CN107944635A - A kind of information propagation forecast model and method for merging the topic factor - Google Patents
A kind of information propagation forecast model and method for merging the topic factor Download PDFInfo
- Publication number
- CN107944635A CN107944635A CN201711328590.0A CN201711328590A CN107944635A CN 107944635 A CN107944635 A CN 107944635A CN 201711328590 A CN201711328590 A CN 201711328590A CN 107944635 A CN107944635 A CN 107944635A
- Authority
- CN
- China
- Prior art keywords
- user
- topic
- model
- function
- propagation
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000004927 fusion Effects 0.000 claims abstract description 19
- 238000005070 sampling Methods 0.000 claims abstract description 16
- 230000008451 emotion Effects 0.000 claims abstract description 5
- 238000011156 evaluation Methods 0.000 claims abstract description 5
- 239000011159 matrix material Substances 0.000 claims description 23
- 230000000926 neurological effect Effects 0.000 claims description 22
- 230000004083 survival effect Effects 0.000 claims description 19
- 230000000694 effects Effects 0.000 claims description 15
- 230000006854 communication Effects 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 8
- 230000007480 spreading Effects 0.000 claims description 6
- 230000001186 cumulative effect Effects 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000003068 static effect Effects 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims 1
- 238000011478 gradient descent method Methods 0.000 claims 1
- 238000010606 normalization Methods 0.000 claims 1
- 239000012141 concentrate Substances 0.000 abstract description 2
- 238000009825 accumulation Methods 0.000 description 3
- 230000001902 propagating effect Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 208000015181 infectious disease Diseases 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000002889 sympathetic effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Tourism & Hospitality (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of information propagation forecast model and method for merging the topic factor, which includes：One is vectorial by means of the topic of LDA topic models study model；One existence analysis model；One negative sampling algorithm comprising modules；The emotion information propagation model of the one fusion topic factor；Data basic assumption module；And experimental duties and corresponding evaluation index, the index are used to predict dissemination of the user to different topic informations under potential network.The present invention can predict the propagation path being forwarded of different topic models exactly, and can be applied to fairly large cascade data and concentrate.
Description
Technical field
The invention belongs to information communication sphere, relates more specifically to a kind of information propagation forecast model for merging the topic factor
And method.
Background technology
Currently, there is the research that many technical methods can be used for information propagation forecast.Traditional Information Propagation Model includes only
Vertical cascade model and linear threshold model.Sometime the node of network is there are two states in two kinds of models given, and one
Kind is enlivens state, and another kind is inactive state, and inactive state can switch to enliven state, otherwise cannot.Both of which is
By setting activation threshold or activation probability directly to judge whether propagated between user.This method have ignored the difference between user
Property, the flexibility of model is limited to a certain extent.
Currently, it is also proposed that information dissemination mechanism research based on community network and based on survival analysis scale-model investigation.Its
In, the Information Propagation Model in community network combines network structure, and extraction propagates relevant user characteristics and content with information
Feature is modeled, including interest similarity, friend's note between user's itself affect power and liveness, the content of information, user
Content of son etc..In addition, in recent years, domestic and international researcher begins attempt to utilize survival analysis model（Survival
Analysis Model）And its propagation rate between variation study user couple（Transmission Rate）, then according to existing
It was observed that network data infer potential information propagation path, be allowed to and the deviation of real information propagation path as far as possible
It is small.Especially, it is assumed that the probability that generation is propagated between user depends on propagation rate between user's infected time and user,
Likelihood by maximizing cascade learns the probability of spreading between user.
In general, user is to meeting personal interest or the topic model of strong sympathetic response can be caused often more to go
They are forwarded to, share and is commented on.Unfortunately, existing more exponential models only consider that user is infected under continuous time
Timestamp and user may show the content of microblog comprising different emotions polarity different infection risks, and have ignored on
The topic that message content is talked about is also information of forecasting in following an important factor for whether can propagating.Therefore, people are urgent
A kind of information propagation prediction method of more efficiently and accurately is wished to, this method is distributed using the topic of model to adjust user
Influence power and neurological susceptibility matrix, and then change the propagation rate between user, and add between the method measure user of negative example sampling
Probability of spreading, finally infers potential information propagation path according to the existing network data observed.
The content of the invention
It is therefore an object of the present invention to propose a kind of information propagation forecast for merging the topic factor, can be efficiently and accurately
It is predicted, and can be applied among the propagation forecast of more massive cascade data collection.
To achieve the above object, the present invention uses following technical scheme：A kind of information propagation forecast for merging the topic factor
Model, it includes：One is vectorial by means of the topic of LDA topic models study model；One existence analysis model, it is used to portray use
Family behavior, by the propagation rate between topic vector adjustment user, and using the probability of spreading of time of fusion attenuation factor
Model Power-Law methods learn the negative log-likelihood function that one group of observable concatenated set minimizes；One negative sampling algorithm
Comprising modules, for overcoming all negative examples to limit model suitable for large-scale data and the balance of optimization object function
Property；The distributed expression algorithm of the emotion information propagation model study user of the one fusion topic factor；And experimental duties and right
The evaluation index answered, the index are used to predict dissemination of the user to different topic informations under potential network.
In an embodiment of the present invention, data preprocessing module is further included；The data preprocessing module turns between user
In the case that hair relation and the relation that is forwarded are unknown, the infected time series of user is only remained as initial cascade data
Collection.
In an embodiment of the present invention, data basic assumption module, sets in the data basic assumption module：Information passes
Broadcast process to occur on static network, which will not change with the change of time；If some not infected node
After being infected by its first father node, it will not be infected be subject to other father nodes again, and infected node
Not infected node can only be infected；The propagation of viewpoint is determined jointly by the influence power of disseminator and the neurological susceptibility of recipient between user
It is fixed.
In an embodiment of the present invention, function is passed through in the survival analysis modelCarry out
Sampling, so as to initialize the influence power matrix and neurological susceptibility matrix of each user in cascade data set, specifically includes following steps：
Propagation rate function between user from the influence power matrix of disseminator, the neurological susceptibility matrix of recipient, microblogging model topic to
What amount and feeling polarities collectively constituted；Calculate the probability density function between user using propagation rate, afterwards to obtain accumulation general
Rate density function；Then survival analysis model is introduced, calculates the existence letter that user is not affected by other infected customer impacts
Number and the risk function for being subject to other infected customer impacts.
In an embodiment of the present invention, in given time window, for a cascade, non-source node user is calculated at certain
A moment infected likelihood and an observable joint likelihood for propagating cascade, and can adding survival probability acquisition one
The likelihood formula of the cascade observed；Assuming that independent mutually between cascade, then one group of observable concatenated set minimizes negative
Log-likelihood function is object function；For bearing example likelihood in object function, truly believed with bearing example user in one group of cascade
Infected frequency carries out probability sampling and replaces the situation that original method considers all negative examples in breath propagation.
In an embodiment of the present invention, influence power matrix and neurological susceptibility matrix are subjected to piecemeal, respectively obtain positive influences
The negative neurological susceptibility vector of force vector, negative effect force vector, positive neurological susceptibility vector sum, then using the boarding steps with projection
Degree descent method is solved；After given network, the information content and Initial travel state, it is applied to task assessment and is carried
Effect of the method gone out in information of forecasting communication process.
The present invention also provides a kind of information propagation prediction method for merging the topic factor, it is characterised in that：Including following step
Suddenly：Step S1：Data are filtered, only retain every cascade infected timestamp of user；Step S2：User is defined first
Between propagation rate function be made of the influence power matrix of disseminator and the neurological susceptibility matrix of recipient, and add topic vector
It is adjusted；In the probability propagation model Power-Law of time of fusion attenuation factor, calculate probability density function and accumulation is general
Rate density function, finally using survival analysis model construction survival function, risk function, and obtains one group of observable cascade collection
Close the negative log-likelihood function minimized；Step S3：It is adopted according to the frequency distribution that negative example occurs in one group of cascade
Sample；Step S4:After given network, the information content and Initial travel state, it is applied to task and assesses proposed side
Effect of the method in information of forecasting communication process.
In an embodiment of the present invention, filtering rule includes in step S1：1）According to the timeliness of model, front and rear will turn
The user that hair exceedes behind week age forwards sequence to remove；2）The activity of the user is defined to forward for the user in data set
The sum of number that other people number and the user is forwarded by other people；3）Choose user activity in data set and exceed some threshold value
User as seed user, forward relation chain for every model, sort from big to small according to the ratio shared by any active ues
The small cascade of ratio is deleted afterwards.
In an embodiment of the present invention, step S3 comprises the following steps：Step S31：Before each negative example of calculating does not normalize
Distribution Value, i.e., the number occurred in concatenated set；Step S32：Count the sum of frequency of all negative examples, and using the value into
Row probability normalizes；Step S33:It is ranked up according to the id of negative example user, and calculates corresponding probability, with cumulative distribution letter
Several forms is arranged on a line segment；Step S34:Line segment is evenly dividing as m sections, it is 0 that left end point 0, which corresponds to probable value, right
The corresponding probable values of endpoint m are 1；After the line segment obtained in the line segment and step S33 is mapped, show that every section of section institute is right
The negative example Customs Assigned Number answered；Step S35：Random sampling is carried out to the value in 0-m, it is to adopt to obtain the corresponding Customs Assigned Number of the value
The negative example of sample, probability size are directly proportional to 3/4 power of negative example frequency；Step S36：According to predetermined negative example sampling number, no
It is disconnected to perform step S35.
Compared with prior art, the present invention can predict the propagation path being forwarded of different topic models exactly, and
Fairly large cascade data is can be applied to concentrate.
Brief description of the drawings
Fig. 1 is one embodiment of the invention, in the schematic configuration view of the information propagation forecast prototype system of the fusion topic factor.
Embodiment
Explanation is further explained to the present invention with specific embodiment below in conjunction with the accompanying drawings.
A kind of information propagation forecast model for merging the topic factor, it includes：One learns model by means of LDA topic models
Topic vector；One existence analysis model, it is used to portray user behavior, passes through the propagation between topic vector adjustment user
Speed, and then one group of Power-Law methods study is observable for the probability of spreading model power method of use time of fusion attenuation factor
The negative log-likelihood function that concatenated set minimizes；One negative sampling algorithm comprising modules, for overcoming all negative examples to limit
Model is suitable for large-scale data and the balance of optimization object function；The emotion information propagation model of the one fusion topic factor
Learn the distributed expression algorithm of user；And experimental duties and corresponding evaluation index, the index are used to predict user latent
To the dissemination of different topic informations under network.Power-Law formula are expressed as:Middle o
(xk) it is to represent big O as xkAny x functions.
In an embodiment of the present invention, data preprocessing module is further included；The data preprocessing module turns between user
In the case that hair relation and the relation that is forwarded are unknown, the infected time series of user is only remained as initial cascade data set
In an embodiment of the present invention, data basic assumption module, sets in the data basic assumption module：Information communication process is sent out
On static network, which will not change with the change of time for life；If some not infected node is by its
After one father node infects, it will not be infected be subject to other father nodes again, and infected node can only infect
Not infected node；The propagation of viewpoint is together decided on by the influence power of disseminator and the neurological susceptibility of recipient between user.
In an embodiment of the present invention, function is passed through in the survival analysis modelCarry out
Sampling, so as to initialize the influence power matrix and neurological susceptibility matrix of each user in cascade data set, specifically includes following steps：
Propagation rate function between user from the influence power matrix of disseminator, the neurological susceptibility matrix of recipient, microblogging model topic to
What amount and feeling polarities collectively constituted；Calculate the probability density function between user using propagation rate, afterwards to obtain accumulation general
Rate density function；Then survival analysis model is introduced, calculates the existence letter that user is not affected by other infected customer impacts
Number and the risk function for being subject to other infected customer impacts.
In an embodiment of the present invention, in given time window, for a cascade, non-source node user is calculated at certain
A moment infected likelihood and an observable joint likelihood for propagating cascade, and can adding survival probability acquisition one
The likelihood formula of the cascade observed；Assuming that independent mutually between cascade, then one group of observable concatenated set minimizes negative
Log-likelihood function is object function；For bearing example likelihood in object function, truly believed with bearing example user in one group of cascade
Infected frequency carries out probability sampling and replaces the situation that original method considers all negative examples in breath propagation.
In an embodiment of the present invention, influence power matrix and neurological susceptibility matrix are subjected to piecemeal, respectively obtain positive influences
The negative neurological susceptibility vector of force vector, negative effect force vector, positive neurological susceptibility vector sum, then using the boarding steps with projection
Degree descent method is solved；After given network, the information content and Initial travel state, it is applied to " prediction stage linkage
State ", the task such as " who will be forwarded " and " prediction of cascade size " assess proposed method in information of forecasting communication process
Effect.
The present invention also provides a kind of information propagation prediction method for merging the topic factor, it comprises the following steps：Step S1：
Data are filtered, only retain every cascade infected timestamp of user；Step S2：The propagation speed between user is defined first
Rate function is made of the influence power matrix of disseminator and the neurological susceptibility matrix of recipient, and is added topic vector and be adjusted；
In the probability propagation model Power-Law of time of fusion attenuation factor, probability density function and cumulative probability density letter are calculated
Number, finally using survival analysis model construction survival function, risk function, and obtains one group of observable concatenated set and minimizes
Negative log-likelihood function；Step S3：It is sampled according to the frequency distribution that negative example occurs in one group of cascade；Step
S4:After given network, the information content and Initial travel state, it is applied to " prediction stage linkage state ", " who will be turned
The task such as hair " and " prediction of cascade size " assesses effect of the proposed method in information of forecasting communication process.
As shown in Figure 1, pretreatment module, it is relatively low to filter out the frequency that user occurs in forwarding relation, so as to reduce pair
The training and prediction of model generate very big interference.Using survival analysis model construction module, by LDA learn topic to
Propagation rate between amount adjustment user, and using the probability of spreading model Power-Law methods of time of fusion attenuation factor, to learn
Practise the negative log-likelihood function that one group of observable concatenated set minimizes.Overcome model excellent using negative sampling algorithm comprising modules
The positive and negative serious imbalance problem of example in change.By three experimental duties and evaluation index, carry out the precision that information of forecasting propagates trend,
Output module.
In an embodiment of the present invention, filtering rule includes in step S1：1）According to the timeliness of model, front and rear will turn
The user that hair exceedes behind week age forwards sequence to remove；2）The activity of the user is defined to forward for the user in data set
The sum of number that other people number and the user is forwarded by other people；3）Choose user activity in data set and exceed some threshold value
User as seed user, forward relation chain for every model, sort from big to small according to the ratio shared by any active ues
The small cascade of ratio is deleted afterwards.
Above is presently preferred embodiments of the present invention, all changes made according to technical solution of the present invention, caused function are made
During with scope without departing from technical solution of the present invention, protection scope of the present invention is belonged to.
Claims (9)
- A kind of 1. information propagation forecast model for merging the topic factor, it is characterised in that：Including：One is vectorial by means of the topic of LDA topic models study model；One existence analysis model, it is used to portray user behavior, by the propagation rate between topic vector adjustment user, and Minimized using the probability of spreading model Power-Law of time of fusion attenuation factor to learn one group of observable concatenated set Negative log-likelihood function；One negative sampling algorithm comprising modules, are suitable for large-scale data and optimization for overcoming all negative examples to limit model The balance of object function；The emotion information propagation model of the one fusion topic factor, it learns the distributed expression algorithm of user；And experimental duties and corresponding evaluation index, the index are used to predict user under potential network to different topic informations Dissemination.
- 2. the information propagation forecast model of the fusion topic factor according to claim 1, it is characterised in that：Further include data Pretreatment module；In the case that the data preprocessing module forwards relation and the relation that is forwarded unknown between user, only retain The infected time series of user is as initial cascade data set.
- 3. the information propagation forecast model of the fusion topic factor according to claim 1, it is characterised in that：Further include data Basic assumption module, sets in the data basic assumption module：Information communication process occurs on static network, and the network is not It can change with the change of time；After if some not infected node is infected by its first father node, it will It will not be infected again be subject to other father nodes, and infected node can only infect not infected node；Seen between user The propagation of point is together decided on by the influence power of disseminator and the neurological susceptibility of recipient.
- 4. the information propagation forecast model of the fusion topic factor according to claim 1, it is characterised in that：The existence point Pass through function in analysis modelSampled, so as to initialize each user in cascade data set Influence power matrix and neurological susceptibility matrix, specifically include following steps：Propagation rate function between user by disseminator influence power Matrix, the neurological susceptibility matrix of recipient, the topic vector sum feeling polarities of microblogging model collectively constitute；Utilize propagation rate meter Calculate user between probability density function, afterwards obtain cumulative probability density function；Then survival analysis model is introduced, calculates and uses Family is not affected by the survival function of other infected customer impacts and is subject to the risk letter of other infected customer impacts Number.
- 5. the information propagation forecast model of the fusion topic factor according to claim 4, it is characterised in that：The given time In window, for a cascade, non-source node user is calculated in sometime infected likelihood and an observable propagation The joint likelihood of cascade, and in the likelihood formula for adding survival probability one cascade that can observe of acquisition；Assuming that between cascade Independent mutually, then the negative log-likelihood function that one group of observable concatenated set minimizes is object function；For target letter Example likelihood is born in number, probability sampling replacement is carried out to bear example user infected frequency in real information propagation in one group of cascade Original method considers the situation of all negative examples.
- 6. the information propagation forecast model of the fusion topic factor according to claim 4, it is characterised in that：It will influence torque Battle array and neurological susceptibility matrix carry out piecemeal, respectively obtain positive influences force vector, negative effect force vector, positive neurological susceptibility vector sum Negative neurological susceptibility vector, is then solved using the stochastic gradient descent method with projection；In given network, the information content After Initial travel state, it is applied to task and assesses effect of the proposed method in information of forecasting communication process.
- A kind of 7. information propagation prediction method for merging the topic factor, it is characterised in that：Comprise the following steps：Step S1:Data are filtered, only retain every cascade infected timestamp of user；Step S2:The propagation rate function between user is defined first by the influence power matrix of disseminator and the neurological susceptibility square of recipient Battle array composition, and add topic vector and be adjusted；In the probability propagation model Power-Law of time of fusion attenuation factor, Probability density function and cumulative probability density function are calculated, finally using survival analysis model construction survival function, risk function, And obtain the negative log-likelihood function that one group of observable concatenated set minimizes；Step S3：It is sampled according to the frequency distribution that negative example occurs in one group of cascade；Step S4：After given network, the information content and Initial travel state, it is applied to task and assesses proposed side Effect of the method in information of forecasting communication process.
- 8. the information propagation prediction method of the fusion topic factor according to claim 7, it is characterised in that：In step S1Filtering rule includes：1）According to the timeliness of model, front and rear forwarding is exceeded to the user behind week age and forwards sequence to remove；2）The number that the activity of the user forwards other people number for the user in data set and the user is forwarded by other people is defined The sum of；3）The user that user activity exceedes some threshold value in selection data set forwards as seed user for every model Relation chain, deletes the small cascade of ratio after sorting from big to small according to the ratio shared by any active ues.
- 9. the information propagation prediction method of the fusion topic factor according to claim 6, it is characterised in that：Step S3 includes Following steps：Step S31：Calculate the Distribution Value before each negative example does not normalize, i.e., the number occurred in concatenated set；Step S32：The sum of frequency of all negative examples is counted, and probability normalization is carried out using the value；Step S33:It is ranked up according to the id of negative example user, and calculates corresponding probability, in the form of cumulative distribution function It is arranged on a line segment；Step S34:Line segment is evenly dividing as m sections, it is 0 that left end point 0, which corresponds to probable value, and the corresponding probable values of right endpoint m are 1； After the line segment obtained in the line segment and step S33 is mapped, the negative example Customs Assigned Number corresponding to every section of section is drawn；Step S35：Random sampling is carried out to the value in 0-m, obtains the negative example that the corresponding Customs Assigned Number of the value is sampling, probability Size is directly proportional to 3/4 power of negative example frequency；Step S36：According to predetermined negative example sampling number, step S35 is constantly performed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711328590.0A CN107944635A (en) | 2017-12-13 | 2017-12-13 | A kind of information propagation forecast model and method for merging the topic factor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711328590.0A CN107944635A (en) | 2017-12-13 | 2017-12-13 | A kind of information propagation forecast model and method for merging the topic factor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107944635A true CN107944635A (en) | 2018-04-20 |
Family
ID=61942886
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711328590.0A Pending CN107944635A (en) | 2017-12-13 | 2017-12-13 | A kind of information propagation forecast model and method for merging the topic factor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107944635A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165239A (en) * | 2018-07-06 | 2019-01-08 | 重庆邮电大学 | Modeling method is propagated based on the D2D information of user property and contents attribute |
| CN110163404A (en) * | 2018-06-12 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of diffusion of information prediction technique, device and server, storage medium |
| CN110932909A (en) * | 2019-12-05 | 2020-03-27 | 中国传媒大学 | Information propagation prediction method, system and storage medium |
| CN111753911A (en) * | 2020-06-28 | 2020-10-09 | 北京百度网讯科技有限公司 | Method and apparatus for fusing models |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107341270A (en) * | 2017-07-28 | 2017-11-10 | 东北大学 | Towards the user feeling influence power analysis method of social platform |
-
2017
- 2017-12-13 CN CN201711328590.0A patent/CN107944635A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107341270A (en) * | 2017-07-28 | 2017-11-10 | 东北大学 | Towards the user feeling influence power analysis method of social platform |
Non-Patent Citations (2)
| Title |
|---|
| DU N, SONG L, WOOH,ET AL.: "Uncover topic-sensitive information diffusion networks", 《IN PROCEEDINGS OF THE 16TH INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS》 * |
| 廖详文等: "基于用户行为的情感影响力和易感性学习", 《计算机学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110163404A (en) * | 2018-06-12 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of diffusion of information prediction technique, device and server, storage medium |
| CN109165239A (en) * | 2018-07-06 | 2019-01-08 | 重庆邮电大学 | Modeling method is propagated based on the D2D information of user property and contents attribute |
| CN109165239B (en) * | 2018-07-06 | 2021-05-18 | 重庆邮电大学 | D2D information propagation modeling method based on user attributes and content attributes |
| CN110932909A (en) * | 2019-12-05 | 2020-03-27 | 中国传媒大学 | Information propagation prediction method, system and storage medium |
| CN110932909B (en) * | 2019-12-05 | 2022-02-18 | 中国传媒大学 | Information propagation prediction method, system and storage medium |
| CN111753911A (en) * | 2020-06-28 | 2020-10-09 | 北京百度网讯科技有限公司 | Method and apparatus for fusing models |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107944635A (en) | A kind of information propagation forecast model and method for merging the topic factor | |
| Beskow et al. | Bot conversations are different: leveraging network metrics for bot detection in twitter | |
| Tang et al. | Analysing information flows and key mediators through temporal centrality metrics | |
| CN103064917B (en) | The high-impact customer group of a kind of specific tendency towards microblogging finds method | |
| Kivelä et al. | Multiscale analysis of spreading in a large communication network | |
| Zhao et al. | Multiple routes transmitted epidemics on multiplex networks | |
| CN103457909B (en) | A kind of Botnet detection method and device | |
| Gao et al. | Measures of node centrality in mobile social networks | |
| EP3291164A1 (en) | Quinquelateral relationship-based kinship network system | |
| Cao et al. | SybilFence: Improving social-graph-based sybil defenses with user negative feedback | |
| CN108022171A (en) | A kind of data processing method and equipment | |
| Xuan et al. | A self-learning information diffusion model for smart social networks | |
| CN108229731A (en) | The user's behavior prediction system and method that more message mutually influence under a kind of much-talked-about topic | |
| Signorelli et al. | A penalized inference approach to stochastic block modelling of community structure in the Italian Parliament | |
| CN103793747A (en) | Sensitive information template construction method in network content safety management | |
| Wang et al. | The impact of different strategy update mechanisms on information dissemination under hyper network vision | |
| CN108520337B (en) | Riadry risk assessment method based on network risk entropy difference | |
| Manasa et al. | Tweet spam detection using machine learning and swarm optimization techniques | |
| CN109858510A (en) | A kind of detection method for http protocol ETag value covert communications | |
| Fang et al. | Social network public opinion research based on S-SEIR epidemic model | |
| Schuman et al. | Variable structure dynamic artificial neural networks | |
| CN108536776A (en) | Unification user malicious act detection method and system in a kind of social networks | |
| Manna et al. | Generalized contact matrices for epidemic modeling | |
| CN112256756B (en) | Influence discovery method based on ternary association diagram and knowledge representation | |
| Kumar et al. | Investigating social network as complex network and dynamics of user activities |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180420 |