KR20090001431A - Method and system for analyzing blog - Google Patents

Abstract

A blog analysis method and system are provided to determine the inclination of blog relation based on a grouping model and determine the inclination of a blog connection network based on the inclination of the determined blog relation. The inclination of plural first blog relations is determined on the basis of a grouping model. The inclination of a blog connection network is determined on the basis of the determined inclination of the first blog relations. A blog analysis system inquires a blog user of plural second blog relations about the inclination and receives the answer. The blog analysis system learns on the basis of the answer and the property data of the second blog relations. The blog analysis system produces the grouping model according to a result of the learning.

Description

Blog Analytics Method and System {METHOD AND SYSTEM FOR ANALYZING BLOG}

1 is a block diagram illustrating a blog analysis system according to an exemplary embodiment of the present invention.

2 is a diagram illustrating a blog network.

3 is a diagram illustrating a blog network.

4 is a block diagram of the blog analysis server shown in FIG. 1.

5 is a flowchart illustrating a blog analysis method according to an embodiment of the present invention.

FIG. 6 is an example of a questionnaire for determining a propensity to blog relationship illustrated in FIG. 5.

FIG. 7 is an example of a classification model for determining acquaintance-oriented tendencies shown in FIG. 5.

FIG. 8 is an example of a classification model for determining information-oriented propensity shown in FIG. 5.

9 is a blog network illustrated to determine the propensity score of the blog network in accordance with the blog analysis method according to an embodiment of the present invention.

10 is an exemplary diagram of a blog network extracted for the accuracy measurement of the propensity determination of the blog network.

<Description of Drawing>

100: blog server, 130: blog analytics server,

132: classification model generation module, 134: relationship propensity determination module,

136: network propensity determination module, 160: database,

200: communication network, 300: user terminal

The present invention relates to a blog analysis method and system.

A social network is a network that expresses the relationship structure of a society as a whole by using relationship information between members in a society. It is called social network analysis to study the relationship structure and derive the unique characteristics of the society. In the past, it was difficult to obtain data indicating the relationship between members in social networks. Therefore, the focus of the existing research was to model the structure and characteristics of social networks without specific data on the relationship between members.

Due to the development of the Internet, these social networks have appeared online. The most distinctive feature of the online social network is that, unlike the existing social network, it stores specific information that can explain the relationship between members and information exchanged by the relationship in a database. Due to these characteristics, recent researches have been conducted to analyze online social networks.

A blog is a kind of personal web site where users can store their articles online, and the blog world is a representative online social network of such blogs. Each blog user can have a relationship with someone else's blog, and these relationships form an online social network. Hereinafter, a social network consisting of a relationship between a blog and a blog is called a blog network.

The goal of an operator or a company running a blogging world is to activate the blogging world as a result of the increased activity of the blog. The propensity that exists in blog networks, such as the propensity to send and receive information mainly for the purpose of acquaintance, mainly affects the activation of blog networks. What kind of blog network has a tendency and how much is used as an important judgment ground for establishing an activation policy for the blog network. However, although various research techniques for analyzing online social networks have been suggested, there are no studies on determining the propensity of blog networks.

Therefore, the technical problem to be achieved by the present invention is to provide a blog analysis method and system that can determine the propensity and extent of the blog network, which is an important basis for establishing the activation policy of the blog world.

According to an embodiment of the present invention, a blog analysis method includes determining a propensity of a plurality of first blog relationships based on a classification model, and based on the propensity of the first blog relationships. Determining the propensity of.

Asking a blog user of a plurality of second blog relationships related to the inclination and receiving a response, learning based on the response and attribute information of the second blog relationship, and according to the result of the learning step. The method may further include generating a classification model.

And assigning a class label divided into a plurality of levels to the propensity of the second blog relationship, wherein the determining of the first blog relationship propensity corresponds to a score assigned to the level corresponding to the propensity score of the first blog relationship. It may include the step of.

The attribute information may include at least one of the number of comments, the number of gifts, the number of posts in the guestbook, the number of scraps, and the number of visits.

The classification model may be generated using any one of a decision tree, a neural network, a Bayesian classification, and a genetic algorithm.

The determining of the first blog relationship propensity may include calculating the propensity score of the first blog relationship by applying attribute information of the first blog relationship to the classification model.

The determining of the blog network propensity comprises: calculating the propensity score of the blog network by adding up the propensity scores of the first blog relationship and dividing the number of blogs included in the blog network and the number of the first blog relationships. It may include.

The propensity may include at least one of acquaintance-oriented propensity and information-oriented propensity.

According to another aspect of the present invention, a method for analyzing a blog includes: asking a blog user of a plurality of first blog relationships related to a propensity of the first blog relationship and receiving a response, the response and attribute information of the first blog relationship Learning based on, generating a classification model according to the result of the learning step, and determining propensity of a plurality of second blog relationships based on the classification model.

And assigning a class label divided into a plurality of levels to the propensity of the first blog relationship, wherein the determining of the second blog relationship propensity corresponds to a score assigned to the level corresponding to a propensity score of the second blog relationship. It may include the step of.

The determining of the second blog relationship propensity may include calculating propensity scores of the second blog relationships by applying attribute information of the second blog relationship to the classification model.

A computer readable medium according to another aspect of the present invention records a program for causing a computer to execute any of the above methods.

A blog analysis system according to another aspect of the present invention is a relationship propensity determining module for determining propensity of a plurality of first blog relationships based on a classification model, and determining a propensity of a blog network based on the propensity of the first blog relationships. It includes a network propensity determination module to.

Receiving a response according to a query related to the propensity from a plurality of blog users of a plurality of second blog relationships, learning based on the response and attribute information of the second blog relationship, and classifying to generate the classification model according to the learning result. It may further include a model generation module.

The classification model generation module assigns a class label divided into a plurality of levels to the propensity of the second blog relationship, and the relationship propensity determining module may correspond to the propensity score of the first blog relationship with a score assigned to the level. have.

The relationship propensity determination module may calculate the propensity score of the first blog relationship by applying attribute information of the first blog relationship to the classification model.

The network propensity determination module may calculate the propensity score of the blog network by adding up the propensity scores of the first blog relations and dividing by any one of the number of blogs included in the blog network and the number of the first blog relations. .

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention.

First, referring to FIGS. 1 to 4, a blog analysis system and a blog connection network targeted for analysis according to an embodiment of the present invention will be described in detail.

1 is a block diagram illustrating a blog analysis system according to an embodiment of the present invention, FIG. 2 is a diagram illustrating a blog network, FIG. 3 is a diagram illustrating a blog network, and FIG. 4 is a diagram of FIG. A block diagram of the illustrated blog analysis server.

Referring to FIG. 1, a blog analysis system according to an embodiment of the present invention includes a blog server 100, a blog analysis server 130, and a database 160, and includes a plurality of user terminals through a communication network 200 ( 300).

The blog server 100 provides a blog service to a user. A blog is a kind of personal web site that allows users to post freely according to their interests. The blog server 100 provides a blog to the user and allows the user to create a document (or post) including text, images, videos, music, and the like. Make sure that you write to your blog. The blog server 100 transmits and displays its own blog document and other user's blog document to the corresponding user terminal 300 according to a user's request.

In addition, the blog server 100 records the actions that the user has performed on his blog and other users' blogs. That is, the blog server 100 may provide information on all situations occurring in a blog, for example, a history of a user's visit to another user's blog, a history of a user's comment on another user's blog document, and a user present to another user. The blog data including the details of the items (items) given and the history related to the articles written in the guest book are recorded in the database 160.

The database 160 receives and stores blog documents and blog data from the blog server 100, and provides corresponding documents and data according to requests from the blog server 100 and the blog analysis server 130. Text, images, videos, etc. according to the types of blog documents can be stored in a separate database, and blog data can be stored in a separate database.

The communication network 200 may include a telephone network as well as a data communication network including a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), the Internet, and the like. It can be used, and it does not matter wired and wireless, and it does not matter which communication method is used.

The user terminal 300 connects to and receives information from the blog server 100 and the blog analysis server 130 through the communication network 200, and displays the blog document from the blog server 100 in a browser (not shown). do. The user terminal 300 is provided with memory means such as a laptop computer, a workstation, a palmtop computer, a personal digital assistant (PDA), a web pad, a mobile communication terminal, as well as a desktop computer, and includes a microprocessor. It can be made of a terminal equipped with a computing power.

The user can write a document on his blog through the user terminal 300, can view his own document, create a link to another person's blog document on his blog, or scrap another person's blog document with his own blog. Can be imported. Users can also view other people's blog documents and write replies to other people's blog documents.

Blog users can establish a relationship between their blog and others' blogs. This relationship forms the online social network, or blog network, in the blog world. Referring to FIG. 2, a rounded rectangle represented by BA to BH represents a blog of each user, and a small rectangle represented by da to dh means a document stored in each blog. The solid arrows indicate that you have linked or scrapped a document on one user blog to another user blog. As a result, the document dc of the blog BD is identical to the article dc of the blog BA. The solid line between the blog and the blog indicates that there is a relationship between the blog and the blog, for example, that the blog BA and the blog BD are related to each other. Referring to FIG. 3, the blog network GB may be more simply represented as a graph. Circles represent each blog, and solid lines represent relationships between blogs. The part G1 enclosed by the dotted lines represents a partial network of the blog network GB. Of course, the blog network may also be a partial network of the entire blog network. Hereinafter, the blog network will be referred to without distinguishing the entire blog network or partial network.

Referring to FIG. 4, the blog analysis server 130 includes a classification model generation module 132, a relationship propensity determination module 134, and a network propensity determination module 136. The classification model generation module 132 generates a decision tree based on the relationship propensity related data and attribute information returned from the blog user, and the relationship propensity determination module 133 attributes the blog relations stored in the database 160. The degree of propensity of the blog relationship is determined based on the information and the decision tree, and the network propensity determination module 136 determines the degree of propensity of the blog network based on the degree of propensity of the blog relationship included in the blog network.

The blog network may have at least two different dispositions depending on the purpose of the article (or document) existing therein. For example, in some blog networks, users write and collect informative articles related to stocks, travel, and so on. The propensity to send and receive texts for the purpose of disseminating and collecting information is defined as information-oriented propensity. On the other hand, in some blog networks, users write articles to maintain and develop relationships with each other, such as regards and posts. As described above, the propensity to send and receive texts among the users for the purpose of acquaintance is defined as acquaintance-oriented propensity. The cyber gifts exchanged in the blog network are more actively traded in the blog network, which has a strong tendency toward friendliness, and the blog network, which has a strong tendency toward information, can be spread and read a lot of informational texts. In general, these two inclinations are mixed within an entire blog network. Of course, blog networks can have a variety of inclinations. The blog analysis server 130 determines the degree with respect to the propensity of any given blog network.

The blog analysis server 130 may be implemented by being included in the blog server 100. The blog analysis server 130 may be implemented as a separate system without being connected to the blog server 100 and the blog database 160 to receive blog data and perform a corresponding function. You may.

Next, a blog analysis method for determining the degree of propensity of the blog network will be described in detail with reference to FIGS. 5 to 9.

FIG. 5 is a flowchart illustrating a blog analysis method according to an exemplary embodiment of the present invention, and FIG. 6 is an example of a questionnaire for determining a propensity to blog relationship illustrated in FIG. 5. FIG. 7 is an example of a classification model for determination of acquaintance-oriented tendencies shown in FIG. 5, FIG. 8 is an example of a classification model for determination of information-oriented tendencies shown in FIG. 5, and FIG. 9 is an embodiment of the present invention. The blog network is illustrated to determine the propensity score of the blog network according to the blog analysis method according to the present invention.

First, the blog analysis server 130 determines the degree of inclination with respect to the relationship between blogs which are the basic units of the blog network, and uses the result to determine the degree of inclination of the blog network. Referring to FIG. 5, in the blog analysis method according to the embodiment of the present invention, a survey step (S410), a class labeling step (S420), a relationship attribute generation step (S430) for classification, and a classification model generation step (S440) And, the network propensity determination step (S450). Each of these steps will be described in more detail. Hereinafter, the propensity of the blog network will be described by using the acquaintance-oriented propensity and information-oriented propensity described above as an example.

Poll

In order to determine the propensity for the relationship between the two blogs, we first survey the blog users. The relationship propensity between the two blogs can be determined by the blog user himself. Therefore, a question related to the propensity is asked through a questionnaire about the relationship that the blog user directly participates in. In other words, the two blog users who are directly involved in each blog relationship are presented with the kinds of articles deemed to have a particular tendency, and how often each user sees the posts in the blogs of other users. The survey results are then quantified to determine the degree of inclination of each relationship. This makes it possible to objectively determine the degree of inclination of the blog relationship compared to directly inquiring the degree of inclination of the blog relationship with the user.

The questionnaire includes questions that ask about information-oriented disposition and questions about acquaintance-oriented disposition. For example, as shown in FIG. 6, a questionnaire for determining an acquaintance-oriented tendency and a questionnaire for determining an information-oriented tendency may be presented to a blog user and the relationship propensity may be determined according to the response. The questionnaire for determining the importance of acquaintance asks how often you look at 'daily life articles', 'personal photos' or how often the user writes for 'safety related articles', and the questionnaire for information-oriented disposition is 'useful information' Ask how often users see "photo of a specific topic" or "interesting and inspiring posts". In response to each question, the user can select one of the following views: "at least once a day", "at least once a week", "at least once a month", "at least once or twice", or "none" You can choose. Of course, the reference time presented in the view (day, week, month, etc.) can be changed as needed, and the number of views can also be changed. Three questions about acquaintance and information-oriented tendency are illustrated, but the number of items can be added or subtracted as needed, and the contents of the items can be different.

The survey may be performed in a manner in which the blog server 100 or the blog analysis server 130 displays a survey page on the user terminal 300 and the user responds thereto. In order to induce the user to participate in the survey, a predetermined reward may be given when the user responds to the survey. You can survey all the relationships that exist in any blog network, but you can also survey some of the relationships.

Such surveys may be optionally performed. That is, the survey is to easily collect the basic data for determining the propensity of the blog relationship, it can be omitted as necessary. Instead, the blog analysis server 130 may extract the same result as the questionnaire by recording a log of an operation of viewing or writing a post in another blog to which the blog user has a relationship.

Class labeling

In the blog analysis method according to an embodiment of the present invention, a class label which is a criterion for determining the degree of propensity of blog relationship is assigned. In order to create a classification model, we first need to determine what we want to determine. This is called a class label. Therefore, in the embodiment of the present invention, the class label is the result that is to be obtained through the survey, that is, the degree of propensity of each relationship. For example, class labels can be divided into levels such as top, middle, and bottom. However, the class label can be set to any number of levels as needed. That is, the degree of inclination can be divided into two levels or four levels or more, not three levels of upper and lower sides. The name of the class label can also be arbitrarily determined.

The survey calculates the propensity score of the blog relationship based on the answer selected by the user, and accordingly, makes the blog relationship belong to one of the class labels. Since two blog users are involved in one relationship, the survey results of the two users are summed and averaged to calculate the propensity score of one blog relationship. Then, the calculated propensity score is converted into a class label.

The acquaintance-oriented propensity score (PE _R ) and the information-oriented propensity score (PE _I ) are calculated as in Equations 1 and 2, respectively.

Where i is the item number, Wi is the score of the i-th item, N (Q _R ) is the number of items with respect to intimacy, and N (Q _I ) is the number of items with information-oriented propensity.

For example, in the survey shown in FIG. 6, the score given to each view is 30 points at least once a day, 15 points at least once a week, and at least once a month. Five points, one or more times or more, can be assigned to one point and zero to none. You can, of course, assign different scores to the view of each question. For each item, the scores of the propensity scores are calculated by adding the scores assigned to the view selected by the user. A blog user answers all three questions about friendliness at all times, and a blog user answers all three questions about friendliness at least once a month. The acquaintance propensity score is 52.5, the average of 90 (30 × 3) and 15 (5 × 3) points. For example, if the propensity score of the blog relationship is 45 points or more, the degree of propensity of the blog relationship can be set to 'high', less than 45 points and 15 points or more, 'medium', and less than 15 points, 'low'. Thus, in this example, the degree of acquaintance-oriented tendency is 'phase'.

Create Relationship Attributes for Classification

The blog analysis method according to an embodiment of the present invention creates a classification model using decision trees, which is one of classification methods of data mining. Of course, in addition to decision trees, classification models can also be created using other classification techniques such as neural networks, Bayesian classification, and genetic algorithms.

In the blog analysis method according to an exemplary embodiment of the present invention, an attribute that is a criterion for decision making is generated in a decision tree. Such an attribute may be generated based on all blog data including historical information and user information related to the blog. In other words, the attributes need to be selected to describe the characteristics of the relationship, and examples of such attributes include the number of comments, the number of gifts, the number of posts left in the guestbook, the number of scraps, the number of visits, and the like. . Here, each number is the sum of the numbers exchanged between two blog users forming one blog relationship.

Depending on the number of comments recorded on the blog, you can determine the degree of relationship inclination, as well as the number of visits and scraps. For example, a gift is an item that is frequently traded among people who know in a blog network, and the number of gifts sent and received may be an important criterion for determining the importance of acquaintance. In addition, people who collect information are often copying articles of interest from other people's blogs, so the number of scraps can be an important criterion for determining information-oriented propensity. The attribute may be variously selected based on the blog data, and the number may be variously set as necessary. The number of attributes can be 30, for example.

Create classification model

The blog analysis server 130 performs training by using data related to a plurality of blog relationships whose actual inclinations are suggested by the blog user as training data, and uses the learned results to propagate the newly given blog relationship. Create a classification model that can automatically determine. That is, the decision tree is generated by learning based on the class label and attribute information of the blog relationship, which has been calculated through the questionnaire. Examples of learning methods for generating decision trees include, but are not limited to, classification and regression trees (CART) algorithms, chi-squared automatic interaction detection (CHAID) algorithms, and C4.5 algorithms. It may be. Such algorithms are well known and will not be described in detail.

The blog analysis server 130 may generate two decision trees for determining an acquaintance-oriented propensity and for determining an information-oriented propensity, and FIGS. 7 and 8 show a decision tree for determining an acquaintance and an information-oriented propensity, respectively. Shows an example. This decision tree consists of a binary tree structure, in which two child nodes are separated from one node and proceed to the terminal node. Each node is given one of its attributes, and the end node is given a class label (CL) indicating the degree of propensity. The degree of propensity of any blog relationship is determined according to the condition information set in each node according to the attribute information of the blog relationship and determined by the class label CL assigned to the end node.

Referring to FIG. 7, the number of comments is given as an attribute of the first node, which is the largest criterion for classification. The propensity of blog relations goes to the left node when the number of comments exceeds 4, and to the right node when the number of comments exceeds 4. In the second node again, branching is based on the number of comments, branching is based on the number of gifts, and branching is based on the number of articles and scraps in the guestbook. If you proceed to the end in this way, the familiarity of the blog relationship becomes the class label (CL) assigned to the end.

Referring to FIG. 8, as in the decision tree for determining intimacy-oriented propensity, the decision tree for determining information-oriented propensity is also provided with the number of comments as an attribute of the first node. The propensity of blog relations is classified into a class label (CL) assigned to the end segment by branching according to the number of comments, the number of scraps, the date of blog creation, the number of gifts, the number of visits, and the degree of visit dependence.

As described above, according to the blog analysis method according to an embodiment of the present invention, two decision trees for determining the degree of acquaintance-oriented tendency and the information-oriented tendency with relations determined by the blog user through a survey are generated. If a blog relationship is given, which is not determined, the attribute information of the blog relationship can be applied to the decision tree to determine the degree of each propensity for this relationship.

Network propensity determination

The blog analysis server 130 determines the propensity of the blog network based on the degree of propensity of the blog relationship determined through the decision tree. Class labels may be divided into levels of 100, 50, and 0, which are converted into scores, respectively, instead of the upper, middle, and lower described above, and the degree of propensity of the determined blog relationship may be calculated based on the class labels. .

As a technique for calculating the degree of propensity of the blog network, the following three techniques can be cited. For example, the degree of propensity of the blog network B1 and B2 shown in FIG. 9 is determined. Referring to FIG. 9, in the blog network B1 and B2 illustrated in the graph, each vertex means a blog and an edge means a blog relationship. It is assumed that the propensity scores of each edge of the blog network B1 are all 10, and the propensity scores of each edge of the blog network B2 are also 10.

First, the degree of propensity of blog network is calculated by summing scores of propensity scores of blog relationships included in blog network. Then, the degree of inclination of the blog network B1 becomes 60, and the degree of inclination of the blog network B2 becomes 100.

Second, as an average technique based on the number of blogs, the degree of propensity of blog networks is calculated by dividing the sum of propensity scores of blog relationships included in blog networks by the number of blogs. In FIG. 9, since the blog network B1 has 4 vertices and the sum of propensity scores of the trunks is 60, the propensity score of the blog network B1 is 15, and the blog network B2 has 10 vertices and the sum of propensity scores of the trunks is 100. The propensity score for blog network B2 is 10. The propensity score of the blog network based on the number of blogs refers to the average propensity score of one blog. The degree of inclination of the blog network according to the average technique is determined to be higher as there are fewer peaks and more edges in the blog network.

The acknowledgment-oriented propensity score (PN _R ) and the information-oriented propensity score (PN _I ) of the blog network according to the average technique based on the number of blogs may be expressed as Equation 3 and Equation 4, respectively.

Where Nv is the number of blogs, PEr is the relationship-oriented propensity score of the relationship, and PEi is the information-oriented propensity score of the relationship.

Third, as an average technique based on the number of blog relationships, the propensity of the blog network is calculated by dividing the sum of the propensity scores of the blog relationships included in the blog network by the number of blog relationships. In FIG. 9, since the blog network B1 has 6 edges and the sum of propensity scores of the trunks is 60, the propensity score is 10, and the blog network B2 has 10 propensity scores and the propensity score of the trunks is 100, so the propensity score is 100. Becomes 10. The propensity score of the blog network based on the number of relationships means the average propensity score of one blog relationship. The degree of inclination of the blog network according to the average technique is determined to be higher as there are more relationships with high propensity scores in the blog network.

The intimacy-oriented propensity score (PN _R ) and the information-oriented propensity score (PN _I ) of the blog network according to the average technique based on the number of relationships may be expressed as Equation 5 and Equation 6, respectively.

Where Ne is the number of relationships, PEr is the relationship-oriented propensity score of the relationship, and PEi is the information-oriented propensity score of the relationship.

In the first example, the degree of propensity depends on the size of the blog network. In other words, since the number of edges of the blog network B1 and B2 is the same, but the number of edges of the blog network B2 is larger than the number of edges of the blog network B1, the decision of the blog network tendency is simply the sum of the relationship propensity scores. It may be determined that the network B1 is lower in inclination than the blog network B2. However, according to the average technique based on the number of blogs and the number of relationships that determine the size of the blog network, as in the second and third examples, the degree of network propensity does not depend on the size of the blog network.

Which of these techniques will be used depends on the nature of the application. For example, an average technique based on the number of blogs can be used in a strategy to activate blog networks using cyber items. Cyber items are often used in blog networks, where there are many relationships between blogs and high relationships. Therefore, the blog network activation strategy using cyber items can be applied to the blog network whose propensity score is calculated by the average technique based on the number of blogs. On the other hand, the average technique based on the number of relationships can be used in the strategy to activate the blog network using informational content. Informational content should be exposed to many people, and relatively many people are active in a blog network with a large number of blogs relative to the number of relationships. Therefore, the blog network activation strategy using informational content can be applied to the blog network whose propensity score is calculated by the average method based on the number of relationships.

Next, how accurate the degree of inclination of the blog relationship and the degree of inclination of the blog network determined by the blog analysis method according to an embodiment of the present invention will be described with reference to the experimental results.

First, a survey was conducted to determine the degree of propensity of blog relations and a decision tree was generated based on the results. The survey subjects randomly selected 35,000 relationships that had been sent and received during the survey. The survey was conducted directly on the blog, which resulted in 1,408 valid relationships. And various experiments were carried out using actual blog data.

First, the accuracy verification of the propensity determination of a blog relationship is demonstrated. The accuracy of the relationship propensity determination was measured in two ways as follows. First, the hit ratio, which is the agreement between the degree of propensity obtained by the generated decision tree and the degree of actual propensity obtained through the survey, was measured. Second, the error between the degree of propensity obtained by the decision tree and the degree of actual propensity obtained through the survey was measured by the difference in the level given to the class label. Measuring the error in this way can distinguish between one level being different, two levels being different, and how much level difference exists when the degree of inclination does not match.

Hit rate

As mentioned earlier, the hit ratio is a calculation of the correspondence between the degree of propensity obtained by the classification model and the actual degree of propensity obtained through the survey. Table 1 shows the results of measuring the hit ratio of blog relationships. The number of levels indicates the number of inclinations divided by how many, for example, the level 3 is the degree of inclination divided by the upper, middle, and lower three levels, and the level number 9 is the degree of inclination divided by the nine levels. . As shown in Table 1, both the acquaintance-oriented tendency and the information-oriented tendency showed high hit rates of 91% or more regardless of the number of levels. As the number of levels increases, the number of class labels that a relationship can have increases, so the hit rate tends to decrease slightly. As described above, it can be seen that the degree of propensity of blog relations can be determined with high accuracy.

error

The error is the difference between the degree of propensity obtained by the classification model and the actual degree of propensity obtained from the survey. Table 2 shows the result of measuring the error of blog relationship. At levels 3, the errors were 0.07 and 0.08, and at levels 9 the errors were 0.22 and 0.23, indicating that the difference between the actual propensity and the propensity suggested by the classification model was ± 0.22 to 0.23. it means. When the number of levels is 3, the maximum error is 2, but when the number of levels is 9, the maximum error may be 8, so the error tends to increase little by little as the number of levels increases. As described above, it can be seen that the degree of propensity of the blog relationship determined according to the embodiment of the present invention has a small error.

Next, the accuracy of the propensity of the blog network determined according to the embodiment of the present invention will be described with reference to FIG. 10.

10 is an exemplary diagram of a blog network extracted for the accuracy measurement of the propensity determination of the blog network.

Since the actual propensity of a given blog network cannot be determined by surveys, it is difficult to measure accuracy directly. Accordingly, the propensity of the propensity of the propensity of the blog network determined according to the embodiment of the present invention is indirectly determined by determining the propensity for the blog network and the randomly selected blog network, which are considered to be strong in one side, and comparing the results with each other. Let's do it.

The following method was used to extract the blog network with strong informational tendency. First, select a blog that has written a lot of scraps by other users. From the blogs that are related to the selected blog, we extract the blog that has scraped the articles of the selected blog. From the extracted blogs, we select 100 blogs in order of the highest number of visits to other blogs. The blog set thus selected becomes a blog network with a strong informational tendency.

And the following method was used to extract the network with strong tendency toward acquaintance. Choose a blog that has a lot of interaction with other blogs. Extract blogs that are related to the selected blog. From the extracted blogs, we select 100 blogs in order of the highest number of visits. The selected blog set becomes a network of blogs with a strong emphasis on friendship.

And we extracted 100 blogs related to this blog around randomly selected blogs.

The blog network thus extracted is schematically illustrated in FIG. 10. In FIG. 10, each arrow means a direction in which a relationship is formed, and an arrow means one side and a neighbor means a neighbor. Let's call a network that has a strong relationship with friends, a network that has a strong relationship with information, a network that has a strong relationship with information, and a randomly extracted blog network. 10 (a) shows an acquaintance-oriented network, (b) an information-oriented network, and (c) a random network.

The scores of the information-oriented network and the random network, and the relation-oriented network and the random network were calculated and compared. As a result, the blog network with strong specific tendency is higher than the randomly extracted blog network.

Table 3 shows the results of comparing the propensity scores of acquaintance-oriented networks and random networks. In the network of intimacy-oriented networks, the tendency obtained by the two average techniques was higher than that of the random network. The difference in values was 1.19 for the average over the number of blogs and 0.27 for the average over the number of relationships, indicating that intimacy networks had 37% and 20% higher values for each average technique than random networks. Therefore, according to the embodiment of the present invention, it was proved that calculating the degree of acquaintance of the blog network can yield a valid result.In the case of acquaintance-oriented propensity, both the average based on the number of blogs and the average based on the number of relationships can be applied. have.

Table 4 shows the result of comparing the propensity scores of the information-oriented networks and the random networks. In the information-oriented network, the average propensity score based on the number of relationships was 0.29 (21%) higher than the random network. However, the average propensity score through blog number was 0.69 (27%) lower than that of random network. The information-oriented tendency is to collect and transmit information, so blog users do not have a duplicate relationship with blogs having the same information. Moreover, information-oriented networks are mainly formed as stars around blogs with a lot of information, and if they have the same number of blogs, they tend to have fewer relationships than random networks. Therefore, it is better to use the average method based on the number of relationships to determine the degree of propensity of information-oriented networks. As a result, it was proved that in the case of the information propensity determination, calculating the degree of information propensity of the blog network based on the number of relationships can yield a valid result.

Embodiments of the invention include a computer readable medium containing program instructions for performing various computer-implemented operations. This medium records a program or process for executing a method of analyzing a blog as described so far. The media may include, alone or in combination with the program instructions, data files, data structures, and the like. Examples of such media include magnetic media such as hard disks, floppy disks and magnetic tape, optical recording media such as CDs and DVDs, floppy disks and program commands such as magnetic-optical media, ROM, RAM, flash memory, and the like. Hardware devices configured to store and perform such operations. Alternatively, the medium may be a transmission medium such as an optical or metal wire, a waveguide, or the like including a carrier wave for transmitting a signal specifying a program command, a data structure, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

Although the preferred embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements of those skilled in the art using the basic concepts of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

As described above, according to the present invention, the degree of propensity of blog relations can be determined based on the classification model, and the degree of propensity of blog connection network can be determined based on the determined degree of propensity of blog relations. The degree of each propensity of the blog network can be used as an important factor in determining the activation policy for the blog network.

Claims (19)

Determining propensity of the plurality of first blog relationships based on the classification model, and Determining a propensity of a blog network based on the propensity of the first blog relationship Blog analysis method comprising a. In claim 1, Asking a blog user of a plurality of second blog relationships related to the inclination and receiving a response, Learning based on the response and attribute information of the second blog relationship, and Generating the classification model according to the result of the learning step Blog analysis method comprising more. In claim 2, And assigning a class label divided into a plurality of levels to the propensity of the second blog relationship. And determining the first blog relationship propensity to map the score assigned to the level to the propensity score of the first blog relationship. In claim 2, The attribute information includes at least one of the number of comments, the number of gifts, the number of articles in the guest book, the number of scraps and the number of visits. In claim 2, The classification model is a blog analysis method generated using any one of a decision tree, neural network, Bayesian classification, and genetic algorithm. In claim 1, And determining the propensity score of the first blog relationship by applying attribute information of the first blog relationship to the classification model. In claim 1, The determining of the blog network propensity comprises: calculating the propensity score of the blog network by adding up the propensity scores of the first blog relationship and dividing the number of blogs included in the blog network and the number of the first blog relationships. Blog analysis method comprising a. In claim 1, The propensity is blog analysis method comprising at least one of the relationship-oriented tendency and information-oriented propensity. Asking a blog user of a plurality of first blog relationships related to the propensity of the first blog relationship and receiving a response; Learning based on the response and the attribute information of the first blog relationship, Generating a classification model according to the result of the learning step, and Determining a propensity of a plurality of second blog relationships based on the classification model Blog analysis method comprising a. In claim 9, And assigning a class label divided into a plurality of levels to the propensity of the first blog relationship. The second blog relationship propensity determining step includes mapping a score assigned to the level to a propensity score of the second blog relationship. In claim 9, And determining the propensity score of the second blog relationship by applying attribute information of the second blog relationship to the classification model. In claim 9, The propensity is blog analysis method comprising at least one of the relationship-oriented tendency and information-oriented propensity. A computer readable medium having recorded thereon a program for causing a computer to execute the method of claim 1. A relationship propensity determination module for determining propensities of the plurality of first blog relationships based on the classification model, and Network propensity determination module for determining the propensity of the blog network based on the propensity of the first blog relationship Blog analysis system comprising a. The method of claim 14, Receiving a response according to a query related to the propensity from a plurality of blog users of a plurality of second blog relationships, learning based on the response and attribute information of the second blog relationship, and classifying to generate the classification model according to the learning result. Blog analysis system further comprising a model generation module. The method of claim 15, The classification model generation module assigns a class label divided into a plurality of levels to the propensity of the second blog relationship. The relationship propensity determination module is configured to map a score assigned to the level to a propensity score of the first blog relationship. Blog Analytics System. The method of claim 14, And the relationship propensity determining module calculates a propensity score of the first blog relationship by applying attribute information of the first blog relationship to the classification model. The method of claim 14, The network tendency determination module adds the propensity scores of the first blog relations and divides one of the number of blogs included in the blog network and the number of first blog relations to calculate a propensity score of the blog network. system. The method of claim 14, The propensity to the blog analysis system comprising at least one of the relationship-oriented tendency and information-oriented propensity.

Images