CN105447128A - Method for predicting spread range of microblog public opinions - Google Patents

Abstract

The invention relates to the field of social network modeling and analysis, in particular to a method for predicting the spread range of microblog public opinion. Proceed in the following order: 1) Construct the propagation network model of the microblog system; 2) Select the sentinel nodes in the microblog propagation network to judge the coverage of public opinion; 3) Use the sentinel monitoring nodes to establish the prediction model of the spread range of microblog public opinion; 4) Carry out empirical statistical analysis of event public opinion in the actual microblog network, and determine the key parameters in the prediction model. The present invention conducts empirical statistical experiments in a social network of 500-1000 microblog nodes, and obtains important parameters of a prediction model. Then write a network microblog crawling program to analyze and count the social network data of 5000-10000 nodes, and use this scale of microblog network to verify the accuracy of the prediction method. The experimental results show that the prediction accuracy is about 83.2%. .

Description

A Method for Predicting the Spread Range of Microblog Public Opinion

technical field

The invention relates to the field of social network modeling and analysis, in particular to a method for predicting the spread range of microblog public opinion.

Background technique

Weibo has become one of the most important new media platforms in modern society. Compared with traditional media, Weibo has the characteristics of timeliness, fragmentation, freedom and openness, and popularity. But anyone can use Weibo to post bad opinions and comments, and after being forwarded and commented by many people, they will quickly spread to the entire social network. Some deceptive remarks can cause damage to social security, and seriously cause social group incidents. Therefore, the relevant government departments must analyze, monitor and predict the public opinion information in Weibo, and make preparations for further management and control.

The existing Internet public opinion information monitoring and analysis mainly focus on two issues: one is to solve the problem of manual processing of massive information, and propose some automatic public opinion analysis systems designed by using computer text analysis and machine learning methods, In order to reduce the manual labor in the process of network public opinion monitoring; the second is to try to solve the problem of the accuracy of network public opinion discovery, and improve the accuracy of public opinion semantic mining in text by improving and optimizing text analysis, clustering algorithms and other methods.

After searching the literature of the existing technology, it is found that the Chinese patent publication number is: CN101661513B, and the patent name is: detection method of network hotspots and public opinion. This technical solution provides a detection method of network hotspots and public opinion in the field of network information processing , which can be applied to the detection and analysis of Weibo public opinion. By collecting microblog text information and comment information within a certain time range, and performing word segmentation and concept mapping processing on the text content of these information, the uncertainty of semantic concepts is eliminated, and finally the features that can reflect the text content are extracted. Then use these content feature data for clustering to form a number of information document collections containing different numbers. According to the number of information documents contained in each collection, it is determined whether it is a hot event in the network, and the information document collection of hot events is praised or criticized. Tendency analysis, so as to grasp the public opinion of netizens on the event, in order to detect Weibo public opinion.

Existing microblog monitoring and analysis methods focus on automated analysis and processing and judgment of public opinion information, ignoring the analysis of public opinion dissemination trends in the entire online social network, and unable to provide network public opinion control personnel with the extent to which public opinion has spread, that is, unable to judge The extent of the spread of public opinion on an event. The invention detects and analyzes microblog public opinion dissemination from the perspective of the overall social network, proposes a method for predicting the degree of microblog public opinion dissemination, and judges the situation of public opinion dissemination by monitoring the information of sentinel nodes.

Contents of the invention

The purpose of the present invention is to solve the above problems, and provide a method for predicting the spread range of microblog public opinion. Through the microblog prediction method, the actual statistical data is used to establish a nonlinear model, and the state of the sentinel node is monitored according to the nature of the public opinion event to determine the microblog. Coverage of public opinion, and provide accurate quantitative data of public opinion dissemination to network public opinion managers.

The technical solution adopted by the present invention to solve the above problems is:

A method for predicting the spread range of microblog public opinion is carried out in the following order:

1) Construct the communication network model of the microblog system: consider each microblog user as a node, and establish links between nodes according to the fans, followers and friends of the microblog to form a complex online social network model; public opinion The scope of dissemination is the coverage of public opinion news;

2) Select sentinel nodes to judge the coverage of public opinion in the Weibo dissemination network;

3) Use sentinel monitoring nodes to establish a prediction model for the spread of microblog public opinion;

An empirical statistical analysis of event public opinion is carried out in the actual microblog network, and the key parameters in the prediction model are determined.

Preferably, the public opinion message coverage rate described in 1) is the ratio of the node set that has learned the message to the entire node set,

$\mathrm{<span\; class="notranslate">\; o</span>}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; V</span>}}<span\; class="notranslate">\; |</span>}{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; |</span>}<span\; class="notranslate">\; ,</span>$

In the formula represents the number of nodes, is the number of all nodes, note that all nodes refer to the total number of nodes within the range of valid users in the Weibo network;

The message dissemination process is a time series T={t1,t2,…,ti,ti+1,…}, and the information coverage rate at the monitoring time t _k is O _k , namely

Preferably, the predictive model in 3) transforms the problem of forecasting information coverage for sentinel nodes in the microblog network into _Vk Merge the events to predict _Ok , study the law between the node subset V _k and the coverage rate O, establish a prediction model, and realize the evaluation of the information coverage rate O _k by detecting the information of the sentinel nodes belonging to V _k ; Choose one of the nodes to spread influence.

Preferably, the sentinel nodes include opinion leader nodes, active nodes in the community, and inactive nodes.

Preferably, the node propagation influence is the product of the degree of the node and the average distance of the indirectly connected nodes, $I\left(i\right)=\frac{out\mathrm{deg}ree\left(i\right)\&times;\underset{j=0}{\overset{\mathrm{no}}{\&Sigma;}}{d}_{ij}}{cou\mathrm{no}t\left(i\right)},$

I(i) indicates the influence of node i, outdegree(i) is the out-degree of node, d _ij indicates the distance between node j indirectly connected with node i, count(i) indicates all other nodes indirectly connected with node i The number; finally establish a statistical method to establish the relationship model between node influence and information coverage

O(I)=f(I),

Taking O(I)=f(I) as the prediction basis, to detect whether several nodes propagate to a certain piece of information, in order to evaluate the information coverage rate, the propagation influence of node j is I _j , then after substitution, O( I _j ), abbreviated as O _j means the information coverage obtained by detecting node j;

S-curve is chosen as the basic model of regression analysis,

The beneficial effects of the present invention are:

The present invention conducts empirical statistical experiments in a social network of 500-1000 microblog nodes, and obtains important parameters of a prediction model. Then write a network microblog crawling program to analyze and count the social network data of 5000-10000 nodes, and use this scale of microblog network to verify the accuracy of the prediction method. The experimental results show that the prediction accuracy is about 83.2%. .

Description of drawings

Fig. 1 is the statistical information map of the present invention's influence small node as the source point;

Fig. 2 is a statistical information map of the present invention with large influential nodes as source points;

Fig. 3 is a statistical information diagram of the medium influence node as the source point of the present invention;

detailed description

Below in conjunction with accompanying drawing and specific embodiment the present invention is described in further detail:

As shown in Figure 1, Figure 2 and Figure 3, a method for predicting the spread range of microblog public opinion according to the present invention, the scope of demonstration is to select 587 students from four colleges of engineering in a certain university, involving 3 grades Current college students in 12 majors and 15 classes. After everyone registers on Sina Weibo, they form online social relationships in a natural way. After forming stable online relationships based on dormitories, friends, classmates, and community activities on campus, new relationships are not allowed to be added. In addition, only nodes within the scope of the university are considered, and other node relationships, such as high school classmates, relatives and friends, are ignored.

Using the Sina Weibo system as an information dissemination platform, random nodes are selected as information sources to release some homogeneous information, such as the promotion of training and learning and commercial promotion activities, and the release of information on college student activities. Only students are allowed to use Weibo to understand and disseminate information, and try to eliminate the interference of offline dissemination. Define a unique id for each test message, marked as M _i , and set a unique id for each student node, recorded as V _j , when students receive M _i to comment and forward normally, and send a message to a public mailbox at the same time e-mail with the titles M _i and V _j . Finally, the trajectory of message transmission is extracted in the email list. Each student’s information is a triplet <M _i , V _j , t _i >, where Mi is the information identification number, V _j is the user identification number, and t _i is the mail The reception time of , here approximates the message propagation arrival time.

In the demonstration, three kinds of influential nodes are used as the propagation source points: nodes with low influence, nodes with high influence and nodes with medium influence, as shown in Figure 1-3 respectively. The x-axis in the figure represents the node influence, and the y-axis represents the information coverage. Each time, 5 homogeneous message disseminations are selected for demonstration, and the error range of information coverage is determined. After observing Figure 1-3, it can be found that there is a certain nonlinear relationship between node influence and information coverage. Nodes with high influence correspond to low information coverage, while nodes with low influence correspond to high information coverage. This rule is consistent with the intuitive analysis in real society. We try to construct a nonlinear model through empirical data to establish the direct relationship between node influence and information coverage.

In Figure 1, nodes with little influence are used as the propagation source, forming a relatively smooth curve. Regression analysis can be used to fit formula (4) O(I _j ). In comparison, there are fewer medium-influence nodes monitored in the empirical evidence, and the interval of medium-influence nodes is relatively sparse.

In Figure 2, nodes with high influence are used as the source of propagation, and the area of nodes with medium influence is more sparse, but the error range is significantly reduced in the interval of nodes with large influence, which is because of the factor of using nodes with large influence as the source of propagation , the error fluctuation of the five empirical processes is small.

In Figure 3, medium-influential nodes are used as the propagation source, and the intervals of medium-influential nodes are no longer sparse, and the error fluctuations are small; the influential nodes show a decreasing trend, and the information coverage error fluctuations become larger; the small-influential nodes As the number of nodes increases, the error fluctuation does not change significantly.

Those skilled in the industry should understand that the present invention is not limited by the above-mentioned embodiments. What are described in the above-mentioned embodiments and the description only illustrate the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will also have Variations and improvements are possible, which fall within the scope of the claimed invention. The protection scope of the present invention is defined by the appended claims and their equivalents.

Claims (5)

1. a method for microblogging public sentiment spread scope prediction, is characterized in that: carry out successively according to following order:

1) build the communication network model of microblog system: each microblog users is considered as a node, set up the company limit between node according to the bean vermicelli of microblogging, concern and friend relation, form a complicated online community network model; Public sentiment spread scope and public sentiment message coverage rate;

2) in microblogging communication network, the sentinel node judging public sentiment coverage is selected;

3) sentry's monitoring node is utilized to set up the forecast model of microblogging public sentiment spread scope;

4) in actual micro blog network, real example statistical study is carried out to event public sentiment, and determine the key parameter in forecast model.

2. the method for a kind of microblogging public sentiment spread scope prediction according to claim 1, is characterized in that: the public sentiment message coverage rate 1) is known the node set of message and the ratio of whole node set,

$O=\frac{\left|\stackrel{\&OverBar;}{V}\right|}{\left|V\right|},$

In formula represent nodes, | V| is whole nodes, notices that whole node refers to the node total number in micro blog network within the scope of validated user;

Message propagation process is time series T={t1, t2 ..., ti, ti+1 ..., monitoring moment t _kinformation coverage be O _k, namely

3. the method for a kind of microblogging public sentiment spread scope prediction according to claim 1, is characterized in that: the forecast model 3) be the problem of micro blog network sentinel node information of forecasting coverage rate change into by the event be merged into is to predict O _k, research Node subsets V _kand the rule between coverage rate O, sets up forecast model, belong to V by detection _kthe information realization of sentinel node to information coverage O _kassessment; A node propagation effect power is selected in sentinel node.

4. the method for a kind of microblogging public sentiment spread scope prediction according to claim 1, is characterized in that: described sentinel node comprises live-vertex in leader of opinion's node, community, inactive node.

5. the method for a kind of microblogging public sentiment spread scope prediction according to claim 1, is characterized in that: described node propagation effect power is the degree of node and the product of indirect communication node mean distance, $I\left(i\right)=\frac{out\mathrm{deg}ree\left(i\right)\&times;\underset{j=0}{\overset{n}{\&Sigma;}}{d}_{ij}}{count\left(i\right)},$

I (i) represents the influence power of node i, the out-degree that outdegree (i) is node, d _ijrepresent the distance between the node j of node i indirect communication, count (i) represents the number of other all nodes of node i indirect communication; Finally set up and first set up the relational model between node influence power and information coverage by statistical method

O(I)＝f(I),

With O (I)=f (I), as basis for forecasting, detect some nodes and whether propagate into certain information, carry out appreciation information coverage rate with this, the propagation effect power of node j is I _j, then O (I is drawn after substituting into _j), be abbreviated as O _jthe information coverage that expression probe node j gets;

Select S curve as the basic model of regretional analysis,