CN106651030A - Method for predicting user participation behavior of hot topic by improved RBF neural network - Google Patents
Method for predicting user participation behavior of hot topic by improved RBF neural network Download PDFInfo
- Publication number
- CN106651030A CN106651030A CN201611189974.4A CN201611189974A CN106651030A CN 106651030 A CN106651030 A CN 106651030A CN 201611189974 A CN201611189974 A CN 201611189974A CN 106651030 A CN106651030 A CN 106651030A
- Authority
- CN
- China
- Prior art keywords
- user
- topic
- cloud
- attribute
- rbf neural
- 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.)
- Granted
Links
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Abstract
The invention discloses a method for predicting a user participation behavior of a hot topic by an improved RBF (Radical Basis Function) neural network, and belongs to the field of computer network information technology analysis. Firstly, the situation that a neural network can take a good fitting effect for a complex nonlinear relationship among user behaviors is considered, and further a user participation behavior prediction model is built by adopting the RBF neural network; secondly, a mapping relationship between a user attribute and the participation behavior has uncertainty, and a cloud theory is introduced for optimizing an activation function of a hidden layer in an RBF; and finally, topic popularity is subjected to exponential function model-based parameter fitting by utilizing time discretization and time slicing methods for a characteristic that the participation behavior of a user is changed with time, so that a topic popularity change trend is obtained.
Description
Technical field
The invention belongs to network topics analysis field, more particularly in social networks the user behavior analysis of much-talked-about topic with
Prediction.
Background technology
In recent years, with the continuous popularization and development of internet, social networks increasingly becomes the weight in many people's lives
Part is wanted, microblogging is one of wherein very representational social networks, it is a kind of social network based on concern mechanism
Network platform, can not only allow the other users that user independently selects oneself interested to be listened to, paid close attention to, and also can freely issue
The message of oneself, the message delivered has broadcast nature simultaneously, i.e., everyone can see, so, microblogging not only has society
The function of network is handed over, while also having both the property of media.Microblog has attracted big portion of China as a kind of new public opinion medium
The subnetting people participate in, and also quickly diffusion becomes the focus incident of entire society to much-talked-about topic therein, and social effectiveness also swashs therewith
Increase.Meanwhile, microblogging also adheres to community network media spirit freely, open and shared, compared with conventional traditional media, gives
The channel of each individual freedom expression exchange, also causes it to become emerging language and propagates platform.It means that predicting certain topic
The temperature being up to, has very important significance on Public Opinion Transmission with control.Can not only deliver early stage predict it can
Can coverage, it is also possible to timely control public opinion trend in development mid-term.
The thing followed also shows the trend of rapid growth to the analysis demand of topic data, therefore, have more
Come the developing state that more researchers begin to focus on topic.And the change of topic temperature can pass through to participate in the people of the topic
Count dynamic change to embody, the prediction of user's participative behavior is roughly divided at present following several：Based on the passing behavior of user
Prediction, based on the prediction of user version interest, prediction based on group influence suffered by user etc..As Zaman et al. exists
《Predicting information spreading in Twitter》It is middle to propose a kind of user based on collaborative filtering model
Behavior prediction method, is predicted by building user profile matrix.Zhang et al. exists《Retweet behavior
prediction using hierarchical dirichlet process》It is middle to propose that a kind of being based on is layered Di Li Cray mistakes
The nonparametric Bayes model of journey, to the interest of user dynamic theme modeling is carried out.Luo et al. exists《Who will
retweet meFinding retweeters in Twitter》Used in sequence learning method based on Pointwise,
For being possible to forward the user of certain microblogging to carry out top-K sequences, according to various determined properties, whether the user can produce forwarding
Behavior.
There is great probabilistic feature because above-mentioned prior art cannot embody user's participative behavior, have ignored
User can not well fit actual feelings in the randomness and ambiguity being made whether when participating in the decision of the topic
Condition, result in can not obtain good prediction effect.Meanwhile, most achievements in research carry out pre- for static participative behavior
Survey, it is impossible to embody user and participate in quantitative dynamic change, therefore the situation of topic can not be perceived.Therefore the present invention adopts mould
Method of the Clouds theory in paste mathematics in combination with RBF neural, makes the forecast model to rise to user's participative behavior
While acting on to good nonlinear fitting, additionally it is possible to embody the randomness of user behavior and the feature of ambiguity.Its difficult point
It is the Feature Selection of user behavior and how represents many Feature Conversions for qualitatively cloud model.
The content of the invention
When the present invention is predicted for neural network algorithm in prior art, local minimum and convergence are easily trapped into
Speed is slow, simultaneously because user behavior complex genesis, it is impossible to accurately embody ambiguity between user property and user behavior with
Randomness, and user's participative behavior with time dynamic the problems such as.The present invention proposes a kind of much-talked-about topic user and participates in row
For Forecasting Methodology.Bean vermicelli of the method research already engaged in topic user, if this can be continued to participate under the influence of various factors
Topic.Meanwhile, it is neural by RBF respectively from user's bean vermicelli unique characteristics attribute, user outside two angles of social attribute
Network carries out user's behavior prediction.Because user behavior has ambiguity and randomness, therefore draw in the learning process of model
Enter Clouds theory, cloud model is replaced into the Gaussian function in RBF neural, meet the not true of user's participative behavior in network topics
It is qualitative.And then the classification problem of user's participative behavior is converted into topic temperature forecasting problem, processed by isochronous surface, and
Parameter fitting is carried out by exponential Function Model, so as to show that topic temperature situation is moved towards.Propose a kind of improved RBF neural
Network hot topic user's participative behavior Forecasting Methodology.Technical scheme is as follows：
A kind of improved RBF neural much-talked-about topic user participative behavior Forecasting Methodology, it is comprised the following steps：
S1：Obtain from the API of existing social platform, or the content obtaining in web page is captured by web crawlers
Social network user data；
S2：The step of extracting association attributes：Participating in topic main cause in view of potential user includes individual subscriber feature
The impact of attribute and user outside social attribute, will extract association attributes in terms of the two；And when doing to the information of user
Between Slice process,
S3：The step of setting up model：The data that user property is carried out based on Cloud transform are fitted, obtaining after Normal Cloud can be with
Higher-dimension cloud is constructed, the number of higher-dimension cloud is the neuron number of hidden layer in RBF neural, and its parameter is hidden layer
The cluster centre and bandwidth of excitation function, is trained by determining after parameter to RBF neural, obtains predicting the user
Whether the forecast model of topic can be participated in；
S4：Prediction and analysis process steps：By predicting following topic trend trend using exponential smoothing, will measure in advance
The much-talked-about topic for going out participates in the time series (y of number1, y2..., yn) do Three-exponential Smoothing calculating, you can fit focus words
The temperature Long-term change trend of topic, so as to carry out the prediction to subsequent time period.
Further, the particular content that step S1 obtains social network user data is under certain hotspot topic
User's participative behavior data and user relationship data；User's participative behavior data include the topic be forwarded and comment on when
Between, the personal information of participating user and historical behavior data；User relationship data includes all of the user participated under the topic
Bean vermicelli and concern user, and their personal information.
Further, step S2 extracts association attributes according to individual subscriber characteristic attribute, mainly includes：
Extract potential user's personal characteristics attribute：The personal characteristics attribute of potential user mainly includes that 1. potential user is whether
For any active ues isActivity (vi)；2. potential user viLabel in whether include and much-talked-about topic identical keyword
isSameTag(Vi)；3. the quantity countOfHF (v of topic is participated in the concern user of potential useri)；4. the pass of potential user
The topic drive inf (v of note useri)；About the unique characteristics attribute x of potential user by more thanikUnified Form description,
Represent potential user viK-th attribute.
Further, step S2 extracts association attributes according to user outside social attribute, mainly includes：Potential user
Outside social attribute mainly include 1. potential user viConcern user whether be certification user isVip (vi,vj)；2. it is potential
User viConcern user whether be leader of opinion isLeader (vi,vj)；3. potential user viIt is identical user's sex to be paid close attention to it
isSameS(vi,vj)；4. potential user viDifferent isDifS (the v of user's sex are paid close attention to from iti,vj)；5. potential user viClose with it
Identical isSameL (the v of note user locationsi,vj)；6. potential user viDifferent isDifL (the v of user locations are paid close attention to from iti,vj)；Together
When, it is also relevant with the corporations' influence power residing for it whether potential user can participate in the much-talked-about topic, therefore defines its team's attribute
For groupInf (vi,Cm), i.e. C residing for potential usermWhether corporations are corporations interested in the much-talked-about topic；By more than
About the unique characteristics attribute x of potential userikUnified Form description, represent potential user viK-th attribute.
Further, the step of step S3 sets up model mainly divides following 4 steps：
S31：User's unique characteristics attribute and user outside social attribute are respectively adopted by Maximum Approach and carry out Cloud transform, cloud
Conversion carries out mathematic(al) manipulation to any irregular data distribution, and it can carry out the fuzzy clustering of soft classification to sample point, make
It becomes the superposition of several different clouds；
S32：By by the property value of Cloud transform in combination with RBF neural, so that it is determined that cloud model hidden layer god
Jing is first；Changed according to Peak Intensity Method cloud, for every dimension attribute X of input layer, n can be obtainediIndividual fitting Normal Cloud, according to higher-dimension cloud
Theoretical and n dimension Normal Cloud Generators construct n and tie up Normal Cloud as the neuron of RBF neural hidden layer, can obtain (n1
×n2×...×nn) individual n dimensions cloud model, i.e., n hidden layer node；
S33：From based on the improved hidden layer neuron of n dimension cloud models desired value is taken as RBF neural hidden layer
The final output value of neuron；
S34：Hidden layer is to being a kind of perceptron algorithm model between output layer in RBF neural, and due to output layer
Node is made up of linear function, and using least square method the weights of connection are solved.
Further, the step S4 prediction and analysis process steps include：
S41：The time series of acquisition is done into Three-exponential Smoothing conversion；
S42：When time series embodies conic section trend, that is, conic section correction model is set up, the model is non-
Linear prediction model, can embody the variation tendency of sequential, predict the development trend of topic temperature.
Advantages of the present invention and have the beneficial effect that：
It is of the invention first, it is contemplated that neutral net can play good plan to non-linear relation complicated between user behavior
Effect is closed, and further user is built using RBF (Radical Basis Function, RBF) neutral net and participated in
Behavior prediction model, can process large scale network topic data when have fast convergence rate, being capable of partial approximation characteristic value
Advantage, and be difficult to be absorbed in local minimum；Secondly as the mapping relations between user property and participative behavior have not
Certainty, is introduced into Clouds theory (Cloud) and the activation primitive of hidden layer in RBF is optimized so that the model can either be abundant
The ambiguity and randomness of expression user's participative behavior, can have good approximation capability for non-linear relation again；Finally, pin
The characteristics of changing over to the participative behavior of user, using time discretization and time dicing method, is carried out to topic temperature
Based on the parameter fitting of exponential Function Model, so as to draw topic temperature variation tendency.
The present invention proposes improved RBF neural much-talked-about topic user participative behavior Forecasting Methodology, can not only be abundant
The ambiguity and randomness of expression user's participative behavior, can have good approximation capability, Er Qieneng for non-linear relation again
The temperature change of topic is enough embodied by user's participative behavior, so as to perceive topic situation, effective public sentiment monitoring and pipe is carried out
Control.
Description of the drawings
Fig. 1 is the entire block diagram that the present invention provides preferred embodiment.
Fig. 2 is that the present invention provides preferred embodiment improved RBF neural much-talked-about topic user participative behavior prediction side
The overview flow chart of method.
Fig. 3 is the forecast model figure of the present invention.
Fig. 4 is the learning algorithm flow chart of the present invention.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, detailed
Carefully describe.Described embodiment is only a part of embodiment of the present invention.
The present invention solves the technical scheme of above-mentioned technical problem：
Entire block diagram of the present invention is illustrated in figure 1, the input for showing the present invention is each of topic lower network structure and user
Item feature, the output after forecast model is whether the bean vermicelli for having participated in topic user, i.e. potential user can participate in the words
That what is inscribed predicts the outcome.The overview flow chart of the present invention is illustrated in figure 2, including：Data module is obtained, attribute module, structure is parsed
Established model module, the common four module of forecast analysis module.Illustrate the detailed implementation process of the present invention, including following four steps
Suddenly：
S1：Obtain data source.Obtaining social network user data can obtain from the API of existing social platform, or
Content in web page is captured by web crawlers.
S2：Extract association attributes.In view of potential user participate in topic main cause include individual subscriber characteristic attribute with
And the impact of user outside social attribute, association attributes will be extracted in terms of the two.And the information to user cooks isochronous surface
Change is processed,
S3：Set up model.The data that user property is carried out based on Cloud transform are fitted, obtain to be constructed after Normal Cloud
Higher-dimension cloud, the number of higher-dimension cloud is the neuron number of hidden layer in RBF neural, and its parameter is hidden layer excitation letter
Several cluster centre and bandwidth.RBF neural is trained by determining after parameter, obtains predicting that the user whether can
Participate in the forecast model of topic.
S4：Prediction and analysis process.By predicting following topic trend trend using exponential smoothing, prediction is drawn
Much-talked-about topic participates in the time series (y of number1, y2..., yn) do Three-exponential Smoothing calculating, you can fit much-talked-about topic
Temperature Long-term change trend, so as to carry out the prediction to subsequent time period.
Above-mentioned steps S1 obtain data source, and the particular content for extracting association attributes is that the user under certain hotspot topic participates in
Behavioral data and user relationship data.User's participative behavior data include time, the participating user that the topic is forwarded and comments on
Personal information and historical behavior data；User relationship data includes that participating in all beans vermicelli of the user under the topic and concern uses
Family, and their personal information etc..
Above-mentioned steps S2 extract association attributes.Main point of following 2 steps.
S21：Extract potential user's personal characteristics attribute.The personal characteristics attribute of potential user mainly includes 1. potential user
Whether it is any active ues isActivity (vi)；2. potential user viLabel in whether include and much-talked-about topic identical is crucial
Word isSameTag (Vi)；3. the quantity countOfHF (v of topic is participated in the concern user of potential useri)；4. potential user
The topic drive inf (v of concern useri)；The present invention is by more than about the unique characteristics attribute x of potential userikUnification
Form is described, and represents potential user viK-th attribute.
S22：Extract potential user outside social attribute.The outside social attribute of potential user mainly includes 1. potential user
viConcern user whether be certification user isVip (vi,vj)；2. potential user viConcern user whether be leader of opinion
isLeader(vi,vj)；3. potential user viIdentical isSameS (the v of user's sex are paid close attention to iti,vj)；4. potential user viWith it
Concern user sex difference isDifS (vi,vj)；5. potential user viIdentical isSameL (the v of user locations are paid close attention to iti,vj)；⑥
Potential user viDifferent isDifL (the v of user locations are paid close attention to from iti,vj)；Meanwhile, whether potential user can participate in the much-talked-about topic
It is also relevant with the corporations' influence power residing for it, therefore its team's attribute is defined for groupInf (vi,Cm), i.e., residing for potential user
CmWhether corporations are corporations interested in the much-talked-about topic.The present invention is by more than about the unique characteristics attribute of potential user
Use xikUnified Form description, represent potential user viK-th attribute.
Above-mentioned steps S3 extract association attributes.Main point of following 4 steps.
S31：User's unique characteristics attribute and user outside social attribute are respectively adopted by Maximum Approach and carry out Cloud transform, cloud
Conversion carries out mathematic(al) manipulation to any irregular data distribution, and it can carry out the fuzzy clustering of soft classification to sample point, make
It becomes the superposition of several different clouds.
S311：Frequency distribution function f (x) of one of user property X is given, according to the actual frequency of X property values point
Cloth situation, can automatically generate the different cloud C (E of some granularitiesxi, Eni, Hei) superposition, wherein each Yun Jun represent one from
Scattered, qualitatively concept, by continuous property value discrete concept is converted to, and can be expressed asWherein aiFor range coefficient；N is to generate discrete concept after conversion
Number.
S312：The crest value position of data distribution function f (x) is found, by the center of gravity position that its attribute value definition is cloud
Put, that is, expect Exi(i=1,2 ..., n), calculate for be fitted f (x), with ExiFor the entropy of desired cloud model, cloud mould is calculated
The distribution function f of typei(x)。
S313：Data distribution f of known cloud model is deducted from f (x)iX (), obtains new data distribution function f ' (x),
And repeat step S312 and S313 on this basis, obtain multiple data distribution functions based on cloud.
S314：According to the distribution letter of known f (x), error of fitting function f ' (x) for finally obtaining and each cloud model
Number, calculates 3 characteristic values based on the qualitativing concept of cloud model, and expectation, entropy and super entropy.
S32：By by the property value of Cloud transform in combination with RBF neural, so that it is determined that cloud model hidden layer god
Jing is first.Changed according to Peak Intensity Method cloud, for every dimension attribute X of input layer, ni fitting Normal Cloud can be obtained.According to higher-dimension cloud
Theoretical and n dimension Normal Cloud Generators construct n and tie up Normal Cloud as the neuron of RBF neural hidden layer, can obtain (n1
×n2×...×nn) individual n dimensions cloud model, i.e., n hidden layer node, the corresponding qualitative cluster of each hidden layer neuron is generally
Read, it is to avoid the cluster process in RBF neural, the qualitativing concept can be with three groups of numerical characteristics
To describe.
S33：One X condition cloud generator is actually based on the n dimension improved hidden layer neurons of cloud model, can will be defeated
The n-dimensional vector for entering is converted to the uncertain numerical value of one group of random distribution, although these numerical value are unequal each other, meets one
Stable distribution, thus from final output value of the desired value as RBF neural hidden layer neuron is wherein taken, this meets people
The characteristics of certainty result is derived from uncertainty in class cognition.
For the input vector x drawn in S32 stepsiCloud model neuron By with
Lower formula is converted to μt：
Wherein, ri1, ri2..., rik, i ∈ [1, n] are often to tie up k normal random number of generation by entropy and super entropy, and cloud model is hidden
Final output value containing layer neuron is：
S34：The output valve of hidden layer neuron can be obtained by S33 steps, hidden layer is to output layer in RBF neural
Between be a kind of perceptron algorithm model, and be made up of linear function due to exporting node layer, therefore asked using least square method
The weights of solution connection.
Above-mentioned steps S4 are predicted and analyzed.Main point of following 2 steps.
S41：The time series of acquisition is done into Three-exponential Smoothing conversion.
Three times smoothing model is
WhereinFor a smooth value,For secondary smooth value,For three smooth values.
S42：When time series embodies conic section trend, that is, set up conic section correction model
Wherein t is current time；L is the time difference of prediction time and current time；For the prediction of subsequent time
Value；at, bt, ctFor conic section correction factor.Its computing formula is as follows
The model is Nonlinear Prediction Models, can embody the variation tendency of sequential, and the development for predicting topic temperature becomes
Gesture.
The present invention is divided user according to the history participative behavior of user using the interactive data of much-talked-about topic in social networks
For participating user and potential user, using based on the improved RBF neural of cloud model each stage of topic potential user is predicted
Behavior, i.e. whether the next stage potential user in topic life cycle can forward or comment under the topic, and by latent
The future trend of topic development can be held in the prediction of user behavior.
The above embodiment is interpreted as being merely to illustrate the present invention rather than limits the scope of the invention.
After the content of the record for having read the present invention, technical staff can make various changes or modifications to the present invention, these equivalent changes
Change and modification equally falls into the scope of the claims in the present invention.
Claims (6)
1. a kind of improved RBF neural much-talked-about topic user participative behavior Forecasting Methodology, it is characterised in that including following step
Suddenly：
S1：Obtain from the API of existing social platform, or the content obtaining captured by web crawlers in web page is social
Network user data；
S2：The step of extracting association attributes：Participating in topic main cause in view of potential user includes individual subscriber characteristic attribute
And the impact of user outside social attribute, association attributes will be extracted in terms of the two；And the information time of doing to user cuts
Pieceization process；
S3：The step of setting up model：The data that user property is carried out based on Cloud transform are fitted, obtain to be built after Normal Cloud
Go out higher-dimension cloud, the number of higher-dimension cloud is the neuron number of hidden layer in RBF neural, its parameter is hidden layer excitation
The cluster centre and bandwidth of function, is trained by determining after parameter to RBF neural, obtains whether predicting the user
The forecast model of topic can be participated in；
S4：Prediction and analysis process steps：By predicting following topic trend trend using exponential smoothing, prediction is drawn
Much-talked-about topic participates in the time series (y of number1, y2..., yn) do Three-exponential Smoothing calculating, you can fit much-talked-about topic
Temperature Long-term change trend, so as to carry out the prediction to subsequent time period.
2. improved RBF neural much-talked-about topic user participative behavior Forecasting Methodology according to claim 1, its feature exists
In it is the user's participative behavior under certain hotspot topic that step S1 obtains the particular content of social network user data
Data and user relationship data；User's participative behavior data include the time that the topic is forwarded and comment on, participating user it is individual
People's information and historical behavior data；User relationship data includes participating in all beans vermicelli of the user under the topic and concern user,
And their personal information.
3. improved RBF neural much-talked-about topic user participative behavior Forecasting Methodology according to claim 1 or claim 2, it is special
Levy and be, step S2 extracts association attributes according to individual subscriber characteristic attribute, mainly includes：
Extract potential user's personal characteristics attribute：The personal characteristics attribute of potential user mainly includes 1. whether potential user is living
Jump user isActivity (vi)；2. potential user viLabel in whether include and much-talked-about topic identical keyword
isSameTag(Vi)；3. the quantity countOfHF (v of topic is participated in the concern user of potential useri)；4. the pass of potential user
The topic drive inf (v of note useri)；About the unique characteristics attribute x of potential user by more thanikUnified Form description,
Represent potential user viK-th attribute.
4. improved RBF neural much-talked-about topic user participative behavior Forecasting Methodology according to claim 3, its feature exists
In step S2 extracts association attributes according to user outside social attribute, mainly includes：The outside social attribute of potential user
It is main to include 1. potential user viConcern user whether be certification user isVip (vi,vj)；2. potential user viConcern use
Whether family is leader of opinion isLeader (vi,vj)；3. potential user viIdentical isSameS (the v of user's sex are paid close attention to iti,vj)；
4. potential user viDifferent isDifS (the v of user's sex are paid close attention to from iti,vj)；5. potential user viIt is identical user locations to be paid close attention to it
isSameL(vi,vj)；6. potential user viDifferent isDifL (the v of user locations are paid close attention to from iti,vj)；Meanwhile, whether potential user
The much-talked-about topic can be participated in also relevant with the corporations' influence power residing for it, therefore define its team's attribute for groupInf (vi,
Cm), i.e. C residing for potential usermWhether corporations are corporations interested in the much-talked-about topic；The relevant potential user by more than
Unique characteristics attribute xikUnified Form description, represent potential user viK-th attribute.
5. improved RBF neural much-talked-about topic user participative behavior Forecasting Methodology according to claim 4, its feature exists
In mainly dividing following 4 steps the step of step S3 sets up model：
S31：User's unique characteristics attribute and user outside social attribute are respectively adopted by Maximum Approach and carry out Cloud transform, Cloud transform
Mathematic(al) manipulation is carried out to any irregular data distribution, it can carry out the fuzzy clustering of soft classification to sample point so as into
For the superposition of several different clouds；
S32：By by the property value of Cloud transform in combination with RBF neural, so that it is determined that the hidden layer neuron of cloud model；
Changed according to Peak Intensity Method cloud, for every dimension attribute X of input layer, n can be obtainediIndividual fitting Normal Cloud, according to higher-dimension Clouds theory
N is constructed with n dimension Normal Cloud Generators and tie up Normal Cloud as the neuron of RBF neural hidden layer, can obtain (n1×n2
×...×nn) individual n dimensions cloud model, i.e., n hidden layer node；
S33：From based on the improved hidden layer neuron of n dimension cloud models desired value is taken as RBF neural hidden layer nerve
The final output value of unit；
S34：Hidden layer is to being a kind of perceptron algorithm model between output layer in RBF neural, and due to exporting node layer
It is made up of linear function, the weights of connection is solved using least square method.
6. improved RBF neural much-talked-about topic user participative behavior Forecasting Methodology according to claim 5, its feature exists
In step S4 predicts and analyze that process steps include：
S41：The time series of acquisition is done into Three-exponential Smoothing conversion；
S42：When time series embodies conic section trend, that is, conic section correction model is set up, the model is non-linear
Forecast model, can embody the variation tendency of sequential, predict the development trend of topic temperature.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611189974.4A CN106651030B (en) | 2016-12-21 | 2016-12-21 | Improved RBF neural network hot topic user participation behavior prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611189974.4A CN106651030B (en) | 2016-12-21 | 2016-12-21 | Improved RBF neural network hot topic user participation behavior prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106651030A true CN106651030A (en) | 2017-05-10 |
| CN106651030B CN106651030B (en) | 2020-08-04 |
Family
ID=58835261
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611189974.4A Active CN106651030B (en) | 2016-12-21 | 2016-12-21 | Improved RBF neural network hot topic user participation behavior prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106651030B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107391712A (en) * | 2017-07-28 | 2017-11-24 | 王亚迪 | A kind of network public opinion trend prediction analysis method |
| CN107944610A (en) * | 2017-11-17 | 2018-04-20 | 平安科技（深圳）有限公司 | Predicted events measure of popularity, server and computer-readable recording medium |
| CN108229731A (en) * | 2017-12-20 | 2018-06-29 | 重庆邮电大学 | The user's behavior prediction system and method that more message mutually influence under a kind of much-talked-about topic |
| CN108564191A (en) * | 2017-12-20 | 2018-09-21 | 重庆邮电大学 | A kind of much-talked-about topic Popularity prediction system and method based on chaology |
| CN108984694A (en) * | 2018-07-04 | 2018-12-11 | 龙马智芯（珠海横琴）科技有限公司 | The processing method and processing device of webpage, storage medium, electronic device |
| CN109214562A (en) * | 2018-08-24 | 2019-01-15 | 国网山东省电力公司电力科学研究院 | A kind of power grid scientific research hotspot prediction and method for pushing based on RNN |
| CN109446236A (en) * | 2018-10-18 | 2019-03-08 | 太原理工大学 | Cement-particle size distribution forecasting method based on random distribution |
| CN110909254A (en) * | 2019-10-31 | 2020-03-24 | 中山大学 | Method and system for predicting question popularity of question-answering community based on deep learning model |
| CN111798018A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Behavior prediction method, behavior prediction device, storage medium and electronic equipment |
| CN112270568A (en) * | 2020-11-02 | 2021-01-26 | 重庆邮电大学 | Social e-commerce platform marketing activity order rate prediction method facing hidden information |
| 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 |
| CN112749335A (en) * | 2020-09-08 | 2021-05-04 | 腾讯科技（深圳）有限公司 | Life cycle state prediction method and device, computer equipment and storage medium |
| CN114091443A (en) * | 2021-10-28 | 2022-02-25 | 广州大学 | Network information propagation index system construction and quantitative evaluation method and system based on deep learning |
| CN115934808B (en) * | 2023-03-02 | 2023-05-16 | 中国电子科技集团公司第三十研究所 | Network public opinion early warning method integrated with association analysis and storm suppression mechanism |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279671A (en) * | 2013-06-03 | 2013-09-04 | 南京大学 | Urban water disaster risk prediction method based on RBF (radial basis function) neural network-cloud model |
| CN104408108A (en) * | 2014-11-18 | 2015-03-11 | 重庆邮电大学 | Hot topic group influence analysis system and method based on grey system theory |
| CN104951836A (en) * | 2014-03-25 | 2015-09-30 | 上海市玻森数据科技有限公司 | Posting predication system based on nerual network technique |
| US20160042277A1 (en) * | 2014-08-05 | 2016-02-11 | Hewlett-Packard Development Company, Lp | Social action and social tie prediction |
| CN105608200A (en) * | 2015-12-28 | 2016-05-25 | 湖南蚁坊软件有限公司 | Network public opinion tendency prediction analysis method |
-
2016
- 2016-12-21 CN CN201611189974.4A patent/CN106651030B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279671A (en) * | 2013-06-03 | 2013-09-04 | 南京大学 | Urban water disaster risk prediction method based on RBF (radial basis function) neural network-cloud model |
| CN104951836A (en) * | 2014-03-25 | 2015-09-30 | 上海市玻森数据科技有限公司 | Posting predication system based on nerual network technique |
| US20160042277A1 (en) * | 2014-08-05 | 2016-02-11 | Hewlett-Packard Development Company, Lp | Social action and social tie prediction |
| CN104408108A (en) * | 2014-11-18 | 2015-03-11 | 重庆邮电大学 | Hot topic group influence analysis system and method based on grey system theory |
| CN105608200A (en) * | 2015-12-28 | 2016-05-25 | 湖南蚁坊软件有限公司 | Network public opinion tendency prediction analysis method |
Non-Patent Citations (1)
| Title |
|---|
| 魏德志 等: "基于混沌理论和改进径向基函数神经网络的网络舆情预测方法", 《物理学报》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107391712A (en) * | 2017-07-28 | 2017-11-24 | 王亚迪 | A kind of network public opinion trend prediction analysis method |
| CN107944610A (en) * | 2017-11-17 | 2018-04-20 | 平安科技（深圳）有限公司 | Predicted events measure of popularity, server and computer-readable recording medium |
| CN108229731B (en) * | 2017-12-20 | 2021-03-23 | 重庆邮电大学 | System and method for predicting user behavior under hot topics under multi-message interaction |
| CN108229731A (en) * | 2017-12-20 | 2018-06-29 | 重庆邮电大学 | The user's behavior prediction system and method that more message mutually influence under a kind of much-talked-about topic |
| CN108564191A (en) * | 2017-12-20 | 2018-09-21 | 重庆邮电大学 | A kind of much-talked-about topic Popularity prediction system and method based on chaology |
| CN108984694A (en) * | 2018-07-04 | 2018-12-11 | 龙马智芯（珠海横琴）科技有限公司 | The processing method and processing device of webpage, storage medium, electronic device |
| CN109214562A (en) * | 2018-08-24 | 2019-01-15 | 国网山东省电力公司电力科学研究院 | A kind of power grid scientific research hotspot prediction and method for pushing based on RNN |
| CN109446236A (en) * | 2018-10-18 | 2019-03-08 | 太原理工大学 | Cement-particle size distribution forecasting method based on random distribution |
| CN109446236B (en) * | 2018-10-18 | 2021-12-21 | 太原理工大学 | Cement particle size distribution prediction method based on random distribution |
| CN111798018A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Behavior prediction method, behavior prediction device, storage medium and electronic equipment |
| CN110909254B (en) * | 2019-10-31 | 2022-05-03 | 中山大学 | Method and system for predicting question popularity of question-answering community based on deep learning model |
| CN110909254A (en) * | 2019-10-31 | 2020-03-24 | 中山大学 | Method and system for predicting question popularity of question-answering community based on deep learning model |
| CN112749335A (en) * | 2020-09-08 | 2021-05-04 | 腾讯科技（深圳）有限公司 | Life cycle state prediction method and device, computer equipment and storage medium |
| CN112749335B (en) * | 2020-09-08 | 2023-11-07 | 腾讯科技（深圳）有限公司 | Lifecycle state prediction method, lifecycle state prediction apparatus, computer device, and storage medium |
| CN112270568A (en) * | 2020-11-02 | 2021-01-26 | 重庆邮电大学 | Social e-commerce platform marketing activity order rate prediction method facing hidden information |
| CN112270568B (en) * | 2020-11-02 | 2022-07-12 | 重庆邮电大学 | Order rate prediction method for social e-commerce platform marketing campaign facing hidden information |
| 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 |
| CN112418525B (en) * | 2020-11-24 | 2022-06-10 | 重庆邮电大学 | Method and device for predicting social topic group behaviors and computer storage medium |
| CN114091443A (en) * | 2021-10-28 | 2022-02-25 | 广州大学 | Network information propagation index system construction and quantitative evaluation method and system based on deep learning |
| CN115934808B (en) * | 2023-03-02 | 2023-05-16 | 中国电子科技集团公司第三十研究所 | Network public opinion early warning method integrated with association analysis and storm suppression mechanism |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106651030B (en) | 2020-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106651030A (en) | Method for predicting user participation behavior of hot topic by improved RBF neural network | |
| Liu et al. | C-RBFNN: A user retweet behavior prediction method for hotspot topics based on improved RBF neural network | |
| Feng et al. | An intelligent decision support system for fuzzy comprehensive evaluation of urban development | |
| CN108229731B (en) | System and method for predicting user behavior under hot topics under multi-message interaction | |
| CN109523021A (en) | A kind of dynamic network Structure Prediction Methods based on long memory network in short-term | |
| Otto et al. | Dynamic perspectives on social characteristics and sustainability in online community networks | |
| CN105243448A (en) | Method and device for predicting evolution trend of internet public opinion | |
| Xiao et al. | User behavior prediction of social hotspots based on multimessage interaction and neural network | |
| Li | Network traffic prediction of the optimized BP neural network based on Glowworm Swarm Algorithm | |
| Wang | Fuzzy comprehensive evaluation of physical education based on high dimensional data mining | |
| Dong et al. | The damped oscillator model (DOM) and its application in the prediction of emotion development of online public opinions | |
| Gao et al. | Meta-circuit machine: Inferencing human collaborative relationships in heterogeneous information networks | |
| Yin et al. | Research on weibo public opinion prediction using improved genetic algorithm based BP neural networks | |
| Chen et al. | A forecasting system of micro-blog public opinion based on artificial neural network | |
| Meng et al. | POI recommendation for occasional groups Based on hybrid graph neural networks | |
| Phillips et al. | Extracting social structure from darkweb forums | |
| Jiang et al. | Prediction model of port throughput based on game theory and multimedia Bayesian regression | |
| CN113850446B (en) | Information diffusion prediction method integrating space-time attention and heterogeneous graph convolution network | |
| Li et al. | Assessing adaptive thermal comfort using artificial neural networks in naturally-ventilated buildings | |
| CN105761152A (en) | Topic participation prediction method based on triadic group in social network | |
| Wang et al. | Predicting information diffusion cascades using graph attention networks | |
| Hou et al. | Research of micro-blog diffusion effect based on analysis of retweet behavior | |
| Çağil et al. | An intelligent simulation model of online consumer behavior | |
| Zhang et al. | Color matching design simulation platform based on collaborative collective intelligence | |
| Yang | Analysis of Network Public Opinion in New Media Based on BP Neural Network Algorithm |
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 |