CN115878907B - Social network forwarding behavior prediction method and device based on user dependency relationship - Google Patents
Social network forwarding behavior prediction method and device based on user dependency relationship Download PDFInfo
- Publication number
- CN115878907B CN115878907B CN202211697432.3A CN202211697432A CN115878907B CN 115878907 B CN115878907 B CN 115878907B CN 202211697432 A CN202211697432 A CN 202211697432A CN 115878907 B CN115878907 B CN 115878907B
- Authority
- CN
- China
- Prior art keywords
- user
- vector
- forwarding
- text
- text content
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Machine Translation (AREA)
Abstract
The embodiment of the invention discloses a social network forwarding behavior prediction method and device based on user dependency, wherein the method comprises the following steps: determining a user group according to the user social relationship to obtain a first embedded vector of each user in the user group; according to the forwarding process and the first embedded vector of the user, calculating to obtain a second embedded vector of the user containing the context dependent vector; performing cross-layer connection and normalization processing according to the sequence of forwarding text contents by the user based on a self-attention mechanism according to the second embedding vector of the user and a preset direction mask, and obtaining the sequence embedding of the forwarding user in the forwarding process; determining a forwarding text embedding vector of the text content according to the text content; according to the forwarding text embedding vector of the text content and the forwarding user sequence embedding, the attention result of the forwarding user sequence embedding to the text content is obtained through calculation, and the prediction probability of the forwarding text content of the user is obtained according to the attention result.
Description
Technical Field
The embodiment of the invention relates to the technical field of social network application, in particular to a social network forwarding behavior prediction method and device based on user dependency.
Background
With the rapid development and popularization of the internet and web2.0 technology, a large number of online social networks have emerged on the internet, which have become the main platform for people to share, spread and acquire information. The media can release the content on the platform, and a large number of users can accelerate the content spreading process through actions such as praise, forwarding and the like.
The research on content diffusion in the prior art mainly aims at understanding the content transmission mechanism, and has important significance for commercial popularization, social hotspot tracking, sensitive content monitoring and the like. By analyzing the characteristics of the content distribution, the distribution range of the content can be adjusted, the distribution user population can be influenced, or the rapid distribution of the content can be avoided.
At present, the prediction about content forwarding behavior mostly depends on an RNN (Recurrent Neural Network ) model to extract the characteristics of the observed forwarding process, but the RNN model considers that the next data of the sequence is influenced by the direct precursor nodes, the longer the distance is, the weaker the influence relationship is, and the processing mode is contrary to the content propagation process. And the existing models mostly consider that the content is transmitted only by means of the influence relationship among users, and the influence of the characteristics of the content, the friend relationship characteristics among users, the attribute characteristics of the users and the like on the content transmission is ignored.
Disclosure of Invention
In view of the foregoing, embodiments of the present invention are provided to provide a social network forwarding behavior prediction method and apparatus based on user dependency that overcome or at least partially solve the foregoing problems.
According to an aspect of the embodiment of the invention, there is provided a social network forwarding behavior prediction method based on user dependency, including:
determining a user group according to the user social relationship to obtain a first embedded vector of each user in the user group;
according to the forwarding process and the first embedded vector of the user, calculating to obtain a second embedded vector of the user containing the context dependent vector; performing cross-layer connection and normalization processing according to the sequence of forwarding text contents by the user based on a self-attention mechanism according to the second embedding vector of the user and a preset direction mask, and obtaining the sequence embedding of the forwarding user in the forwarding process;
determining a forwarding text embedding vector of the text content according to the text content;
according to the forwarding text embedding vector of the text content and the forwarding user sequence embedding, the attention result of the forwarding user sequence embedding to the text content is obtained through calculation, and the prediction probability of the forwarding text content of the user is obtained according to the attention result.
Optionally, determining the user group according to the user social relationship, and obtaining the first embedded vector of each user in the user group further includes:
acquiring a user social relationship, determining a user group according to the user social relationship, and constructing a user social relationship network; the user social relation network is constructed by a graph data structure and comprises a plurality of user nodes and neighbor nodes connected with the user nodes;
and inputting the vector of each user node to the graph neural network model aiming at the user group to obtain a first embedded vector of each user in the user group.
Optionally, for the user group, inputting the vector of each user node into the graph neural network model, and obtaining the first embedded vector of each user in the user group further includes:
inputting vectors of user nodes and vectors of neighbor nodes of the user nodes into a graph neural network model aiming at any user node in a user group, calculating influence of each neighbor node on the user nodes, and splicing a plurality of obtained attention results of the user nodes according to the influence on the basis of a multi-head attention mechanism; outputting an average value of a plurality of attention results as an embedded vector of the user node;
And splicing the preset random vector with the embedded vector of the user node to obtain a first embedded vector of the user.
Optionally, calculating a second embedded vector of the user including the context dependency vector according to the forwarding process and the first embedded vector of the user further includes:
calculating the influence of the preceding forwarding user on the following forwarding user according to the forwarding sequence of the forwarding process, and obtaining the context dependency vector of the following forwarding user according to the weighted calculation of the influence;
and fusing the first embedded vector of the user and the context dependency vector of the user to obtain a second embedded vector of the user containing the context dependency vector.
Optionally, determining the forwarding text embedding vector of the text content according to the text content further comprises:
determining a release time embedded vector, a text background and content embedded vector and a text age embedded vector of the text content according to the text content;
and splicing the release time embedded vector, the text background and the content embedded vector of the text content and the text age embedded vector to obtain a forwarding text embedded vector of the text content.
Optionally, determining the release time embedded vector, the text background and the content embedded vector of the text content according to the text content, and the text age embedded vector further comprises:
Determining the time characteristics of the release time of the text content according to a plurality of preset dimensions according to the release time of the text content; the plurality of preset dimensions includes daily, weekly, and/or holiday dimensions; the time characteristic is represented by single thermal coding;
determining time embedded vectors of a plurality of preset dimensions according to the time characteristics of the release time of the text content of the plurality of preset dimensions;
splicing the time embedded vectors with a plurality of preset dimensions to obtain a release time embedded vector of the text content;
obtaining a corresponding text content embedding vector according to the text content, and constructing a corresponding hot content embedding vector according to the hot content;
based on the attention mechanism, calculating the similarity between the hot content and the text content according to the text content embedding vector and the hot content embedding vector, and calculating to obtain a background feature vector of the text content according to the similarity;
calculating weights of the background feature vector and the text content embedding vector, and calculating to obtain the text background and the content embedding vector according to the weights; the weight is obtained based on a gating mechanism;
based on the time attention mechanism, determining the age characteristics of the text content, and obtaining the text age embedding vector according to the age characteristics.
Optionally, according to the forwarding text embedding vector and the forwarding user sequence embedding of the text content, calculating to obtain an attention result of the forwarding user sequence embedding on the text content, and according to the attention result, obtaining the prediction probability of the forwarding text content of the user further includes:
calculating influence on each user embedded by the forwarding user sequence according to the forwarding text embedding vector of the text content, and obtaining the attention result of the forwarding user sequence embedding on the text content according to the influence and the forwarding user sequence embedding;
and according to the attention result, obtaining the prediction probability of forwarding the text content to the next user by the user in the forwarding sequence.
Optionally, obtaining the predicted probability of the user forwarding the text content to the next user in the forwarding order further includes:
and calculating an attention result based on the multi-layer perceptron to obtain a prediction probability of forwarding the text content to a next user by the user in the forwarding sequence.
Optionally, the method further comprises:
predicting to obtain target forwarding probability of the user group, and optimizing the prediction model according to the target forwarding probability and the sample forwarding probability in the training set.
According to another aspect of the embodiment of the present invention, there is provided a social network forwarding behavior prediction apparatus based on user dependency, the apparatus including:
The first user vector module is suitable for determining a user group according to the social relationship of the users to obtain a first embedded vector of each user in the user group;
the second user vector module is suitable for calculating a second embedded vector of the user containing the context dependency vector according to the forwarding process and the first embedded vector of the user; performing cross-layer connection and normalization processing according to the sequence of forwarding text contents by the user based on a self-attention mechanism according to the second embedding vector of the user and a preset direction mask, and obtaining the sequence embedding of the forwarding user in the forwarding process;
the text vector module is suitable for determining a forwarding text embedding vector of the text content according to the text content;
and the prediction module is suitable for obtaining the attention result of the text content by the forwarding user sequence embedding according to the forwarding text embedding vector of the text content and the forwarding user sequence embedding, and obtaining the prediction probability of the user forwarding the text content according to the attention result.
According to yet another aspect of an embodiment of the present invention, there is provided a computing device including: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
The memory is used for storing at least one executable instruction, and the executable instruction enables the processor to execute the operation corresponding to the social network forwarding behavior prediction method based on the user dependency relationship.
According to still another aspect of the embodiments of the present invention, there is provided a computer storage medium having at least one executable instruction stored therein, where the executable instruction causes a processor to perform operations corresponding to the social network forwarding behavior prediction method based on a user dependency as described above.
According to the social network forwarding behavior prediction method and device based on the user dependency relationship, provided by the embodiment of the invention, the prediction model for text content propagation based on the user social network is established by analyzing the correlation between the behavior of forwarding text content in the user social network and various characteristics, so that the propagation range of the text content can be predicted, the propagation of the text content can be conveniently controlled, a platform is helped to recommend more proper text content, more accurate popularization is performed, reasonable control of the content is realized, and the development of the text platform is promoted. Furthermore, the invention uses a single context feature extraction mode to extract the influence of the past forwarding user on the current user node, and obtains the context dependency vector. And the influence among users is calculated by using a self-attention mechanism and a preset direction mask, so that the long-term dependence problem is relieved. According to the invention, the transducer is used as a sequence processing tool, and different from the existing deep learning model, the transducer module completely abandons the structure of the RNN and other sequence models, and adopts an attention mechanism to construct the model, so that any two nodes in the sequence can be directly associated, and the loss of characteristics is avoided. The invention greatly relieves the long-term dependence problem in the RNN model and avoids the problem that influence relation cannot be formed between non-adjacent nodes. Meanwhile, according to the invention, through the result of characteristic analysis of the forwarding behavior, a corresponding prediction model is determined aiming at the characteristics of the forwarding process, so that the prediction of the forwarding behavior of the user is completed.
The foregoing description is only an overview of the technical solutions of the embodiments of the present invention, and may be implemented according to the content of the specification, so that the technical means of the embodiments of the present invention can be more clearly understood, and the following specific implementation of the embodiments of the present invention will be more apparent.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
FIG. 1 illustrates a flow diagram of a social network forwarding behavior prediction method based on user dependencies, according to one embodiment of the invention;
FIG. 2 shows an overall architecture schematic of a predictive model;
FIG. 3 shows a schematic diagram of a social network forwarding behavior prediction apparatus based on user dependency according to one embodiment of the present invention;
FIG. 4 illustrates a schematic diagram of a computing device, according to one embodiment of the invention.
Detailed Description
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
FIG. 1 shows a flow chart of a social network forwarding behavior prediction method based on user dependency according to one embodiment of the present invention, as shown in FIG. 1, the method includes the steps of:
step S101, determining a user group according to the user social relationship, and obtaining a first embedded vector of each user in the user group.
In this embodiment, a prediction model is used to predict the probability of a user forwarding text content in a user group based on the dependency relationship between users. The user dependency relationship comprises a user social relationship, a friend making condition of a user and the like, and the influence of the user social relationship on forwarding behavior is analyzed based on the user social relationship. Specifically, the user social relationship may adopt a graph data structure manner of the friend relationship network in fig. 2 to construct a user social relationship network by using each user, i.e. a user group, included in the user social relationship. In the graph data structure, each node represents a single user, and the edges represent the characteristics of the relationships between the users. The user social relation network comprises a plurality of user nodes and neighbor nodes which are determined according to the connecting edges and are connected with the user nodes. For the user relationship, the relationship characteristics of the user in the friend network can be extracted through a GAT (Graph Attention Networks, graphic neural network) model, so as to obtain a first embedded vector of the user. GAT is part of the predictive model in this embodiment, which is responsible for extracting the embedded vectors of users based on the user's social relationships. When the GAT is used for obtaining the first embedded vector of the user, sample data of the social relationship network of the user is required to be collected in advance, the GAT is trained according to the sample data, and finally the embedded vector of each user in the user group can be obtained according to the trained GAT. The sample data comprises a user social relation network such as G (U, E) and a content set S based on a preset range, wherein U is each user in the user social relation network, E is a friend relation (i.e. a dependent relation) among different users, and single content S epsilon S, at time t by user U epsilon U 0 Emitting at a predetermined time range such as DeltaT obs In the time frame, the content s is propagated in the user social relationship network G through various online or offline approaches, and the propagation process can be recorded as a time sequence process C= [ (u) 1 ,t 1 ),(u 2 ,t 2 ),…(u i ,t i )…(u j ,t j ),...(u n ,t n )]Wherein, when i < j, t i ＜t j And t 1 ≥t 0 ，t n ＜t 0 +ΔT obs And acquiring the content propagated by a plurality of users in the user social relationship network within a preset time range.
The main structure of the GAT is composed of a graph annotation layer, and the input of the graph annotation layer is a group of node expression vectors h= [ h ] 1 ,h 2 ,h 3 ,...,h n ]I.e. the vector of each user node, the output is also a set of node expression vectors h '= [ h ]' 1 ,h′ 2 , h′ 3 ,...,h′ n ]. In this embodiment, the input is a vector of user nodes (such as various features of the user nodes themselves), and the input is an embedded vector of each user node in the user group. Aiming at the expression vector of each node in a certain user social relation network, the drawing attention layer acquires the expression vector of the neighbor node of the drawing attention layer through the structural characteristics of the drawing, and calculates the attention coefficient. Influence alpha of neighbor node j of node i on node i ij The calculation formula is as follows:
wherein W is GAT As a result of the parameters after the training,for each set of neighbor nodes of node i, α ij The influence of the neighbor node j on the node i is strong and weak. In practical application, F represents a feedforward neural network. LeakyReLU is the activation function.
For the middle layer of the GAT model, in order to make the result of the self-attention mechanism more stable, a multi-head attention mechanism can be adopted, namely, a plurality of sets of weights are adopted to learn the coefficients of the attention mechanism together, so that the fluctuation of a single coefficient due to the data set or the model initialization is avoided. The method can be specifically adopted as follows:
h i ＝[h ik ]
wherein σ () is an activation function, W k Is the coefficient of the kth attention mechanism, h ik Attention result corresponding to kth attention mechanism [ is the same as that of the kth attention mechanism ]]For splicing operation, [ h ] ik ]Attention results from multiple heads, such as k attention results, are stitched as current results.
For the last layer of the GAT model, since the last layer is directly used for outputting, the results of multiple attention mechanisms can be processed in an average manner, so that the user embedded vector based on the user social relationship network is obtained, and here, the results of connecting the attention mechanisms are not adopted only in the last layer.
Further, in view of the imperfection of the social relationship network of the user, the graph embedding process may introduce errors and other problems, and the obtained user embedded vector is usually in error with the attribute of the user, so as to eliminate the errors, the embodiment adopts another set of vectors to model the attribute of the user which is not observed, and the vectors represent the inherent attribute of the user which is irrelevant to the historical behavior and are not usually relevant to the social relationship network of the user. For each user U e U, a preset random vector v is used u ∈R 1×d The universality of the embedded vector of the user is increased. Wherein R is 1×d For the 1*d dimension real space, a random vector v is preset u The situation may be set up without limitation here. The first embedded vector of the user is determined by the user embedded vector obtained by embedding the user social relation network and a preset random vector, and finally the first embedded vector e of each user in the user group is obtained u The first embedded vector of the user will be used for subsequent sequence processingEtc., in particular, e u ＝[h u ,v u ]I.e. the user presets the random vector v u User embedding vector h obtained by embedding with user social relation network u Splicing to obtain a first embedded vector e of the user u U denotes the user.
Step S102, calculating a second embedded vector of the user containing the context dependent vector according to the forwarding process and the first embedded vector of the user; and performing cross-layer connection and normalization processing according to the sequence of forwarding text contents by the user based on a self-attention mechanism according to the second embedding vector of the user and a preset direction mask, and obtaining the sequence embedding of the forwarding user in the forwarding process.
The forwarding process, that is, a series of processes that text content is forwarded by multiple users, includes a forwarding process that, for example, forwards text content from user 1 to user 2 and from user 2 to user 4 … …, where the forwarding process involves a forwarding order among forwarding users, according to the forwarding process, a attention mechanism may be used to extract a dependency relationship of each user, and on the basis of retaining a first embedding vector of a user, a final embedding result of a user may also be adjusted by a change in direction order of forwarding text content.
Specifically, for the first embedded vector of each user, an attention mechanism is adopted to calculate the influence of each previous forwarding user with a previous forwarding sequence on the current user (the subsequent forwarding user), and the weighted summation is carried out based on each influence to obtain the vector of the current forwarding process, wherein the vector characterizes the state of the content when the current user forwards, fuses the information of each user in the historical forwarding process, and reflects the mutual influence among the users in the forwarding process. Specifically, the influence of the kth forwarding user on the jth forwarding user can be calculated by adopting the following formula:
wherein, the liquid crystal display device comprises a liquid crystal display device,for predicting model parameters, adjustments may be made according to a training process,<>representing an inner product operation e k A first embedded vector, e, for the kth user j The first embedded vector for the j-th user, k is less than j. e, e l The value range of l is 1 to j-1 for the first embedded vector of the first user. The attention mechanism adopted in the formula calculation has directionality, namely only the user forwarded before has influence on the user forwarded after, and otherwise, the attention mechanism does not have influence. Physically, only previously forwarded users will have an impact on later forwarded users, which is not consistent with the way the natural language processing model processes, because in text content, later appearing text content may document the former content. After calculation based on the attention mechanism, the attention weight may be normalized using a softmax function, etc.
For the user node in the forwarding process, the context dependency vector is the weighted summation of the first embedded vector and the influence of each user forwarded in advance, as shown in the following formula:
d j the context dependency vector of the jth user is calculated according to the first embedded vector of each user before j and the influence of each user on the jth user.
Since there is no dependency between the calculation results, the whole forwarding process can be calculated together in order to increase the calculation speed. Specifically, for a forwarding process with length l, a forwarding process matrix E εR is constructed d×l Where d represents the user embedding dimension, each row in matrix E represents the first embedding vector of the user at that location, and to ensure directionality, a preset direction mask M εR is introduced l×l When i<j is M i,j =0, otherwise M i,j The = - ≡i.e. only the previously forwarded user has an influence on the subsequently forwarded user, whereas no attention mechanism influence weight matrix is calculatedThe calculation method is as follows:
wherein A is E R l×l ，A ij The influence of the ith user on the j users is represented. In the calculation, if i>j is M ij The term goes to 0 after softmax operation, thus masking the effect of the post-forwarding user on the previously forwarded user. Using matrix a, matrix operations may be used to accelerate the computation of the context dependency vector for each stage.
D＝AE T
Wherein the ith row of matrix D represents the context dependency vector of the ith user, matrix A contains the influence among the users, A ij The influence of the ith user on the j users is represented. Each row in matrix E represents the first embedded vector of the user at that location, T being the transpose operation. The ith row corresponds to D in matrix D i 。
The characteristics of the user include a first embedded vector e of the user i And its context dependency vector d i However, considering that a single vector is required for subsequent processing, the two vectors can be fused in a manner of connection, addition, multiplication and the like, in the embodiment, a gating mechanism is adopted to select and fuse effective information in the two vectors, coefficients are calculated according to values of the two vectors, and the two vectors are fused in proportion according to the coefficients, so that a single vector of user characteristics is obtained.
u i ＝ i ⊙ i +(1- i )⊙ i
Wherein, the liquid crystal display device comprises a liquid crystal display device,are all prediction model parameters, and are adjusted according to the training process, g i Characterizing e i And d i Middle e i Importance of 1- i Represents e i And d i D in (d) i By g i Weighted average is carried out on the first embedded vector of the user and the context dependency vector, and the second embedded vector u of the user containing the context dependency vector is obtained through fusion i . As a result of the term multiplication of the vectors, the Hadmard function can be used.
The forwarding sequence embedding process mainly uses a transducer module. Since the present embodiment predicts whether or not a single user forwards, instead of predicting a forwarding sequence, only a transducer encoder section is used here, and the attention mechanism in the encoder can effectively extract the association relationship between user nodes. In each layer of the transducer, the input data contains a second embedded vector u of the user of the context dependent vector for each user i Arranging according to the order of forwarding text content of each user to obtainWhere l represents the sequence length, d up For the dimensions set according to the implementation. In order to better reflect the influence of the position characteristics of each user, the position characteristics are added in a preset position code (a fixed position code can be adopted) of each position, and the final result is mainly finished through a self-attention mechanism, and the specific formula is as follows:
wherein SA represents self-attention mechanism, f x Representing a softmax function, W Q 、W K 、W V All are prediction model parameters, which can be adjusted according to the training process, and M is a preset direction mask. For output SA (C o ) And performing cross-layer connection and normalization processing, wherein the following formula is shown:
e c ＝BatchNorm(SA(C o )+C o )
wherein, batchNorm is the normalization function. Here, a cross-layer connection structure such as ResNet can be used to avoid the occurrence of training Gradient vanishing problem. In order to achieve better processing effect, the above operation can be repeated for a plurality of times to obtain the forwarding user sequence embedded e of the final forwarding process c ＝{e c1 ,e c2 ,...,e cn }. Wherein e c1 ,e c2 ,...,e cn I.e. a sequence of users of n users arranged in the order in which the text content is forwarded by the n users. The forwarding user sequence embedding contains a plurality of user dependency relationships such as user social relationship, user forwarding relationship and the like, and the user dependency relationship has great influence on the user to perform social network forwarding.
Step S103, determining a forwarding text embedding vector of the text content according to the text content.
The forwarding behavior characteristics for text content can be summarized into multiple categories, such as publication source, publication time, text content, etc. The processing of various characteristics comprises the characteristic extraction of text content, the characteristic extraction of release time, the characteristic extraction of heat attenuation mode and the like.
The distribution time characteristic of the text content is a limited set, and no network relation structure exists, so that the processing mode is completely different from the text characteristic. In combination with daily work and rest time arrangement, modeling of release time according to periodic change of work and rest time is selected, for example, according to current working mode, time features in texts are determined by selecting a plurality of preset dimensions, for example, time feature embedding is selected according to dimensions of daily, weekly, holidays and the like.
The time of day characteristic refers to the time range that the content release time is on the current day, the week time characteristic refers to the range that the release time is on the current week, and the holiday time characteristic refers to whether the content release time is currently a holiday. Determining the release time of content based on the above time characteristics uses three one-hot codes, e.gExpressed, e pd For time of day feature e pw For the time of week feature e ph Is a holiday time feature. Wherein the above time characteristics are represented by single thermal codes, each single thermal code belonging to differentThe real space of the preset dimension is set according to the implementation situation, and the specific preset dimension is not limited herein. After the single thermal encoding of each temporal feature is obtained, the single thermal encoding is converted into a corresponding temporal embedding vector through a linear layer, as follows:
h pd ＝W pd e pd
h pw ＝W pw e pw
h ph ＝W ph e ph
wherein e pd ,e pw ,e ph One-time codes respectively representing time of day characteristics, time of week characteristics and holiday time characteristics,corresponding to the above three, respectively the respective time embedded vectors,to predict model parameters, adjustments may be made according to a training process. d, d d,w,h Representing the dimension of each time embedded vector, T d ,T w ,T h The specific values of the dimensions are set according to implementation conditions. The release time characteristic is commonly represented according to the time embedding vectors of the three, so as to obtain a release time embedding vector:
h pt ＝[h pd ,h pw ,h ph ]
Wherein [ among others ]]Representing vector concatenation, h pt I.e. the issue time embedded vector.
According to the text content, the text content issued by the user can be converted through a natural language processing model to obtain the text content characteristics, such as a large-scale pre-training model BERT, so as to obtain a corresponding embedded vector v of the text content p 。
Considering that when the text is released, the contents such as hot topics and the like influence the forwarding of the text contents, the text forwarding can be predicted by combining the hot topic contents. Specifically, a group of contents such as hot topics (hereinafter referred to as hot) are built by embedding the hot topic contentsContent) Hot content embedding vector V h ∈R 50×d The dimension of R is illustrated, and may be specifically set according to the implementation. Although the hot content can represent public opinion and background information of the current text content, the hot content is not combined with the specific text content, the forwarding condition of the specific text content cannot be determined, the hot content is numerous, and the hot content which has influence on the current text content is obtained through screening and is used as the background information of the current text content. In order to filter out the hot content which is irrelevant to the current text content in the hot content, an attention mechanism can be adopted to filter out the data of each hot topic, and the similarity with the currently processed text content is calculated. The calculation is specifically performed according to the following formula:
v pt ＝f p (W p v p +b p )
V ht ＝f h (W h V h +b h )
Wherein W is p ,W h ,b p ,b h For predicting model parameters, v can be adjusted according to training process pt 、V ht According to the respective function f p 、f h Embedding vector v into text content p Embedding vector V for hot content h Calculated N h I.e. the number of hot content, V hti Represents V ht I.e. the embedding result of the i-th hot content, beta i Representing the importance of the ith hot content to the text content, beta i And obtaining the similarity between the text content and the hot content by calculating. After obtaining the embedded vector of each hot content and the importance of the embedded vector to the text content, the background characteristic of the text content can be obtained through weighted summation.
Wherein v is bg A background feature vector representing the current text content may be fused using a gating unit in order to fuse features within the text with background features. And calculating and outputting a gating mechanism through the background feature vector and the text content embedded vector, namely outputting the proportion of the background feature vector in the final result by using a gating unit.
h hg ＝[v bg ,v p ]
α＝sigmoid(f(W hg h hg +b hg ))
Wherein h is hg Representing the input of gating information, W hg 、b hg For the prediction model parameters, f is an activation function, and α represents the weight occupied by the background feature vector. And according to the background feature weight, weighting and summing the background feature vector and the text content embedding vector, and calculating to obtain the text background and content embedding vector containing the background feature vector.
e bg ＝α*v bg +(1-α)*v p
Considering that the text content has a phenomenon of heat decay, namely, the heat is reduced with time, the self-attention mechanism cannot completely extract relevant features of the heat decay, so that the time-attention mechanism is also required to be used for modeling the text content, the heat decay features are extracted, and the heat of the text content can be defined as an 'age feature' of the text content, and the heat decay occurs with time.
The age characteristic of the text content is similar to the characteristic of the release time, but the age characteristic has no periodicity and no special value, so the text content can be obtained by using the age characteristic for embedding without considering periodicity, special value and the like through the following formula:
h dt ＝E dt e dt
wherein e dt ∈R D Is age-characteristic one-hot code, R D The number of age ranges, and D is the dimension set according to the implementation.In order to predict the parameters of the model,and adjusting according to the training process. />Is a text age embedding vector.
Splicing the text age embedded vector with the release time embedded vector of the text content, the text background and the content embedded vector to obtain e age ＝[h dt ,h pt,bg ]And obtaining the forwarding text embedded vector containing the text age, the release time, the text background characteristic and the text content characteristic.
The characteristics of the user dependency relationship and the characteristics of the forwarded text content play an important role in forwarding behavior prediction, and content information can have more accurate expression capability by combining the characteristics of the user dependency relationship and the characteristics of the forwarded text content, so that the accuracy of the prediction is improved.
Here, steps S101 to S102 are related calculations of the embedded vector of the user, step S103 is related calculations of the embedded vector of the text content, and the execution sequence of both may be the same, and the execution sequence of steps S101 to S102 and step S103 is not limited here.
Step S104, according to the forwarding text embedding vector and the forwarding user sequence embedding of the text content, attention results of the forwarding user sequence embedding on the text content are obtained through calculation, and the prediction probability of the forwarding text content of the user is obtained according to the attention results.
And calculating influence on the sequence embedding of the forwarding user by using the forwarding text embedding vector, wherein the calculated attention mechanism coefficient is as follows:
wherein w is f ∈R d For predicting model parameters, the parameters can be adjusted according to the training process, e age Plays a role of gating information, controls the characteristic intensity of entering the subsequent processing of the text, such as the text age, the release time, the text background characteristic, the text content characteristic and the like, and e ci Refers to the embedding of the ith user in the forwarding user sequence embedding,is the attention result of the ith user. As a result of the term multiplication of the vectors, the Hadmard function can be used. Calculating for forwarding user sequence embedding containing n users to obtain attention results of the whole forwarding user sequence embedding on forwarding text content, wherein the attention results are cascade embedded vectors:
according to the above attention results, for any time t i Concatenating attention results e cas Obtaining user u i For the next user u i+1 Forwarding text content s i Is shown in the following formula:
wherein, MLP (Multilayer Perceptron, multi-layer perceptron),the probability that each user in the user group forwards text content for the next forwarding user is represented, and the user group is the user group specified according to the user social relationship.
Optionally, the present embodiment further includes the following steps:
step S105, predicting the target forwarding probability of the user group, and optimizing the prediction model according to the target forwarding probability and the sample forwarding probability in the training set.
In this embodiment, when predicting the probability that a user will forward text content to the next user, the prediction model is trained based on sample data collected in advance. According to the training set in the collected sample data, the sample forwarding probability of all user groups in the training set can be obtained, and according to the probability of the sample forwarding user, each prediction model parameter of the prediction model in the training process can be adjusted. And calculating the forwarding probability of the whole user group according to the probability of predicting the text content to the next user by the user, wherein the forwarding probability is as follows:
The objective of optimizing the prediction model is to make L reach the maximum value as much as possible, and during optimization, all parameters in the prediction model are optimized by taking the result of L as a reference through an Adam optimizer and a back propagation algorithm, and finally the prediction model parameters obtained through training are all prediction model parameters when the prediction model is used for prediction.
The architecture diagram of the prediction model is shown in fig. 2, and the input information includes a friend relation network of the user, a forwarding process of text content, release content (i.e. text content) and hot content, and the prediction model includes different stages such as representation learning, user dependence, time sequence dependence, external dependence and prediction. In the representation learning stage, graph representation learning is carried out on a friend relation network (namely a user social network), for example, a first embedded vector of a user is obtained by using GAT extraction; for the forwarding process, a forwarding sequence (such as a forwarding process matrix formed by a text forwarding sequence by a user), time attenuation in the forwarding process and the like are obtained and embedded; for published content and hot content, text embedding (i.e., text content embedding vectors) and the like may be extracted. In the user dependence phase, the influence of each user in the forwarding process is weighted and summed through an attention mechanism to form a context dependence vector (i.e. for user sequence embedding in fig. 2), and the context dependence vector is fused with a first embedding vector of the user (i.e. background information fusion in fig. 2) to form a second embedding vector of the user. The second embedded vector of the user is used as input of a time sequence dependency stage, the time sequence dependency characteristics in the sequence are extracted through processing of a transducer encoder, and the sequence embedding of the forwarding user in the forwarding process is obtained, wherein the adding of position codes is included, and the transducer encoder outputs the sequence embedding of the forwarding user by adopting the modes of multi-head attention, weighted summation, normalization, feedforward network and the like. In the external dependence and prediction stage, feature vectors composed of the hot content and the feature of the released content, namely forwarding text embedding vectors, are weighted and summed to the influence of the obtained forwarding user sequence embedding to form cascade embedding vectors, namely attention results (namely external dependence attention in fig. 2), and finally the probability of forwarding text content for each user is output through an MLP model according to the attention results.
According to the social network forwarding behavior prediction method based on the user dependency relationship, provided by the embodiment of the invention, the prediction model for text content propagation based on the user social network is established by analyzing the correlation between the behavior of forwarding text content in the user social network and various characteristics, so that the propagation range of the text content can be predicted, the propagation of the text content can be conveniently controlled, a platform is helped to recommend more proper text content, more accurate popularization is realized, reasonable control of the content is realized, and the development of the text platform is promoted. Furthermore, the invention uses a single context feature extraction mode to extract the influence of the past forwarding user on the current user node, and obtains the context dependency vector. And the influence among users is calculated by using a self-attention mechanism and a preset direction mask, so that the long-term dependence problem is relieved. According to the invention, the transducer is used as a sequence processing tool, and different from the existing deep learning model, the transducer module completely abandons the structure of the RNN and other sequence models, and adopts an attention mechanism to construct the model, so that any two nodes in the sequence can be directly associated, and the loss of characteristics is avoided. The invention greatly relieves the long-term dependence problem in the RNN model and avoids the problem that influence relation cannot be formed between non-adjacent nodes. Meanwhile, according to the invention, through the result of characteristic analysis of the forwarding behavior, a corresponding prediction model is determined aiming at the characteristics of the forwarding process, so that the prediction of the forwarding behavior of the user is completed.
Fig. 3 is a schematic structural diagram of a social network forwarding behavior prediction device based on user dependency according to an embodiment of the present invention. As shown in fig. 3, the apparatus includes:
the first user vector module 310 is adapted to determine a user group according to the social relationship of the users, and obtain a first embedded vector of each user in the user group;
a second user vector module 320 adapted to calculate a second embedded vector of the user comprising the context dependency vector based on the forwarding process and the first embedded vector of the user; performing cross-layer connection and normalization processing according to the sequence of forwarding text contents by the user based on a self-attention mechanism according to the second embedding vector of the user and a preset direction mask, and obtaining the sequence embedding of the forwarding user in the forwarding process;
a text vector module 330 adapted to determine a forwarding text embedding vector for the text content based on the text content;
the prediction module 340 is adapted to calculate an attention result of the forwarding user sequence embedding on the text content according to the forwarding text embedding vector and the forwarding user sequence embedding of the text content, and obtain a prediction probability of the forwarding text content of the user according to the attention result.
Optionally, the first user vector module 310 is further adapted to:
Acquiring a user social relationship, determining a user group according to the user social relationship, and constructing a user social relationship network; the user social relation network is constructed by a graph data structure and comprises a plurality of user nodes and neighbor nodes connected with the user nodes;
and inputting the vector of each user node to the graph neural network model aiming at the user group to obtain a first embedded vector of each user in the user group.
Optionally, the first user vector module 310 is further adapted to:
inputting vectors of user nodes and vectors of neighbor nodes of the user nodes into a graph neural network model aiming at any user node in a user group, calculating influence of each neighbor node on the user nodes, and splicing a plurality of obtained attention results of the user nodes according to the influence on the basis of a multi-head attention mechanism; outputting an average value of a plurality of attention results as an embedded vector of the user node;
and splicing the preset random vector with the embedded vector of the user node to obtain a first embedded vector of the user.
Optionally, the second user vector module 320 is further adapted to:
calculating the influence of the preceding forwarding user on the following forwarding user according to the forwarding sequence of the forwarding process, and obtaining the context dependency vector of the following forwarding user according to the weighted calculation of the influence;
And fusing the first embedded vector of the user and the context dependency vector of the user to obtain a second embedded vector of the user containing the context dependency vector.
Optionally, the text vector module 330 is further adapted to:
determining a release time embedded vector, a text background and content embedded vector and a text age embedded vector of the text content according to the text content;
and splicing the release time embedded vector, the text background and the content embedded vector of the text content and the text age embedded vector to obtain a forwarding text embedded vector of the text content.
Optionally, the text vector module 330 is further adapted to:
determining the time characteristics of the release time of the text content according to a plurality of preset dimensions according to the release time of the text content; the plurality of preset dimensions includes daily, weekly, and/or holiday dimensions; the time characteristic is represented by single thermal coding;
determining time embedded vectors of a plurality of preset dimensions according to the time characteristics of the release time of the text content of the plurality of preset dimensions;
splicing the time embedded vectors with a plurality of preset dimensions to obtain a release time embedded vector of the text content;
obtaining a corresponding text content embedding vector according to the text content, and constructing a corresponding hot content embedding vector according to the hot content;
Based on the attention mechanism, calculating the similarity between the hot content and the text content according to the text content embedding vector and the hot content embedding vector, and calculating to obtain a background feature vector of the text content according to the similarity;
calculating weights of the background feature vector and the text content embedding vector, and calculating to obtain the text background and the content embedding vector according to the weights; the weight is obtained based on a gating mechanism;
based on the time attention mechanism, determining the age characteristics of the text content, and obtaining the text age embedding vector according to the age characteristics.
Optionally, the prediction module 340 is further adapted to:
calculating influence on each user embedded by the forwarding user sequence according to the forwarding text embedding vector of the text content, and obtaining the attention result of the forwarding user sequence embedding on the text content according to the influence and the forwarding user sequence embedding;
and according to the attention result, obtaining the prediction probability of forwarding the text content to the next user by the user in the forwarding sequence.
Optionally, the prediction module 340 is further adapted to:
and calculating an attention result based on the multi-layer perceptron to obtain a prediction probability of forwarding the text content to a next user by the user in the forwarding sequence.
Optionally, the apparatus further comprises:
The optimizing module 350 is adapted to predict a target forwarding probability of the user population, and perform optimizing processing on the prediction model according to the target forwarding probability and the sample forwarding probability in the training set.
The above descriptions of the modules refer to the corresponding descriptions in the method embodiments, and are not repeated herein.
The embodiment of the invention also provides a nonvolatile computer storage medium, and the computer storage medium stores at least one executable instruction, wherein the executable instruction can execute the social network forwarding behavior prediction method based on the user dependency relationship in any of the method embodiments.
FIG. 4 illustrates a schematic diagram of a computing device, according to an embodiment of the invention, the particular embodiment of which is not limiting of the particular implementation of the computing device.
As shown in fig. 4, the computing device may include: a processor 402, a communication interface (Communications Interface) 404, a memory 406, and a communication bus 408.
Wherein:
processor 402, communication interface 404, and memory 406 communicate with each other via communication bus 408.
A communication interface 404 for communicating with network elements of other devices, such as clients or other servers.
The processor 402 is configured to execute the program 410, and may specifically perform relevant steps in the social network forwarding behavior prediction method embodiment based on the user dependency.
In particular, program 410 may include program code including computer-operating instructions.
The processor 402 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included by the computing device may be the same type of processor, such as one or more CPUs; but may also be different types of processors such as one or more CPUs and one or more ASICs.
Memory 406 for storing programs 410. Memory 406 may comprise high-speed RAM memory or may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
Program 410 may be specifically operative to cause processor 402 to perform the social network forwarding behavior prediction method based on user dependencies in any of the method embodiments described above. Specific implementation of each step in the procedure 410 may refer to corresponding descriptions in the corresponding steps and units in the social network forwarding behavior prediction embodiment based on the user dependency, which are not described herein. It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the apparatus and modules described above may refer to corresponding procedure descriptions in the foregoing method embodiments, which are not repeated herein.
The algorithms or displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general-purpose systems may also be used with the teachings herein. The required structure for a construction of such a system is apparent from the description above. In addition, embodiments of the present invention are not directed to any particular programming language. It should be appreciated that the teachings of embodiments of the present invention described herein may be implemented in a variety of programming languages, and the above description of specific languages is provided for disclosure of preferred embodiments of the present invention.
In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in the above description of exemplary embodiments of the invention, various features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be construed as reflecting the intention that: i.e., an embodiment of the invention that is claimed, requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.
Those skilled in the art will appreciate that the modules in the apparatus of the embodiments may be adaptively changed and disposed in one or more apparatuses different from the embodiments. The modules or units or components of the embodiments may be combined into one module or unit or component and, furthermore, they may be divided into a plurality of sub-modules or sub-units or sub-components. Any combination of all features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be used in combination, except insofar as at least some of such features and/or processes or units are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.
Various component embodiments of the invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that some or all of the functionality of some or all of the components according to embodiments of the present invention may be implemented in practice using a microprocessor or Digital Signal Processor (DSP). Embodiments of the present invention may also be implemented as a device or apparatus program (e.g., a computer program and a computer program product) for performing a portion or all of the methods described herein. Such a program embodying the embodiments of the present invention may be stored on a computer readable medium, or may have the form of one or more signals. Such signals may be downloaded from an internet website, provided on a carrier signal, or provided in any other form.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Embodiments of the invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names. The steps in the above embodiments should not be construed as limiting the order of execution unless specifically stated.
Claims (9)
1. A social network forwarding behavior prediction method based on user dependency relationship is characterized by comprising the following steps:
determining a user group according to a user social relationship, and obtaining a first embedded vector of each user in the user group;
according to the forwarding process and the first embedded vector of the user, calculating to obtain a second embedded vector of the user containing the context dependent vector; performing cross-layer connection and normalization processing according to the sequence of forwarding text contents by the user based on a self-attention mechanism according to the second embedding vector of the user and a preset direction mask, and obtaining a forwarding user sequence embedding in a forwarding process;
determining a forwarding text embedding vector of the text content according to the text content;
calculating the influence of the forwarding text embedding vector of the text content on each user embedded by the forwarding user sequence, and obtaining the attention result of the forwarding user sequence embedding on the text content according to the influence and the forwarding user sequence embedding; according to the attention result, obtaining the prediction probability of the user forwarding the text content to the next user in the forwarding sequence;
the determining a forwarding text embedding vector of the text content according to the text content further comprises:
Determining a release time embedded vector, a text background and content embedded vector and a text age embedded vector of the text content according to the text content;
splicing the release time embedded vector, the text background and the content embedded vector of the text content and the text age embedded vector to obtain a forwarding text embedded vector of the text content; determining the time characteristics of the release time of the text content according to a plurality of preset dimensions according to the release time of the text content; the plurality of preset dimensions includes daily, weekly, and/or holiday dimensions; the time characteristic is represented by single thermal coding; determining time embedded vectors of a plurality of preset dimensions according to the time characteristics of the release time of the text content of the plurality of preset dimensions; splicing the time embedded vectors with the preset dimensions to obtain the release time embedded vector of the text content; obtaining a corresponding text content embedding vector according to the text content, and constructing a corresponding hot content embedding vector according to the hot content; based on an attention mechanism, calculating the similarity between the hot content and the text content according to the text content embedding vector and the hot content embedding vector, and calculating to obtain a background feature vector of the text content according to the similarity; calculating weights of the background feature vector and the text content embedding vector, and calculating to obtain a text background and the content embedding vector according to the weights; the weight is obtained based on a gating mechanism; based on a time attention mechanism, determining age characteristics of text content, and obtaining text age embedding vectors according to the age characteristics.
2. The method of claim 1, wherein determining a user population based on the user social relationship, obtaining a first embedded vector for each user in the user population, further comprises:
acquiring a user social relationship, determining a user group according to the user social relationship, and constructing a user social relationship network; the user social relation network is constructed by a graph data structure and comprises a plurality of user nodes and neighbor nodes connected with the user nodes;
and inputting the vector of each user node to the graph neural network model aiming at the user group to obtain a first embedded vector of each user in the user group.
3. The method of claim 2, wherein the inputting the vector of each user node into the graph neural network model for the user group to obtain the first embedded vector of each user in the user group further comprises:
inputting vectors of the user nodes and vectors of neighbor nodes of the user nodes into a graph neural network model aiming at any user node in the user group, calculating influence of each neighbor node on the user nodes, and splicing a plurality of obtained attention results of the user nodes according to the influence on the basis of a multi-head attention mechanism; outputting an average value of a plurality of attention results as an embedded vector of the user node;
And splicing the preset random vector with the embedded vector of the user node to obtain a first embedded vector of the user.
4. The method of claim 1, wherein calculating a second embedded vector for the user that includes the context dependency vector based on the forwarding process and the first embedded vector for the user further comprises:
calculating the influence of the preceding forwarding user on the following forwarding user according to the forwarding sequence of the forwarding process, and obtaining the context dependency vector of the following forwarding user according to the influence weighting calculation;
and fusing the first embedded vector of the user and the context dependency vector of the user to obtain a second embedded vector of the user containing the context dependency vector.
5. The method of claim 1, wherein the deriving a predicted probability of the user forwarding text content to the next user in the forwarding order based on the attention result further comprises:
and calculating the attention result based on the multi-layer perceptron to obtain the prediction probability of forwarding the text content to the next user by the user in the forwarding sequence.
6. The method according to any one of claims 1-5, further comprising:
Predicting to obtain target forwarding probability of the user group, and optimizing the prediction model according to the target forwarding probability and the sample forwarding probability in the training set.
7. A social network forwarding behavior prediction apparatus based on user dependency, the apparatus comprising:
the first user vector module is suitable for determining a user group according to the social relationship of the users to obtain a first embedded vector of each user in the user group;
the second user vector module is suitable for calculating a second embedded vector of the user containing the context dependent vector according to the forwarding process and the first embedded vector of the user; performing cross-layer connection and normalization processing according to the sequence of forwarding text contents by the user based on a self-attention mechanism according to the second embedding vector of the user and a preset direction mask, and obtaining a forwarding user sequence embedding in a forwarding process;
the text vector module is suitable for determining a forwarding text embedding vector of the text content according to the text content;
the prediction module is suitable for calculating the influence of the forwarding text embedding vector of the text content on each user embedded by the forwarding user sequence, and obtaining the attention result of the forwarding user sequence embedding on the text content according to the influence and the forwarding user sequence embedding; according to the attention result, obtaining the prediction probability of the user forwarding the text content to the next user in the forwarding sequence;
The text vector module is further adapted to: determining a release time embedded vector, a text background and content embedded vector and a text age embedded vector of the text content according to the text content;
splicing the release time embedded vector, the text background and the content embedded vector of the text content and the text age embedded vector to obtain a forwarding text embedded vector of the text content; determining the time characteristics of the release time of the text content according to a plurality of preset dimensions according to the release time of the text content; the plurality of preset dimensions includes daily, weekly, and/or holiday dimensions; the time characteristic is represented by single thermal coding; determining time embedded vectors of a plurality of preset dimensions according to the time characteristics of the release time of the text content of the plurality of preset dimensions; splicing the time embedded vectors with the preset dimensions to obtain the release time embedded vector of the text content; obtaining a corresponding text content embedding vector according to the text content, and constructing a corresponding hot content embedding vector according to the hot content; based on an attention mechanism, calculating the similarity between the hot content and the text content according to the text content embedding vector and the hot content embedding vector, and calculating to obtain a background feature vector of the text content according to the similarity; calculating weights of the background feature vector and the text content embedding vector, and calculating to obtain a text background and the content embedding vector according to the weights; the weight is obtained based on a gating mechanism; based on a time attention mechanism, determining age characteristics of text content, and obtaining text age embedding vectors according to the age characteristics.
8. A computing device, comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is configured to store at least one executable instruction, where the executable instruction causes the processor to perform operations corresponding to the social network forwarding behavior prediction method based on a user dependency as set forth in any one of claims 1 to 6.
9. A computer storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the user dependency based social network forwarding behavior prediction method according to any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211697432.3A CN115878907B (en) | 2022-12-28 | 2022-12-28 | Social network forwarding behavior prediction method and device based on user dependency relationship |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211697432.3A CN115878907B (en) | 2022-12-28 | 2022-12-28 | Social network forwarding behavior prediction method and device based on user dependency relationship |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115878907A CN115878907A (en) | 2023-03-31 |
| CN115878907B true CN115878907B (en) | 2023-08-11 |
Family
ID=85755723
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211697432.3A Active CN115878907B (en) | 2022-12-28 | 2022-12-28 | Social network forwarding behavior prediction method and device based on user dependency relationship |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115878907B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116091260B (en) * | 2023-04-07 | 2023-07-25 | 吕梁学院 | Cross-domain entity identity association method and system based on Hub-node |
| CN116842398A (en) * | 2023-06-27 | 2023-10-03 | 哈尔滨工业大学 | Topic perception information forwarding prediction method and system based on shielding self-attention network |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112418525A (en) * | 2020-11-24 | 2021-02-26 | 重庆邮电大学 | Method and device for predicting social topic group behaviors and computer storage medium |
| CN113139134A (en) * | 2021-04-26 | 2021-07-20 | 中国人民解放军国防科技大学 | Method and device for predicting popularity of user-generated content in social network |
| CN113807616A (en) * | 2021-10-22 | 2021-12-17 | 重庆理工大学 | Information diffusion prediction system based on space-time attention and heterogeneous graph convolution network |
| CN113901298A (en) * | 2021-10-29 | 2022-01-07 | 杭州师范大学钱江学院 | Social network forwarding behavior prediction method based on microblog text embedded learning |
| CN115168743A (en) * | 2022-07-27 | 2022-10-11 | 天津大学 | Crowd-sourcing body forwarding behavior prediction system based on heterogeneous graph neural network |
| CN115293437A (en) * | 2022-08-15 | 2022-11-04 | 重庆邮电大学 | Relationship prediction method based on social network platform dynamic interaction |
| CN115470991A (en) * | 2022-09-15 | 2022-12-13 | 重庆邮电大学 | Network rumor propagation prediction method based on user short-time emotion and evolutionary game |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11074500B2 (en) * | 2017-06-20 | 2021-07-27 | Battelle Memorial Institute | Prediction of social media postings as trusted news or as types of suspicious news |
-
2022
- 2022-12-28 CN CN202211697432.3A patent/CN115878907B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112418525A (en) * | 2020-11-24 | 2021-02-26 | 重庆邮电大学 | Method and device for predicting social topic group behaviors and computer storage medium |
| CN113139134A (en) * | 2021-04-26 | 2021-07-20 | 中国人民解放军国防科技大学 | Method and device for predicting popularity of user-generated content in social network |
| CN113807616A (en) * | 2021-10-22 | 2021-12-17 | 重庆理工大学 | Information diffusion prediction system based on space-time attention and heterogeneous graph convolution network |
| CN113901298A (en) * | 2021-10-29 | 2022-01-07 | 杭州师范大学钱江学院 | Social network forwarding behavior prediction method based on microblog text embedded learning |
| CN115168743A (en) * | 2022-07-27 | 2022-10-11 | 天津大学 | Crowd-sourcing body forwarding behavior prediction system based on heterogeneous graph neural network |
| CN115293437A (en) * | 2022-08-15 | 2022-11-04 | 重庆邮电大学 | Relationship prediction method based on social network platform dynamic interaction |
| CN115470991A (en) * | 2022-09-15 | 2022-12-13 | 重庆邮电大学 | Network rumor propagation prediction method based on user short-time emotion and evolutionary game |
Non-Patent Citations (1)
| Title |
|---|
| 基于分层注意力的信息级联预测模型;张志扬 等;计算机科学(06);第201-209页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115878907A (en) | 2023-03-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115878907B (en) | Social network forwarding behavior prediction method and device based on user dependency relationship | |
| WO2021047593A1 (en) | Method for training recommendation model, and method and apparatus for predicting selection probability | |
| CN110781321B (en) | Multimedia content recommendation method and device | |
| US11381651B2 (en) | Interpretable user modeling from unstructured user data | |
| US11551026B2 (en) | Dynamic reconfiguration training computer architecture | |
| CN112053004A (en) | Method and apparatus for time series prediction | |
| CN112541124B (en) | Method, apparatus, device, medium and program product for generating a multitasking model | |
| CN116415654A (en) | Data processing method and related equipment | |
| CN110287461A (en) | Text conversion method, device and storage medium | |
| WO2020236483A1 (en) | Cross-lingual information retrieval and information extraction | |
| CN113536144B (en) | Social network information propagation scale prediction method and device | |
| CN112990444B (en) | Hybrid neural network training method, system, equipment and storage medium | |
| CN110110233A (en) | Information processing method, device, medium and calculating equipment | |
| CN111881972B (en) | Black-out user identification method and device, server and storage medium | |
| CN113065013A (en) | Image annotation model training and image annotation method, system, device and medium | |
| CN116684330A (en) | Traffic prediction method, device, equipment and storage medium based on artificial intelligence | |
| Kamuni et al. | Enhancing End-to-End Multi-Task Dialogue Systems: A Study on Intrinsic Motivation Reinforcement Learning Algorithms for Improved Training and Adaptability | |
| US11604962B2 (en) | Method and system for training a machine learning system using context injection | |
| CN116993185A (en) | Time sequence prediction method, device, equipment and storage medium | |
| Kassawat et al. | Incorporating joint embeddings into goal-oriented dialogues with multi-task learning | |
| CN116401522A (en) | Financial service dynamic recommendation method and device | |
| CN115836288A (en) | Method and apparatus for generating training data | |
| CN113259163B (en) | Web service quality prediction method and system based on network topology perception | |
| CN118043802A (en) | Recommendation model training method and device | |
| CN112581185B (en) | Advertisement user gender and age estimation method, system, computer and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |