thus enhancing object processing performance.

[0081] S24, Perform object processing based on the interaction behavior characteristics of all objects within the first time period.

[0082] In some feasible implementations, when processing objects based on the interaction behavior characteristics of all objects within a first time period, for each object, it can be determined whether the object is in an active state at the first moment based on the interaction behavior characteristics of that object within the first time period.

[0083] Specifically, for each object, the row corresponding to the object in the target feature matrix can be determined, and the feature vector formed by the elements of that row can be determined as the interactive behavior feature of the object in the first time period.

[0084] After determining the interactive behavior characteristics of the object in the first time period, the interactive behavior characteristics of the object in the first time period can be processed by the state prediction model to obtain the predicted probability that the object is in an active state at the first moment.

[0085] When the predicted probability that an object is active at the first moment is greater than or equal to the probability threshold, it indicates that the object is likely to be active at the first moment, and thus it can be determined that the object is active at the first moment. When the predicted probability that an object is active at the first moment is less than the probability threshold, it indicates that the object is likely to be inactive at the first moment, and thus it can be determined that the object is inactive at the first moment.

[0086] The active state of any object at any time includes, but is not limited to, being online or interacting with other objects. No restrictions are imposed here.

[0087] The state prediction model is a pre-trained neural network model, which can be built based on Hidden Markov Model (HMM), Support Vector Machine (SVM) model, Recurrent Neural Network (RNN) or Long Short-Term Memory (LSTM) network, etc., without any restrictions.

[0088] The first moment is the moment following the end of the first time period, meaning that the active state of the object in the future can be predicted based on the interactive behavior characteristics of the object in the first time period.

[0089] In some feasible implementations, when processing objects based on the interaction behavior characteristics of all objects within a first time period, for each object, in response to the friend recommendation operation triggered at the first moment, friend recommendations are made to that object based on the interaction behavior characteristics of all objects within the first time period.

[0090] Wherein, the first moment is the moment following the end of the first time period.

[0091] The friend recommendation action can be either opening the friend list or discovering friends; the specific action can be determined based on the actual application scenario and is not limited here.

[0092] For example, see Figure 5 , Figure 5 This is a schematic diagram illustrating a scenario of recommending friends, as demonstrated in an embodiment of this application. Figure 5 As shown, for any given object, when the user controlling the object clicks the "Discover" touch button for finding new friends at the first moment, a friend recommendation list can be displayed to the object, and the recommended friends will be displayed in the friend recommendation list.

[0093] Specifically, the similarity of the behavior of the object to that of each other can be determined based on the interactive behavior characteristics of the object during the first time period and the interactive behavior characteristics of each of the other objects during the first time period.

[0094] If Figure 4 The set of nodes corresponding to all time slices in the time slice is represented as {T1, T2, ..., T...} M Let M be the number of time slices within the first time period. Then, the behavioral similarity between any two objects can be expressed as:

[0095]

[0096] For example, for Figure 3 There is:

[0097] s(v1,v2)=w(v1,u1)·w(v2,u1)+w(v1,u2)·w(v2,u2)=1×2+0×1=2.

[0098] In other words, the behavioral similarity between all objects can be represented as a matrix. That is, S = WW T =(U∑)(U∑) T .

[0099] Approximately, the behavioral similarity between all objects can be represented as S = ZZ. T =(U d ∑ d (U) d ∑ d ) T .

[0100] Based on this, the behavioral similarity between all objects can be determined directly from the target feature matrix Z.

[0101] Furthermore, based on the behavioral similarity among all objects, at least one recommended friend can be identified for each object, and each recommended friend can be displayed through a recommendation list.

[0102] For each object, objects whose behavior is more similar to that object than a similarity threshold can be identified as recommended friends of that object, and / or, the top preset number of objects with the highest behavior similarity to that object can be identified as recommended friends of that object. The specific details can be determined based on the actual application scenario requirements and are not limited here.

[0103] Utilizing interactive behavior features for friend matching or recommendation helps expand the user's social network and increase user engagement. Predicting user activity levels based on these features allows for real-time differentiation and association between active and inactive users, improving processing efficiency.

[0104] In some feasible implementations, after determining the target feature matrix based on the first dataset corresponding to the first time period, the target feature matrix can be updated after a second time period following the first time period, since time is constantly progressing, to obtain the updated target feature matrix.

[0105] For details on how to update the target feature matrix, please refer to [link / reference]. Figure 6 , Figure 6 This is a flowchart illustrating the target feature matrix update method provided in this application embodiment, as shown below. Figure 6As shown, the target feature matrix update method provided in this application embodiment may include the following steps:

[0106] S61, in response to the second time period following the first time period, obtain the second dataset.

[0107] In some feasible implementations, the feature vector composed of each row of matrix elements in the updated target feature matrix represents the interactive behavior features of an object within the target time period.

[0108] The target time period consists of a first time period and a second time period.

[0109] Specifically, when updating the target feature matrix, a third dataset can be obtained. The third dataset includes the interaction data of all objects in the object set within the target time period. Then, based on the third dataset, a new target feature matrix can be determined in the same way as the target feature matrix determined based on the first dataset, thereby updating the original target feature matrix.

[0110] Optionally, in updating the target feature matrix, a second dataset can also be obtained, which includes interaction data of all objects in the object set during a second time period.

[0111] S62, determine the second weight matrix based on the second dataset.

[0112] In some feasible implementations, the matrix element in the i-th row and u-th column of the second weight matrix represents the number of objects that interact with object i in the u-th time slice of the second time period, where u is a positive integer less than or equal to K, and K is the number of time slices in the second time period.

[0113] S63, Update the target feature matrix according to the second weight matrix.

[0114] In some feasible implementations, the target feature matrix can be updated according to the second weight, thereby performing object processing based on the interactive behavior characteristics of all the objects within the target time period. For example, determining the active state of each object at a second time point, and making friend recommendations for each object at the second time point, etc.

[0115] The second moment can be the moment following the end of the second time period.

[0116] In some feasible implementations, when updating the target feature matrix based on the second weight matrix, the second weight matrix can be decomposed into three matrices: the second singular value matrix, the second left singular matrix, and the second right singular matrix.

[0117] The second left singular matrix is ​​used to characterize the participation intensity of each object in different object interaction modes within the second time period. The singular values ​​on the diagonal of the second singular value matrix represent the importance of different interaction modes within the second time period. The second right singular matrix represents the participation degree of each time slice within the second time period in different object interaction modes.

[0118] Further, determine the first d largest singular values ​​in the second singular value matrix, and the second singular value vectors corresponding to the first d largest singular values ​​in the second left singular matrix and the second singular value matrix; where d is a positive integer and is less than or equal to the minimum value of K and N.

[0119] Furthermore, the first feature matrix can be determined based on the first d largest singular values ​​in the second singular value matrix and the second singular value vectors corresponding to the first d largest singular values ​​in the second left singular matrix and the second singular value matrix, and then the target feature matrix can be updated based on the first feature matrix.

[0120] In some feasible implementations, when updating the target feature matrix based on the first feature matrix, the first feature matrix and the original target feature matrix can be concatenated to obtain the updated target feature matrix.

[0121] In this context, the vector formed by each row of the first feature matrix represents the interactive behavior characteristics of different objects during the second time period.

[0122] For example, if the second weight matrix is ​​represented as W′, then the singular value decomposition of the second weight matrix yields:

[0123]

[0124] Where, ∑′ d U′ represents the first d largest singular values ​​of the second singular value matrix ∑. d Let d represent the d left singular vectors corresponding to the second left singular matrix, then the first characteristic matrix can be determined as U′. d ∑′ d .

[0125] Based on this, the updated target feature matrix is: [U d Σ d ,U′ d Σ′ d ], [] represents matrix concatenation operations.

[0126] In some feasible implementations, when updating the target feature matrix based on the first feature matrix, the target feature matrix and the first feature matrix can be concatenated to obtain a first recombined matrix. For example, the target feature matrix is ​​U. d Σ dThe first characteristic matrix is ​​U′ d Σ′ d Then the first recombination matrix is ​​[U d Σ d U′ d Σ′ d ].

[0127] Furthermore, the first recombined matrix is ​​decomposed into an orthogonal triangular matrix to obtain a first orthogonal matrix and a first upper triangular matrix.

[0128] The first orthogonal matrix is ​​the orthogonalized representation of the cross-time period (cross the first time period and the second time period) interaction behavior features of each object, and the first upper triangular matrix is ​​used to represent the orthogonalized representation coefficients of the cross-time period interaction behavior features of each object.

[0129] For example, the first orthogonal matrix obtained by performing orthogonal triangular decomposition on the first recombined matrix is ​​Q1, and the first upper triangular matrix obtained is R1.

[0130] Furthermore, a diagonal matrix is ​​constructed based on the first right singular matrix obtained by performing singular value decomposition on the first weight matrix and the second right singular matrix obtained by performing singular value decomposition on the second weight matrix, and the diagonal matrix is ​​then subjected to orthogonal triangular decomposition to obtain the second upper triangular matrix.

[0131] The first right singular matrix represents the participation level of each time slice in different object interaction modes in the first time period, the second right singular matrix represents the participation level of each time slice in different object interaction modes in the second time period, and the second upper triangular matrix is ​​used to represent the orthogonalized expression coefficients of object interaction modes across time periods.

[0132] For example, the first right singular matrix obtained by performing singular value decomposition on the first weight matrix is: The second right singular matrix obtained by performing singular value decomposition on the second weight matrix is: The constructed diagonal matrix is The second upper triangular matrix, R2, is obtained by performing orthogonal triangular decomposition on the diagonal matrix.

[0133] Furthermore, the updated target feature matrix can be determined based on the first orthogonal matrix, the first upper triangular matrix, and the second upper triangular matrix.

[0134] Specifically, the coefficient adjustment matrix can be determined based on the first orthogonal matrix, the first upper triangular matrix, and the second upper triangular matrix. The number of rows and columns in the coefficient adjustment matrix is ​​the same as the number of rows and columns in the target feature matrix. Based on this, each element in the coefficient adjustment matrix corresponds to a matrix element at the same position in the target feature matrix. For each position, the value of the matrix element at that position in the target feature matrix is ​​multiplied by the value of the matrix element at that position in the coefficient adjustment matrix, ultimately yielding the updated target feature matrix.

[0135] Alternatively, the second recombination matrix can be determined based on the first and second upper triangular matrices. Specifically, the second recombination matrix can be determined by multiplying the transpose of the second upper triangular matrix with the first upper triangular matrix.

[0136] The second recombination matrix is ​​used to represent the coupling strength between object interaction patterns and time slices across time periods, revealing the synergistic correlation between object behavioral interaction characteristics and object interaction patterns.

[0137] For example, if the first upper triangular matrix is ​​R1 and the second upper triangular matrix is ​​R2, then the second recombination matrix is ​​R1R2. T .

[0138] Furthermore, singular value decomposition is performed on the second recombined matrix to determine the third singular value matrix and the third left singular matrix obtained by the decomposition. Then, the first d largest singular values ​​in the third singular value matrix and the third singular value vectors corresponding to the first d largest singular values ​​in the third left singular matrix and the third singular value matrix are determined.

[0139] Based on this, the updated target feature matrix can be determined according to the first orthogonal matrix, the third singular value vector, and the third singular value vector.

[0140] For example, performing singular value decomposition on the second recombined matrix determines the resulting third singular value matrix and third left singular matrix, and then determining the first d largest singular values ​​within the third singular value matrix. The third singular value vector corresponding to the first d largest singular values ​​in the third left singular matrix and the third singular value matrix is: The first orthogonal matrix is ​​Q1, and the third singular value vector is And the third singular value vector is U R d In this case, the updated target feature matrix is

[0141] In other words, if the first weight matrix is ​​represented as W and the second weight matrix is ​​represented as W′, then concatenating the first weight matrix and the second weight matrix yields W″=[W,W′], where [] represents the matrix concatenation operation.

[0142] After performing singular value decomposition on W′, we get:

[0143]

[0144] Based on the orthogonal triangular decomposition algorithm, we can obtain:

[0145] Q1,R1←QR([U d Σ d U′ d Σ′ d ])

[0146]

[0147] Where Q1 and Q2 are orthogonal matrices, and R1 and R2 are upper triangular matrices.

[0148] because:

[0149]

[0150] therefore:

[0151]

[0152] Therefore, for the matrix Singular value decomposition yields:

[0153]

[0154] therefore:

[0155]

[0156] in,

[0157] in,

[0158] If we perform singular value decomposition on W″, we can determine the third singular value matrix and the third left singular matrix obtained from the decomposition, and determine the first d largest singular values ​​Σ″ within the third singular value matrix. d And the third singular value vector U″ corresponding to the first d largest singular values ​​in the third left singular matrix and the third singular value matrix. d And then according to U″ d and Σ″ d Determine the updated target feature matrix, because therefore That is, the updated target feature matrix is Therefore, the updated target feature matrix can be determined based on the aforementioned first orthogonal matrix, third singular value vector, and third singular value vector.

[0159] Among them, the third left singular matrix is ​​used to characterize the participation intensity of each object in different object interaction modes within the target time period, the singular values ​​on the diagonal of the third singular value matrix represent the importance of different interaction modes within the target time period, and the third right singular matrix represents the participation degree of each time slice in different object interaction modes within the third time period.

[0160] It should be noted that the object interaction modes involved in the embodiments of this application can be concentrated interaction during peak periods, sparse interaction during off-peak periods, or object interaction caused by sudden events (such as game tasks), etc., and are not limited here. For example, the object interaction modes in the embodiments of this application may include off-peak interaction on weekdays, peak interaction on weekends, concentrated interaction caused by task release, etc.

[0161] In the above embodiments, by updating the target feature matrix on the new time period data (second weight matrix), the computational latency can be significantly reduced, the computational complexity can be reduced, and the recalculation of the entire data can be avoided. This is beneficial to improving the update efficiency of the target feature matrix and improving the object processing efficiency.

[0162] Meanwhile, based on the QR decomposition method, not only can the update efficiency of the target feature matrix be improved, but the interpretability and discriminative power of the feature space can also be enhanced. It can not only reveal the correlation of cross-time interaction patterns and enhance the ability to characterize complex behaviors, but also improve the feature expression effect of the updated target behavior feature matrix.

[0163] It should be noted that the collection and processing of interactive data in the embodiments of this application should strictly comply with the requirements of relevant national laws and regulations in actual application, obtain the informed consent or separate consent of the subject corresponding to the interactive data, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject. It is understood that in the specific implementation of this application, when the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of relevant countries and regions.

[0164] See Figure 7 , Figure 7 This is a schematic diagram of the structure of the object processing apparatus provided in the embodiments of this application. The object processing apparatus provided in the embodiments of this application includes:

[0165] Data acquisition module 71 is used to acquire a first dataset, which includes the interaction data of all objects in the object collection during a first time period;

[0166] The feature processing module 72 is used to determine a first weight matrix based on the first dataset mentioned above. The matrix element in the i-th row and j-th column of the first weight matrix represents the number of objects that interact with object i in the j-th time slice of the first time period. i is a positive integer less than or equal to N, j is a positive integer less than or equal to M, N is the number of objects in the object set mentioned above, and M is the number of time slices in the first time period mentioned above.

[0167] The feature processing module 72 is used to perform singular value decomposition on the first weight matrix and determine the interaction behavior features of each object in the first time period based on the decomposition result.

[0168] The object processing module 73 is used to process objects based on the interaction behavior characteristics of all the above objects during the first time period.

[0169] In some feasible implementations, when the feature processing module 72 performs singular value decomposition on the first weight matrix and determines the interaction behavior features of each object within the first time period based on the decomposition results, it is used for:

[0170] Perform singular value decomposition on the first weight matrix to determine the first singular value matrix and the first left singular matrix obtained from the decomposition.

[0171] Determine the first d largest singular values ​​in the first singular value matrix and the first singular value vector in the first left singular matrix corresponding to the first d largest singular values ​​in the first singular value matrix; where d is a positive integer and is less than or equal to the minimum value of M and N;

[0172] The target feature matrix is ​​determined based on the first d largest singular values ​​in the first singular value matrix and the first singular value vector; wherein, the feature vector composed of the matrix elements in each row of the target feature matrix respectively represents the interactive behavior features of the object in the first time period.

[0173] In some feasible implementations, when the object processing module 73 performs object processing based on the interaction behavior characteristics of all the objects during the first time period, it is used to:

[0174] For each of the above objects, determine whether the object is in an active state at the first moment based on the interactive behavior characteristics of the object during the first time period.

[0175] For each of the above objects, in response to the friend recommendation operation triggered at the first moment mentioned above, friend recommendations are made to that object based on the interaction behavior characteristics of all the above objects during the first time period mentioned above.

[0176] The first moment mentioned above refers to the moment following the end of the first time period mentioned above.

[0177] In some feasible implementations, for each of the above-mentioned objects, when the object processing module 73 determines whether the object is in an active state at the first moment based on the object's interactive behavior characteristics during the first time period, it is used to:

[0178] The interaction behavior characteristics of the object during the first time period are processed by the state prediction model to obtain the predicted probability that the object is in an active state at the first moment.

[0179] In response to the predicted probability being greater than or equal to the probability threshold, it is determined that the object is in an active state at the first time point; in response to the predicted probability being less than the probability threshold, it is determined that the object is in an inactive state at the first time point.

[0180] In some feasible implementations, for each of the aforementioned objects, when the object processing module 73 recommends friends to that object based on the interaction behavior characteristics of all the aforementioned objects during the aforementioned first time period, it is used to:

[0181] Based on the interactive behavior characteristics of the object during the first time period and the interactive behavior characteristics of each of the other objects during the first time period, the similarity of the object's behavior with each of the other objects is determined.

[0182] At least one recommended friend is determined based on the similarity of the above behaviors, and each of the above recommended friends is displayed in a recommendation list;

[0183] Among them, at least one of the recommended friends includes at least one of the following:

[0184] Objects whose behavior is more similar to this object than a similarity threshold;

[0185] The first preset number of objects whose behavior is most similar to that of the given object.

[0186] In some feasible implementations, the data acquisition module 71 described above is further used for:

[0187] In response to a second time period following the first time period, a second dataset is obtained, which includes interaction data of all objects in the object set during the second time period.

[0188] The aforementioned feature processing module 72 is also used for:

[0189] The second weight matrix is ​​determined based on the second dataset mentioned above. The matrix element in the i-th row and u-th column of the second weight matrix represents the number of objects that interact with object i in the u-th time slice of the second time period. u is a positive integer less than or equal to K, and K is the number of time slices in the second time period.

[0190] The target feature matrix is ​​updated according to the second weight matrix. The feature vector composed of each row of matrix elements in the updated target feature matrix represents the interactive behavior features of the object in the target time period. The target time period consists of the first time period and the second time period.

[0191] The object processing module 73 described above is also used for:

[0192] Object processing is performed based on the interaction behavior characteristics of all the aforementioned objects within the aforementioned target time period.

[0193] In some feasible implementations, when the feature processing module 72 updates the target feature matrix according to the second weight matrix, it is used to:

[0194] Perform singular value decomposition on the above second weight matrix to determine the second singular value matrix and the second left singular matrix obtained from the decomposition;

[0195] Determine the first d largest singular values ​​in the second singular value matrix and the second singular value vector in the second left singular matrix corresponding to the first d largest singular values ​​in the second singular value matrix; where d is a positive integer and is less than or equal to the minimum value of K and N;

[0196] The first feature matrix is ​​determined based on the first d largest singular values ​​in the second singular value matrix and the second singular value vector. The vector formed by the matrix elements of each row of the first feature matrix represents the interactive behavior characteristics of different objects in the second time period.

[0197] The target feature matrix is ​​updated based on the first feature matrix.

[0198] In some feasible implementations, when the feature processing module 72 updates the target feature matrix according to the first feature matrix, it is used to:

[0199] The target feature matrix and the first feature matrix are concatenated to obtain the first recombination matrix. The first recombination matrix is ​​then decomposed into an orthogonal triangular matrix to obtain a first orthogonal matrix and a first upper triangular matrix. The first orthogonal matrix is ​​an orthogonalized representation of the cross-time interaction behavior features of each of the above objects, and the first upper triangular matrix is ​​used to represent the orthogonalized representation coefficients of the cross-time interaction behavior features of each of the above objects.

[0200] A diagonal matrix is ​​constructed based on the first right singular matrix obtained by performing singular value decomposition on the first weight matrix and the second right singular matrix obtained by performing singular value decomposition on the second weight matrix. The diagonal matrix is ​​then subjected to orthogonal triangular decomposition to obtain a second upper triangular matrix. The first right singular matrix represents the participation degree of each time slice in the first time period in different object interaction modes, and the second right singular matrix represents the participation degree of each time slice in the second time period in different object interaction modes. The second upper triangular matrix is ​​used to represent the orthogonalized expression coefficients of cross-time period object interaction modes.

[0201] Based on the first orthogonal matrix, the first upper triangular matrix, and the second upper triangular matrix, the updated target feature matrix is ​​determined.

[0202] In some feasible implementations, when the feature processing module 72 determines the updated target feature matrix based on the first orthogonal matrix, the first upper triangular matrix, and the second upper triangular matrix, it is used to:

[0203] The second recombination matrix is ​​determined based on the first upper triangular matrix and the second upper triangular matrix. The second recombination matrix is ​​used to represent the coupling strength between the object interaction mode and the time slice across time periods.

[0204] Perform singular value decomposition on the second recombined matrix to determine the third singular value matrix and the third left singular matrix obtained from the decomposition.

[0205] Determine the first d largest singular values ​​in the aforementioned third singular value matrix, and the third singular value vector in the aforementioned third left singular matrix that corresponds to the first d largest singular values ​​in the aforementioned third singular value matrix;

[0206] Based on the first orthogonal matrix, the first d largest singular values ​​in the third singular value matrix, and the third singular value vector, the updated target feature matrix is ​​determined.

[0207] In specific implementation, the aforementioned object processing device can execute the above-described functions through its built-in functional modules. Figure 2 and / or Figure 6 The implementation methods provided for each step are detailed in the above-mentioned implementation methods, and will not be repeated here.

[0208] See Figure 8 , Figure 8 This is a schematic diagram of the structure of the electronic device provided in an embodiment of this application. For example... Figure 8 As shown, the electronic device 800 in this embodiment may include: a processor 801, a network interface 804, and a memory 805. Furthermore, the electronic device 800 may also include: an object interface 803, and at least one communication bus 802. The communication bus 802 is used to implement communication between these components. The object interface 803 may include a display screen and a keyboard; optionally, the object interface 803 may also include a standard wired interface or a wireless interface. The network interface 804 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 805 may be a high-speed RAM or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory 805 may also be at least one storage device located remotely from the aforementioned processor 801. Figure 8 As shown, the memory 805, which is a computer-readable storage medium, may include an operating system, a network communication module, an object interface module, and a device control application.

[0209] exist Figure 8 In the illustrated electronic device 800, the network interface 804 provides network communication functionality; the object interface 803 is primarily used to provide an input interface for objects; and the processor 801 can be used to call the device control application program stored in the memory 805 to achieve:

[0210] Obtain the first dataset, which includes the interaction data of all objects in the object collection during the first time period;

[0211] The first weight matrix is ​​determined based on the first dataset mentioned above. The matrix element in the i-th row and j-th column of the first weight matrix represents the number of objects that interact with object i in the j-th time slice of the first time period. i is a positive integer less than or equal to N, j is a positive integer less than or equal to M, N is the number of objects in the object set mentioned above, and M is the number of time slices in the first time period mentioned above.

[0212] Singular value decomposition is performed on the first weight matrix, and the interaction behavior characteristics of each of the above objects in the first time period are determined based on the decomposition results.

[0213] Object processing is performed based on the interaction behavior characteristics of all the aforementioned objects during the first time period.

[0214] In some feasible implementations, when the processor 801 performs singular value decomposition on the first weight matrix and determines the interaction behavior characteristics of each object within the first time period based on the decomposition result, it is used for:

[0215] Perform singular value decomposition on the first weight matrix to determine the first singular value matrix and the first left singular matrix obtained from the decomposition.

[0216] Determine the first d largest singular values ​​in the first singular value matrix and the first singular value vector in the first left singular matrix corresponding to the first d largest singular values ​​in the first singular value matrix; where d is a positive integer and is less than or equal to the minimum value of M and N;

[0217] The target feature matrix is ​​determined based on the first d largest singular values ​​in the first singular value matrix and the first singular value vector; wherein, the feature vector composed of the matrix elements in each row of the target feature matrix respectively represents the interactive behavior features of the object in the first time period.

[0218] In some feasible implementations, when the processor 801 performs object processing based on the interaction behavior characteristics of all the objects during the first time period, it is used to:

[0219] For each of the above objects, determine whether the object is in an active state at the first moment based on the interactive behavior characteristics of the object during the first time period.

[0220] For each of the above objects, in response to the friend recommendation operation triggered at the first moment mentioned above, friend recommendations are made to that object based on the interaction behavior characteristics of all the above objects during the first time period mentioned above.

[0221] The first moment mentioned above refers to the moment following the end of the first time period mentioned above.

[0222] In some feasible implementations, for each of the above-mentioned objects, when the processor 801 determines whether the object is active at the first moment based on the object's interactive behavior characteristics during the first time period, it is used to:

[0223] The interaction behavior characteristics of the object during the first time period are processed by the state prediction model to obtain the predicted probability that the object is in an active state at the first moment.

[0224] In response to the predicted probability being greater than or equal to the probability threshold, it is determined that the object is in an active state at the first time point; in response to the predicted probability being less than the probability threshold, it is determined that the object is in an inactive state at the first time point.

[0225] In some feasible implementations, when the processor 801 recommends friends to each of the aforementioned objects based on the interaction behavior characteristics of all the aforementioned objects during the aforementioned first time period, it is used to:

[0226] Based on the interactive behavior characteristics of the object during the first time period and the interactive behavior characteristics of each of the other objects during the first time period, the similarity of the object's behavior with each of the other objects is determined.

[0227] At least one recommended friend is determined based on the similarity of the above behaviors, and each of the above recommended friends is displayed in a recommendation list;

[0228] Among them, at least one of the recommended friends includes at least one of the following:

[0229] Objects whose behavior is more similar to this object than a similarity threshold;

[0230] The first preset number of objects whose behavior is most similar to that of the given object.

[0231] In some feasible implementations, the processor 801 is further configured to:

[0232] In response to a second time period following the first time period, a second dataset is obtained, which includes interaction data of all objects in the object set during the second time period.

[0233] The second weight matrix is ​​determined based on the second dataset mentioned above. The matrix element in the i-th row and u-th column of the second weight matrix represents the number of objects that interact with object i in the u-th time slice of the second time period. u is a positive integer less than or equal to K, and K is the number of time slices in the second time period.

[0234] The target feature matrix is ​​updated according to the second weight matrix. The feature vector composed of each row of matrix elements in the updated target feature matrix represents the interactive behavior features of the object in the target time period. The target time period consists of the first time period and the second time period.

[0235] Object processing is performed based on the interaction behavior characteristics of all the aforementioned objects within the aforementioned target time period.

[0236] In some feasible implementations, when the processor 801 updates the target feature matrix according to the second weight matrix, it is used to:

[0237] Perform singular value decomposition on the above second weight matrix to determine the second singular value matrix and the second left singular matrix obtained from the decomposition;

[0238] Determine the first d largest singular values ​​in the second singular value matrix and the second singular value vector in the second left singular matrix corresponding to the first d largest singular values ​​in the second singular value matrix; where d is a positive integer and is less than or equal to the minimum value of K and N;

[0239] The first feature matrix is ​​determined based on the first d largest singular values ​​in the second singular value matrix and the second singular value vector. The vector formed by the matrix elements of each row of the first feature matrix represents the interactive behavior characteristics of different objects in the second time period.

[0240] The target feature matrix is ​​updated based on the first feature matrix.

[0241] In some feasible implementations, when the processor 801 updates the target feature matrix according to the first feature matrix, it is used to:

[0242] The target feature matrix and the first feature matrix are concatenated to obtain the first recombination matrix. The first recombination matrix is ​​then decomposed into an orthogonal triangular matrix to obtain a first orthogonal matrix and a first upper triangular matrix. The first orthogonal matrix is ​​an orthogonalized representation of the cross-time interaction behavior features of each of the above objects, and the first upper triangular matrix is ​​used to represent the orthogonalized representation coefficients of the cross-time interaction behavior features of each of the above objects.

[0243] A diagonal matrix is ​​constructed based on the first right singular matrix obtained by performing singular value decomposition on the first weight matrix and the second right singular matrix obtained by performing singular value decomposition on the second weight matrix. The diagonal matrix is ​​then subjected to orthogonal triangular decomposition to obtain a second upper triangular matrix. The first right singular matrix represents the participation degree of each time slice in the first time period in different object interaction modes, and the second right singular matrix represents the participation degree of each time slice in the second time period in different object interaction modes. The second upper triangular matrix is ​​used to represent the orthogonalized expression coefficients of cross-time period object interaction modes.

[0244] Based on the first orthogonal matrix, the first upper triangular matrix, and the second upper triangular matrix, the updated target feature matrix is ​​determined.

[0245] In some feasible implementations, when the processor 801 determines the updated target feature matrix based on the first orthogonal matrix, the first upper triangular matrix, and the second upper triangular matrix, it is used to:

[0246] The second recombination matrix is ​​determined based on the first upper triangular matrix and the second upper triangular matrix. The second recombination matrix is ​​used to represent the coupling strength between the object interaction mode and the time slice across time periods.

[0247] Perform singular value decomposition on the second recombined matrix to determine the third singular value matrix and the third left singular matrix obtained from the decomposition.

[0248] Determine the first d largest singular values ​​in the aforementioned third singular value matrix, and the third singular value vector in the aforementioned third left singular matrix that corresponds to the first d largest singular values ​​in the aforementioned third singular value matrix;

[0249] Based on the first orthogonal matrix, the first d largest singular values ​​in the third singular value matrix, and the third singular value vector, the updated target feature matrix is ​​determined.

[0250] It should be understood that in some feasible implementations, the processor 801 described above may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

[0251] In specific implementation, the aforementioned electronic device 800 can perform the above-described actions through its built-in functional modules. Figure 2 and / or Figure 6 The implementation methods provided for each step are detailed in the above-mentioned implementation methods, and will not be repeated here.

[0252] This application also provides a computer-readable storage medium storing a computer program that is executed by a processor to implement... Figure 2 and / or Figure 6 The methods provided in each step are detailed in the implementation methods provided in the above steps, and will not be repeated here.

[0253] The aforementioned computer-readable storage medium can be an internal storage unit of the object processing apparatus or electronic device provided in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device. The aforementioned computer-readable storage medium can also include magnetic disks, optical disks, read-only memory (ROM), or random access memory (RAM), etc. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of the electronic device. The computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0254] This application provides a computer program product, which includes a computer program that is executed by a processor. Figure 2 and / or Figure 6 The methods provided for each step in the process.

[0255] The terms "first," "second," etc., used in the claims, description, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or electronic device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or electronic devices. References to "embodiment" herein mean that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The presentation of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. The term "and / or" as used in this application's description and appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0256] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Those skilled in the art can implement the described functions using different methods for each specific application, but such implementations should not be considered beyond the scope of this application.

[0257] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. An object processing method, characterized in that, The method includes: Obtain a first dataset, which includes interaction data of all objects in the object collection during a first time period; A first weight matrix is ​​determined based on the first dataset, wherein the matrix element in the i-th row and j-th column of the first weight matrix represents the number of objects that interact with object i in the j-th time slice of the first time period, i is a positive integer less than or equal to N, j is a positive integer less than or equal to M, N is the number of objects in the object set, and M is the number of time slices in the first time period. Singular value decomposition is performed on the first weight matrix, and the interaction behavior characteristics of each object in the first time period are determined based on the decomposition results. Object processing is performed based on the interaction behavior characteristics of all the objects during the first time period.

2. The method according to claim 1, characterized in that, The step of performing singular value decomposition on the first weight matrix and determining the interaction behavior characteristics of each object within the first time period based on the decomposition results includes: Perform singular value decomposition on the first weight matrix to determine the first singular value matrix and the first left singular matrix obtained from the decomposition; Determine the first d largest singular values ​​in the first singular value matrix, and the first singular value vector in the first left singular matrix corresponding to the first d largest singular values ​​in the first singular value matrix; where d is a positive integer, and is less than or equal to the minimum value of M and N; The target feature matrix is ​​determined based on the first d largest singular values ​​in the first singular value matrix and the first singular value vector; wherein, the feature vector composed of the matrix elements in each row of the target feature matrix respectively represents the interactive behavior features of the object in the first time period.

3. The method according to claim 1 or 2, characterized in that, The object processing based on the interaction behavior characteristics of all the objects during the first time period includes at least one of the following: For each object, it is determined whether the object is in an active state at the first moment based on the object's interactive behavior characteristics during the first time period; For each of the objects, in response to the friend recommendation operation triggered at the first moment, friend recommendations are made to the object based on the interaction behavior characteristics of all the objects in the first time period; Wherein, the first moment is the moment following the end of the first time period.

4. The method according to claim 3, characterized in that, For each object, determining whether the object is active at the first moment based on its interactive behavior characteristics during the first time period includes: The interactive behavior characteristics of the object during the first time period are processed by the state prediction model to obtain the predicted probability that the object is in an active state at the first moment. In response to the predicted probability being greater than or equal to a probability threshold, it is determined that the object is in an active state at the first time moment; in response to the predicted probability being less than the probability threshold, it is determined that the object is in an inactive state at the first time moment.

5. The method according to claim 3, characterized in that, For each of the objects, the step of recommending friends to that object based on the interaction behavior characteristics of all the objects within the first time period includes: Based on the interaction behavior characteristics of the object during the first time period and the interaction behavior characteristics of each of the other objects during the first time period, the similarity of the object's behavior with each of the other objects is determined. At least one recommended friend is determined based on the similarity of the behaviors described, and each of the recommended friends is displayed in a recommendation list. The at least one recommended friend includes at least one of the following: Objects whose behavior is more similar to this object than a similarity threshold; The first preset number of objects whose behavior is most similar to that of the given object.

6. The method according to claim 2, characterized in that, The method further includes: In response to a second time period following the first time period, a second dataset is obtained, the second dataset including interaction data of all objects in the object set during the second time period; A second weight matrix is ​​determined based on the second dataset, wherein the matrix element in the i-th row and u-th column of the second weight matrix represents the number of objects that interact with object i in the u-th time slice of the second time period, u is a positive integer less than or equal to K, and K is the number of time slices in the second time period. The target feature matrix is ​​updated according to the second weight matrix, and the feature vector composed of each row of matrix elements in the updated target feature matrix respectively represents the interactive behavior features of the object in the target time period; wherein, the target time period is composed of the first time period and the second time period; Object processing is performed based on the interactive behavior characteristics of all the objects within the target time period.

7. The method according to claim 6, characterized in that, The step of updating the target feature matrix according to the second weight matrix includes: Perform singular value decomposition on the second weight matrix to determine the second singular value matrix and the second left singular matrix obtained from the decomposition; Determine the first d largest singular values ​​in the second singular value matrix, and the second singular value vector in the second left singular matrix corresponding to the first d largest singular values ​​in the second singular value matrix; where d is a positive integer and is less than or equal to the minimum value of K and N; The first feature matrix is ​​determined based on the first d largest singular values ​​in the second singular value matrix and the second singular value vector. The vector formed by the matrix elements of each row of the first feature matrix represents the interactive behavior characteristics of different objects in the second time period. The target feature matrix is ​​updated based on the first feature matrix.

8. The method according to claim 7, characterized in that, The step of updating the target feature matrix based on the first feature matrix includes: The target feature matrix and the first feature matrix are concatenated to obtain a first recombination matrix. The first recombination matrix is ​​then decomposed into an orthogonal triangular matrix to obtain a first orthogonal matrix and a first upper triangular matrix. The first orthogonal matrix is ​​an orthogonalized representation of the cross-time interaction behavior features of each object, and the first upper triangular matrix is ​​used to represent the orthogonalized representation coefficients of the cross-time interaction behavior features of each object. A diagonal matrix is ​​constructed based on the first right singular matrix obtained by performing singular value decomposition on the first weight matrix and the second right singular matrix obtained by performing singular value decomposition on the second weight matrix. The diagonal matrix is ​​then subjected to orthogonal triangular decomposition to obtain a second upper triangular matrix. The first right singular matrix represents the participation degree of each time slice in the first time period in different object interaction modes, and the second right singular matrix represents the participation degree of each time slice in the second time period in different object interaction modes. The second upper triangular matrix is ​​used to represent the orthogonalized expression coefficients of object interaction modes across time periods. The updated target feature matrix is ​​determined based on the first orthogonal matrix, the first upper triangular matrix, and the second upper triangular matrix.

9. The method according to claim 8, characterized in that, The step of determining the updated target feature matrix based on the first orthogonal matrix, the first upper triangular matrix, and the second upper triangular matrix includes: The second recombination matrix is ​​determined based on the first upper triangular matrix and the second upper triangular matrix. The second recombination matrix is ​​used to represent the coupling strength between the object interaction pattern and the time slice across time periods. Perform singular value decomposition on the second recombined matrix to determine the third singular value matrix and the third left singular matrix obtained from the decomposition; Determine the first d largest singular values ​​in the third singular value matrix, and the third singular value vector in the third left singular matrix that corresponds to the first d largest singular values ​​in the third singular value matrix; The updated target feature matrix is ​​determined based on the first orthogonal matrix, the first d largest singular values ​​in the third singular value matrix, and the third singular value vector.

10. An object processing apparatus, characterized in that, The device includes: The data acquisition module is used to acquire a first dataset, which includes the interaction data of all objects in the object collection during a first time period. The feature processing module is used to determine a first weight matrix based on the first dataset, wherein the matrix element in the i-th row and j-th column of the first weight matrix represents the number of objects that interact with object i in the j-th time slice of the first time period, i is a positive integer less than or equal to N, j is a positive integer less than or equal to M, N is the number of objects in the object set, and M is the number of time slices in the first time period. The feature processing module is used to perform singular value decomposition on the first weight matrix and determine the interaction behavior features of each object in the first time period based on the decomposition result. The object processing module is used to process objects based on the interaction behavior characteristics of all the objects during the first time period.

11. An electronic device, characterized in that, It includes a processor and a memory, which are interconnected; The memory is used to store computer programs; The processor is configured to execute the method according to any one of claims 1 to 9 when the computer program is invoked.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is executed by a processor to implement the method of any one of claims 1 to 9.

13. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images