Topic propagation method based on time-domain psychology

Abstract

The invention belongs to the technical field of topic propagation, and particularly relates to a topic propagation method based on time-domain psychology, comprising: acquiring topic propagation data including user information, marketing topic information, a user relationship network and community information; extracting related attributes of the topic propagation data; inputting the related attributes into the topic propagation model to obtain a topic propagation prediction result; performing topic propagation according to a topic propagation prediction result; according to the method, the uncertainty of psychological states of users in different time domains and the action relation of adjacent nodes are analyzed, so that a measurement standard is provided for state conversion of the users; according to the cross-community marketing topic propagation influence and the conversion problem thereof, the influence of the marketing topic in a cross-community space and an information propagation rule are fully analyzed, so that the cross-community marketing topic propagation rule is deeply and comprehensively analyzed.

Description

Technical Field

[0001] The present invention belongs to the technical field of topic propagation, and specifically relates to social platform marketing topic data analysis, and in particular to a cross-community propagation model for marketing topics on an e-commerce platform. Background Art

[0002] With the increasing popularity and development of the Internet, people can quickly and easily obtain information from various Internet social platforms. These social platforms have also become an important platform for the dissemination of various information on the Internet. With the popularity of social networks and the rapid growth in the number of users, more and more companies are aware of the importance of marketing on social networks. Taking social networking platforms such as Facebook, Instagram, and WeChat as examples, these social platforms have become important channels for companies to interact with consumers, promote products, and establish brand image. By using social networks for marketing, companies can interact with users in real time, accurately locate target audiences, and expand brand influence through sharing and word-of-mouth communication.

[0003] At present, there are two main directions for the research on cross-community marketing topic propagation: one is to establish an information propagation model based on the classic epidemic model SIR, and to predict the propagation trend of cross-community marketing topics on the Internet by improving the SIR model. The other is to summarize the characteristics of the actual propagation trend of topics on the network, and predict the forwarding behavior of users based on deep learning algorithms and combined with propagation characteristics.

[0004] In response to the problem of user cognitive drive measurement, Pu et al. designed an adaptive knowledge accumulation framework based on the cognitive process in the human brain. This framework can continuously learn in multiple fields and even generalize to new and unseen fields. Yang et al. proposed a scalable hierarchical heterogeneous multi-core architecture for learning and analyzing various human cognitive activities, including behavioral selection, context-dependent learning, etc. From the above studies, it can be found that scholars mostly analyze and model the relationship between cognition and behavior, and there are relatively few studies on the extraction and analysis of the multidimensional and complex causes of cognitive drive. In response to the problem of uncertainty in the user's temporal psychological state, Hao et al. proposed a multi-scale sentence embedding method CAMSE, which encodes sentences into embedding tensors based on contextual self-attention and multi-scale technology. Huang et al. integrated emotional intelligence and attention mechanisms to improve the LSTM network, and at the same time used the emotional semantic information hidden in the text topic and context to achieve the representation and classification of emotions. Obviously, when studying psychological states, the above literature tends to classify and grade emotions, while the behavioral impact of user psychological changes in continuous time needs further research. In view of the multi-level nature of cross-community marketing topic propagation, Yu et al. proposed that identifying key nodes is a process of diffusion from the center to the outside, and regarded it as a regression problem. Wen et al. used Jensen-Shannon divergence and log-sigmoid transformation functions to construct community networks to show the relationship between communities. In summary, scholars have conducted in-depth research on the multi-layer coupling and mutual influence of cross-community topic propagation, but the interconnection and transformation between communities still deserve further study. Summary of the invention

[0005] In summary, the current research methods of topic propagation have the following problems: 1. How to effectively measure the multidimensional and complex causes driven by users' own cognition has become a challenge. 2. In the process of cross-community marketing topic propagation, in addition to the factors of the topic itself, the user's temporal psychological state has an important impact on the attention paid to the topic and its propagation in related communities; however, analyzing the specific impact of this psychological state on topic propagation has become a challenge. 3. In the process of marketing topic propagation in different communities, the trends and characteristics of topic propagation in different communities are different, but communities need to be associated and converted. How to combine the differences between communities and the state conversion between community users has become a challenge.

[0006] In order to solve the problems existing in the above prior art, the present invention proposes a topic propagation method based on time domain psychology, which comprises:

[0007] Step A: Obtain user historical topic propagation data, which includes user information, user relationship network, marketing topic information, and social platform community network information; pre-process the user historical topic propagation data;

[0008] Step B: extract relevant attributes from the preprocessed user topic propagation data;

[0009] Step C: constructing a topic propagation model according to relevant attributes, the topic propagation model including: a user-driven cognitive measurement mechanism, a time-domain psychological state description mechanism, and a cross-community marketing information propagation model; the user-driven cognitive measurement mechanism is used to obtain the proportional coefficient of the user cognitive description mechanism parameters; the time-domain psychological state description mechanism is used to obtain the high-dimensional vector representation of the user; the cross-community marketing information propagation model predicts the topic propagation results according to the high-dimensional vector representation of the user and the proportional coefficient of the user cognitive description mechanism parameters;

[0010] Step D: Predict the propagation trend of the topic to be propagated based on the topic propagation model.

[0011] Beneficial effects of the present invention:

[0012] Starting from the causes of user cognitive drive, the present invention uses potential causal algorithms to quantify the multidimensional and complex causes of cognitive drive; the present invention uses word vector embedding and LSTM to describe the user's temporal psychology and more accurately capture the changes in the user's psychological state; finally, based on the correlation between communities, the present invention combines user cognitive drive and temporal psychology to propose a cross-community marketing topic propagation model, thereby deeply and comprehensively analyzing the laws of cross-community marketing topic propagation. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a flow chart of the present invention;

[0014] Figure 2 A schematic diagram of the potential causal algorithm of the present invention for measuring user cognitive driving factors;

[0015] Figure 3 The word vector and LSTM of the present invention measure the user's temporal psychology;

[0016] Figure 4 It is the state transition diagram of the CC-SIR model of the present invention;

[0017] Figure 5 Schematic diagram of the CC-SIR model of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] A topic propagation method based on time domain psychology, the method comprising:

[0020] Step A: Obtain user historical topic propagation data, which includes user information, user relationship network, marketing topic information, and social platform community network information; pre-process the user historical topic propagation data;

[0021] Step B: extract relevant attributes from the preprocessed user topic propagation data;

[0022] Step C: constructing a topic propagation model according to relevant attributes, the topic propagation model including: a user-driven cognitive measurement mechanism, a time-domain psychological state description mechanism, and a cross-community marketing information propagation model; the user-driven cognitive measurement mechanism is used to obtain the proportional coefficient of the user cognitive description mechanism parameters; the time-domain psychological state description mechanism is used to obtain the high-dimensional vector representation of the user; the cross-community marketing information propagation model predicts the topic propagation results according to the high-dimensional vector representation of the user and the proportional coefficient of the user cognitive description mechanism parameters;

[0023] Step D: Predict the propagation trend of the topic to be propagated based on the topic propagation model.

[0024] like Figure 1 The figure shows the overall process of the present invention, indicating that the input of the present invention is the basic information of users on the social platform, the historical behavior data of users and the basic information of marketing topics, and the output after the prediction model is the prediction result of whether the user will place an order for the marketing topic product.

[0025] In this embodiment, the specific process of the present invention includes the following steps:

[0026] S1: Get data. The way to get data is to download open source data sets online. What needs to be obtained here is the interaction of marketing topics during their life cycle and the historical behavior data of users who have interacted with the marketing topics. The marketing topic participation situation needs to obtain the time of marketing topic interaction, basic information of the interacting users, and basic information of the marketing topic; the historical behavior of the marketing topic interactor includes the marketing topic information that the user has interacted with in history. And pre-process the data.

[0027] S2: Extract relevant attributes. Extract relevant attributes based on the basic information and historical behavior of active platform users in the life cycle of a marketing topic, the basic information and interaction information of the marketing topic, etc.

[0028] S3: Build a model. First, use the potential causal algorithm to quantify the multidimensional and complex causes of cognitive drive. Then, in order to more accurately capture the changes in user psychological states, use word vector embedding and LSTM to depict user temporal psychology. Finally, based on the correlation between communities, a cross-community marketing topic propagation model is proposed in combination with user cognitive drive and temporal psychology.

[0029] In this embodiment, obtaining the data source includes:

[0030] S11: Obtaining raw data: Raw data can be obtained through social network public APIs or by directly downloading existing data sources.

[0031] S12: Simple data cleaning. The raw data usually obtained is unstructured and cannot be used directly for data analysis. Most unstructured data can be structured through simple data cleaning. For example, duplicate data can be deleted, invalid information can be cleaned, etc.

[0032] In this embodiment, the specific process of extracting relevant attributes includes: in a social platform, a user's judgment on a cross-community marketing topic is affected by many factors, such as: the user's activity level, the influence of neighboring users on themselves, the popularity of the topic when the user is exposed to the topic, the degree of association between the topic diffusion and the community, etc.

[0033] S21, user driving force Dri(ui);

[0034] After a user posts a topic, his fans browse, comment on and forward it, causing the topic to gradually spread in the social network. Num[read(ui)], Num[comt(ui)] and Num[ret(ui)] represent the average number of views, average number of comments and average number of forwarding of all microblogs published by user ui some time before the topic was launched. Since the number of views is usually much larger than the other two and the influence of the number of views on the spread of the topic is not significant, a weakening coefficient μ is added before the average number of views. The driving force of the user in the process of information dissemination can be defined as:

[0035]

[0036] S22, User activity Act(u i );

[0037] User activity indicates the i The recent level of activity in topic dissemination. It is generally believed that the more active the user, the more likely they are to participate in topic dissemination and the greater their driving force in topic dissemination.

[0038] User activity is defined as:

[0039] Act(u i )=η×Num[orig(u i )]+Num[retw(u i )]

[0040] Among them, Num[orig(u i )] and Num[retw(u i )] represent user u i The number of original microblogs and forwarded microblogs published in the period before the topic spreads. Generally, users’ original microblogs are far less than forwarded microblogs, so we define a y[0,1] as a weakening coefficient to reduce the impact of the number of original microblogs on the calculation of user activity.

[0041] S23, topic popularity P(t);

[0042] Topic popularity refers to the traffic size of a topic, that is, the popularity of a topic. Each relatively independent topic will experience a process in which its popularity gradually increases and then gradually fades. The traffic of a topic will slowly decline over time and slowly reach a low level, indicating that the topic has left the public's view. This decline process is similar to the half-life of physical elements. This article introduces the half-life function (1 / 2) to describe the decline process of topic traffic. The definition of message popularity is as follows:

[0043]

[0044] Among them, t and t0 represent the current time and the starting time of the topic propagation heat decay process respectively, and T is the average propagation period of the topic.

[0045] S24, topic diffusion Diff(topic);

[0046] Topic diffusion refers to the degree of spread of a topic in social networks, that is, the popularity and spread of the topic. The level of topic diffusion reflects the popularity and spread of a topic in social networks, and is an important indicator to measure the popularity and spread of a topic in social networks. A high topic diffusion indicates that the topic is well-known and spreadable in social networks, and the content and activities of the topic are likely to attract more attention and participation. A low topic diffusion indicates that the topic is not well-known and spreadable in social networks, and the content and activities of the topic are not likely to attract more attention and participation.

[0047] Diff(topic)=Num[brow(topic)]+Num[comm(topic)+Num[like(topic)]+Num[C(topic)]

[0048] S25, community relevance I(A,B);

[0049] Community relevance indicates the degree of relevance between communities. The degree of relevance can be expressed by community similarity. The more similar the community, the more similar the community topics are. Otherwise, the opposite is true. Measuring community relevance can help study the dissemination rules of similar community marketing topics.

[0050]

[0051] Among them, A i , B j It represents the keyword sets extracted from different communities, and I represents the mutual information.

[0052] In this embodiment, building a model includes: building a user-driven cognitive measurement mechanism, building a time-domain psychological description mechanism, and building a cross-community marketing topic propagation model. In the first stage, the algorithm design of the user-driven cognitive measurement mechanism is constructed. Starting from the multidimensionality and complexity of the user cognitive driving factors, considering the advantages of the potential causal algorithm in analyzing multi-factor and potential factor models, a marketing topic user cognitive driving measurement method based on the potential causal algorithm is proposed. First, the user's personal information, social relations and community information are vectorized, and different community users participating in topic T are extracted. A propensity score model is established based on user information and user topic acceptance, and the causal effect is calculated based on the model results of the control group and the experimental group. The proportional coefficient of the parameters of the user cognitive description mechanism is obtained by normalization. In the second stage, starting from the uncertainty of the psychological state of users in different time domains, a user time-domain psychological state description mechanism based on word vector embedding and LSTM is constructed by using word vector embedding and LSTM to represent the effectiveness of high-dimensional space and temporal space information. First, the user's published information is one-hot encoded as the input of the word vector model, and then the model is trained, and finally the user's high-dimensional vector representation is obtained according to the model parameter matrix. In the third stage, a cross-community marketing information dissemination model CC-SIR is proposed. First, mutual information is used to measure the correlation between communities. Then, the cross-community model state transition equation is defined by combining user cognitive drive and time domain psychology. Finally, the cross-community marketing topic dissemination model CC-SIR is obtained.

[0053] The specific steps are:

[0054] S31: Build a user-driven cognitive measurement mechanism.

[0055] First, a propensity score model is established to vectorize the user's personal information, social relationships, and community information. The propensity score model is used to obtain the user's causal effect, and the cognitive driving force coefficient is obtained based on the average causal effect, such as Figure 2As shown. First, the information vector is used as the input X of the model. In the data screening process, the data is divided into a treatment group (factor applied) and a control group (factor not applied) according to whether a specific factor is applied. Determine the marketing topic T that all users browse together, and set the target variable Y according to whether the user accepts the content of the topic: 1 for acceptance and 0 for non-acceptance.

[0056] In order to evaluate the user's acceptance of the topic, usually, if the user's subsequent browsing of related topics increases significantly after browsing the topic, the marketing topic is considered successful and the target variable is set to 1. The calculation method of acceptance Acc is as follows:

[0057]

[0058] Among them, num before Indicates the number of views on related topics before browsing topic T, num after It represents the number of views on related topics after browsing topic T, and Acc represents the final acceptance, which is the target variable Y.

[0059] First, the input data is normalized, assuming that Xi represents the i-th eigenvalue in the feature matrix. The normalized value is Xnorm, which is defined as follows:

[0060] Xnorm=(Xi-Xmin) / (Xmax-Xmin)

[0061] Among them, Xmin and Xmax represent the minimum and maximum values ​​of the feature among all users respectively. The feature vector of each user is a row of data, that is, a 1×n vector, where n is the number of features.

[0062] Use the Sigmoid function to convert the linear combination into a probability value and get the final predicted value:

[0063]

[0064] Among them, the Sigmoid function is defined as:

[0065] Sigmoid(x)=(1+e -x ) -1

[0066] Use mean square error as the loss function Loss of the propensity score model:

[0067]

[0068] After completing the propensity score model and prediction, each sample will receive a propensity score. Then, sample matching is performed to match each intervention sample with a control sample with the highest score.

[0069] In order to better obtain matching objects, based on the extraction of important user features, combined with user propensity scores, a complete match is performed. Specifically, a control sample is matched for each intervention group sample in the control group to obtain a sample matching tuple (X, X'). First, the propensity score difference is calculated: for the intervention group sample X and the control group sample X', their propensity scores are p(X) and p(X') respectively. The propensity score difference is defined as:

[0070] Δp=|p(X)-p(X′)|

[0071] Calculate the user feature similarity. For each user feature vector X and X', calculate the Euclidean distance:

[0072]

[0073] Calculate the comprehensive matching score, combine the propensity score difference and feature similarity, and define the comprehensive matching score Score(X,X′):

[0074] Score(X,X′)=α·Δp+β·Dist(X,X′)

[0075] Among them, α and β are weight parameters used to balance the impact of propensity score differences and feature similarities on the matching score.

[0076] Select the best match. In the control group, select the control sample X′ with the smallest matching score for each intervention group sample X, that is:

[0077]

[0078] The lower the matching score Score(X,X′), G0 represents the control group, indicating that the matching tuples of X in the treatment group and X′ in the control group are closer, and the better the result of estimating the causal effect.

[0079] According to the potential causal model, the individual causal effect is calculated for each individual. The individual causal effect represents the difference in the results of the individual when it receives the intervention and when it does not receive the intervention. X″ represents the user feature with the lowest matching score with the control group corresponding to X. The causal effect CE is calculated as follows:

[0080]

[0081] Based on calculating the causal effect CE for each treatment group user and the matched control group user, and then depending on whether the relevant factor is applied, the average causal effect of ATT in each case is calculated as follows:

[0082]

[0083] Among them, n represents the number of people in the experimental group, and i represents the number of people.

[0084] Calculate the mean of each influencing factor and normalize it to get the proportional coefficient of the user cognitive driving force mechanism. i as follows:

[0085]

[0086] In social networks, the factors that determine the influence of messages in cross-community marketing topic dissemination include both message factors, user factors, and community factors. Message factors include topic popularity and topic diffusion, and user factors include activity, cognitive matching, and time domain psychology. Community factors include community acceptance and the degree of connection between communities. Combining these three factors, the cognitive driving force function of user UI is constructed as follows:

[0087] Con(ui)=β0+β1*Soc(ui)+β2*Com(ui)+β3*Per(ui)

[0088] Among them, parameters β1, β2, and β3 are causal effect coefficients obtained by causal inference from the potential causal model. β1, β2, and β3 represent the weights of social factors, community factors, and personal factors, respectively, and describe the proportion of the three factors in constituting cognitive driving force.

[0089] S32: Construct a time domain psychological description mechanism.

[0090] Extract information such as comments posted by users 24 hours before and after they are exposed to marketing topics.

[0091] Likes, favorites, etc. Take 30 minutes as a slice to describe the browsing topic information based on user behavior.

[0092] Comments, likes, reposts and other information content describe the user's temporal psychological state. The obtained information is vectorized in high dimensions through word vector embedding to obtain the user's temporal psychological state feature set, such as Figure 3 shown.

[0093] The user-posted information is one-hot encoded as the input data set D = (x1, y1), ..., (xn, yn) of the word vector model, where x i ∈X is the input, y iThe weight matrix W, W', W is the final word vector matrix required, and W' is the vector output by the hidden layer mapped to a vector z with the same size as the vocabulary i , the vector z i Use the Softmax function to convert it into the probability distribution of the predicted central word Yu i Do cross loss:

[0094]

[0095] Where N is the total number of samples, is the cross entropy loss value.

[0096] By iteratively minimizing the loss function, use Adam back propagation to update W, W' until convergence. Use W and the text of the extracted keywords to calculate the user's time domain vector:

[0097] Vec(Text(ui,t))=X i *W

[0098] The obtained user word vector is used as the input of the LSTM model to obtain the description of the user's psychological state in the current time domain.

[0099] In the process of marketing topic dissemination, due to the uncertainty of users' psychological states at different times, the dissemination of marketing topics is huge. Different psychological states at different times will prompt users to make completely different decisions, and the role of users' neighboring nodes is also huge. Therefore, this paper introduces LSTM network to analyze users' time domain psychology and consider the influence of neighboring users. Based on the analysis in Chapter 2 and the output of word vector embedding in the previous section, we can get the output of LSTM:

[0100] h t =σ(W o [h t-1 ,V t ]+b o )*tanh(σ(W f ·[h t-1 ,V t ]+b f ))*C t-1 +σ(W i ·[h t-1 ,V t ]+b i )*tanh(W C ·[h t-1 ,V t ]+b C )

[0101] Among them, bo, bf, bi, bc are the biases of the output gate, forget gate and input gate respectively. Ht represents the output of the LSTM model. After the softmax function, the time domain psychological coefficient O(t) is obtained as follows:

[0102] O(t)=softmax(h t )

[0103] Considering the influence of adjacent nodes and topics, the time domain psychological coefficient δ(t) is:

[0104]

[0105] Among them, the δ(t) coefficient is the result of measuring the user's temporal psychological state at time t.

[0106] S33. Build a cross-community marketing topic dissemination model.

[0107] S331 Cross-community state transition mechanism;

[0108] SIR propagation mechanism, studying the propagation of marketing topics in a cross-community context. Due to the repetitiveness of the cross-community marketing topic life cycle, a cross-community marketing topic iterative propagation model CC-SIR is established, and a cross-model community topic inter-layer state transition mechanism is constructed, such as Figure 4 As shown. In order to facilitate the description of the model and simplify the theoretical expression, this paper takes three communities as examples for research, namely community 0, community 1, and community 2. The topic first spreads in community 0, and then spreads to community 1 after being forwarded by critical users. Similarly, the propagation path of community 2 is also forwarded by critical users in community 1.

[0109] Assume that for an independent topic, the infection rate is λ, that is, the probability of going from state S to state I is λ. When the topic spreads across communities 1 and 2, there is a certain relationship between the infection rate and the community correlation. Since the correlation between community 1 and community 0 is relatively large, the spread effect of the topic in the two communities will also converge.

[0110] The topic propagation between communities will produce inter-layer conversion after being forwarded by critical users, and the conversion probability is directly related to the correlation between the two. The threshold ε for judging the correlation is specified, the cross-community state conversion equation is given, and the cross-model state conversion (i.e., I a ->S b )The probability is as follows:

[0111]

[0112] S332: Topic propagation mechanism;

[0113] Considering the repetitiveness of the life cycle in the process of cross-community marketing topics, this paper couples the SIR model with different community layers to construct the CC-SIR cross-community marketing topic propagation model. In this model, users in each topic layer have three states: susceptible, infected, and recovered. The construction of this model is based on the following assumptions (such as Figure 4 shown):

[0114] 1. Due to the short-term and explosive nature of topic propagation, this paper does not consider population dynamics such as birth, death, and mobility in the research process. It is assumed that the number of user nodes in the social network is equal at any time during the research period. Therefore, the sum of the user status ratios in the same community layer model at any time is:

[0115] S i +I i +R i =1

[0116] 2. The spread of topics is similar to the spread of infectious diseases. When a new user comes into contact with a user who has already spread a topic, there will inevitably be a certain rate of infection.

[0117] 3. During the dissemination process of cross-community marketing topics, critical users in different communities will directly inherit the susceptibility status, infection status, and immunity status of the topic.

[0118] Based on the above assumptions, this paper defines the propagation rules of topics in social networks as follows: at the same community level, when an infected node comes into contact with a susceptible node, the susceptible node will be converted into an infected node with a probability of λ without considering the influence of other factors. As the popularity of the topic decreases and time goes by, the infected node will be converted into an immune node with a probability of μ.

[0119] Based on the CC-SIR model and the above propagation rules, the dynamic equations of k cross-community marketing topic communities are obtained as follows:

[0120]

[0121] Due to the unidirectional nature of information dissemination in the community layer of each marketing topic, the state transition of users in the same layer is also unidirectional. That is, the user state can only change from susceptible state to infected state and then to immune state. Assuming that a user node ui has m neighbor nodes, the probability that n of the neighbors forward the topic of this layer follows the binomial distribution:

[0122]

[0123] The probability that any user ui forwards the topic at this level at time t is:

[0124]

[0125] Combined with the mean field theory, the infection rate at time t is obtained Let α represent the community C i The correlation with its predecessor community is Rlv(ci-1,ci), and the correlation threshold is set to ε, and the kinetic equation is as follows:

[0126]

[0127] By combining the above mean field theory with the CC-SIR model, the final dynamic equation of cross-community marketing topics is obtained, which accurately describes the propagation and evolution of topics in different communities, such as Figure 5 This model provides a theoretical basis for understanding and predicting the spread trend of topics in complex social networks.

[0128] The above embodiments further illustrate the purpose, technical solutions and advantages of the present invention in detail. It should be understood that the above embodiments are only preferred implementation modes of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made to the present invention within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A topic propagation method based on time domain psychology, characterized in that: include: Step A: Obtain user historical topic propagation data, which includes user information, user relationship network, marketing topic information, and social platform community network information; Preprocess the user's historical topic propagation data; Step B: extract relevant attributes from the preprocessed user topic propagation data; Step C: constructing a topic propagation model according to relevant attributes, the topic propagation model including: a user-driven cognitive measurement mechanism, a time-domain psychological state description mechanism, and a cross-community marketing information propagation model; the user-driven cognitive measurement mechanism is used to obtain the proportional coefficient of the user cognitive description mechanism parameters; the time-domain psychological state description mechanism is used to obtain the high-dimensional vector representation of the user; the cross-community marketing information propagation model predicts the topic propagation results according to the high-dimensional vector representation of the user and the proportional coefficient of the user cognitive description mechanism parameters; Step D: Predict the propagation trend of the topic to be propagated based on the topic propagation model.

2. The topic propagation method based on time domain psychology according to claim 1 is characterized in that: Preprocessing the user's historical topic propagation data includes: cleaning the historical topic propagation data to obtain structured data, and deleting duplicate data and invalid data in the structured data.

3. The topic propagation method based on time domain psychology according to claim 1 is characterized in that: The extracted topic propagation data related attributes include: user driving force, user activity, topic popularity, topic diffusion and community relevance.

4. The topic propagation method based on time domain psychology according to claim 3 is characterized in that: The user driving force includes: obtaining the average number of views, average number of comments and average number of reposts within a period of time after the user posts the topic information; setting a weakening coefficient, multiplying the average number of views by the weakening coefficient, and then calculating the sum of the average number of comments and the average number of reposts to obtain the user driving force; User activity is the sum of the number of topics posted by the user before the topic spreads and the number of topics forwarded by the user; Topic popularity is to comment on topic popularity using the half-life function; Topic diffusion refers to the degree of spread of a topic in social networks, that is, the popularity and spread of the topic; The community association is to calculate the keyword sets extracted from different communities; use mutual information to calculate the similarity between two communities; set a similarity threshold, compare the calculated similarity with the similarity threshold, and those greater than the similarity threshold are associated communities, otherwise they are unassociated communities.

5. The topic propagation method based on time domain psychology according to claim 1 is characterized in that: Constructing a user-driven cognitive measurement mechanism includes: establishing a propensity score model; vectorizing user information, user relationship networks, and community information, and using a propensity score model to process the vectorized data to obtain the user's causal effect; and constructing a cognitive driving force function based on the causal effect.

6. The topic propagation method based on time domain psychology according to claim 5, characterized in that: The cognitive driving force function expression is: Con(ui)=β0+β1*Soc(ui)+β2*Com(ui)+β3*Per(ui) Among them, β0 is a hyperparameter, β1, β2, β3 are the weights of social factors, community factors and personal factors, ui is the user, Soc(ui) is the extracted user social information vector, Com(ui) is the user community information vector, and Per(ui) is the user personal information vector.

7. The topic propagation method based on time domain psychology according to claim 1 is characterized in that: The time-domain psychological state description mechanism includes: obtaining the user's comment data on the topic; embedding the comment data into word vectors to obtain the user's time-domain psychological state feature set; inputting the user's time-domain psychological state feature set into the LSTM model to obtain the user's current time-domain psychological state description.

8. The topic propagation method based on time domain psychology according to claim 1 is characterized in that: The cross-community marketing information dissemination model includes: constructing a cross-community state transition mechanism and a topic propagation mechanism; constructing a cross-community state transformation equation based on the cross-community state transformation mechanism, and calculating the cross-model state transformation probability based on the cross-community state transformation equation; processing topics according to the topic propagation mechanism, and obtaining a cross-community marketing information dissemination model based on the cross-model state transformation probability.

Images